Transactions Journal of

of the ASME Heat Transfer

EDITORIAL STAFF Published Quarterly by The American Society of Mechanical Engineers
Editor, J. J. JAKLITSCH, JR. VOLUME 103 • NUMBER 3 • AUGUST 1981
Production Editor,
ANNOUNCEMENTS
STELLA ROBINSON
452 Second National Symposium on Numerical Methods in Heat Transfer
Editorial Prod. Asst.,
BETH DARCHI 491 HTD Best Paper of the Year
507 Nominations for Heat Transfer Division Distinguished Service Award
HEAT TRANSFER DIVISION
545 Mandatory excess-page charges
Chairman, A. S. RATHBUN, JR. 590 Change of address form for subscribers
Secretary, F. A. KULACKI
Senior Technical Editor, K. T. YANG TECHNICAL PAPERS
Technical Editor, M. EPSTEIN
409 Pioneering Contributions to Our Knowledge of Convective Heat Transfer: One Hundred Years of
Technical Editor, G. M. FAETH
Heat Transfer Research (80-HT-137)
Technical Editor, R. H. PLETCHER
E. R. G. Eckert
Technical Editor, V. E. SCHROCK
415 Effect of Plenum Length and Diameter on Turbulent Heat Transfer in a Downstream Tube and on
Technical Editor, R. K. SHAH
Plenum-Related Pressure Losses
Technical Editor, R. SIEGEL
S. C. Lau, E. M. Sparrow, and J. W. Ramsey
Technical Editor, R. VISKANTA
423 Analytic Prediction of the Friction Factor for Turbulent Flow in Internally Finned Channefs
M. J. Scott and R. L. Webb
BOARD ON Heat Transfer in a Capillary Flow Emerging from a Reservoir
429
COMMUNICATIONS E. Papoutsakis and D. Ramkrishna
Chairman and Vice-President 436 Effects of Axial Conduction and Convective Boundary Conditions in Slug Flow inside a Circular
M. J. RABINS Tube
B. Vlck and M. N. Ozisik
Members-at-Large
441 Heat Transfer from a Plate Elevated above a Host Surface and Washed by a Separated Flow
W. BEGELL
Induced by the Elevation Step
D. BLAINE
J. CALLAHAN E. M. Sparrow, F. Sarnie, and S. C. Lau
D. KOENIG 440 Unsteady Stagnation Point Heat Transfer with Blowing or Suction
M. KUTZ T. Sano
F. LANDIS 453 Surface Contamination, Rejuvenation, and the Reproducibility of Results in Nucleate Pool Boiling
J. W. LOCKE K. A. Joudi and D. D. James
J. E. ORTLOFF 459 Destabllzation of Film Boiling Due to Arrival of a Pressure Shock- -Part I: Experimental
C. PHILLIPS A. Inoue and S. G. Bankoff
K. REID 465 Destabilization of Film Boiling Due to Arrival of a Pressure Shock—Part II: Analytical
A. Inoue, A. Ganguli, and S. G. Bankoff
Business Staff
472 An Experimental Investigation of Boiling Heat Transfer of Flourocarbon R-11 Refrigerant for
345 E. 47th St.
Concentric-Tube Thermosyphon
New York, N. Y. 10017
N. Seki, S. Fukusako, and K. Koguchi
(212) 644-7789
Mng. Dir., Publ., J. J. FREY 478 Boiling Burnout During Crossflow over Cylinders, beyond the Influence of Gravity
M. Z. Hasan, M. M. Hasan, R. Eichhorn, and J. H. Lienhard
OFFICERS OF THE ASME 485 Density-Wave Oscillations in Sodium Heated Once-Through Steam Generator Tubes
President, ROBERT B. GAITHER
H. C. Una!
Treasurer, ROBERT A. BENNETT 492
Application of Mixing Length Theory to Wavy Turbulent Liquid—Gas Interface
S. Levy and J. M. Healzer
501
A Numerical Solution to Dropwise Evaporation
508 J. J. Rizza
Journal of Heat Transfer (ISSN 0022-1481) is Cooling of Freely Falling Liquid Drops with a Shell of an Immiscible Volatile Liquid
edited and published quarterly at the offices of
rhe American Society of Mechanical Engineers. United 514 Y. H. Mori, K. Nagai, H. Funaba, and K. Komotori
Engineering Center, 345 E. 47th St., New York, N. Y.
10017. A S M E - T W X No. 7 1 0 - 5 8 1 - 5 2 6 7 . New York
A Three-Dimensional Air-Vapor-Droplet Local Interaction Model for Spray Units
Second-class postage paid at New York. N. Y.. and at 522 S. J . . Palaszewski, L. M. Jiji, and S. Weinbaum
additional mailing offices.
CHANGES OF ADDRESS must be received at Society Natural Convection from a Horizontal Cylinder—Laminar Regime
headquarters seven weeks before they are to be 528 B. Farouk and S.i. Giiceri
effective. Please send old label and new address. Transient Natural Convection Heat Transler of Water with Maximum Density Effect and
PRICES: To members. $30.00, annually: to Supercooling (80-HT-74)
nonmembers, $60.00. Single copies. $20.00 each. 535
Add $5.00 for postage to countries outside the L. Roblllard and P. Vasseur
United States and Canada. Flow Structure with Natural Convection in Inclined Air-Filled Enclosures
STATEMENT from By-Laws. The Society shall not be 540 S. J. M. Linthorst, V. M. M. Schlnkel, and C. J. Hoogendoorn
responsible for statements or opinions advanced in
sapors or . . . printed in its publications (B7.1, para. 3). Mixed Convection Heat Transfer from Horizontal Cylinders in the Crossflow of a Finite Water
COPYRIGHT © 1981 by the American Society of Layer
Mechanical Engineers. Reprints from this publication 546 W. D. Bennon and F. P. Incropera
may be made on condition that full credit be given the
TRANSACTIONS OF THE ASME, A Horizontal Flow Past a Partially Heated, Infinite Vertical Cylinder
JOURNAL OF HEAT TRANSFER, and the author,
552
L. S. Yao and F. F. Chen

I
and date of publication be stated.
INDEXED by the Engineering Index, Inc. Developing Laminar Free Convection in an Open Ended Vertical Annulus with a Rotating Inner
Cylinder
M. A. I. El-Shaarawi and A. Sarhan
{Contents continued on page 458)

Downloaded 04 Feb 2011 to 194.85.80.107. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 559 Laminar Mixed Convection in a Shrouded Fin Array
S. Aclrarya and S. V. Patankar
586 Combined Free and Forced Convection in Vertical Tubes with Radial Internal Fins
C. Prakasfi and S. V. Pataakar
573 Sonie Fundarnentai Relationships for Tubular Heat Exchanger Thermal Performance
I(. P. Singh

579 A General Theory of Wet Surface Heat Exchangers and its Application to Regen@ratlve
Evaporative Cooling
I.L. Maclaine-cross and P.J. Banks
586 A Paraaretric Study sf a Particulate Director Contact Heat Exchanger (78-WA/WI-16)
R. Letari
591 Corrduction of Heat across Rectangular Cellular Enclosures
J. Eftekhar, G . Darkazalll, and A. Haji-Sheikh

TECHNICAL NOTES
A Closed-form Analytical Solution for Freezing Adjacent to a Plane Wall Cooled by Forced
Convection
C. F. Hsu and E. M. Sparrow
Phase Change around a Circular Cylinder
V. J. bunardini
Heat Conduction through "'fin-Yang" Bodies
J. H. Lienhard
Yrairsient Thermal Expansion of Solids During Inert Heating, Phase Change, and Surface
Gasification
L. W. Munter
Extension of the Adams-Welty Ruid Bed Heat Transfer Model to the Packed Bed Case
R. L. Adanis
Thermal Storage Regenerator Parametersfor Almost Constant Gas isc charge Temperature
F. E. Ramie
Effect of Knudson Murs~beron Dropwise Condensation
1.1. Panaka
Eruatuni on a previously published paper by 12.
E. Wickox

Transactions of the ASME

Downloaded 04 Feb 2011 to 194.85.80.107. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 Pioneering Contributions to Our
E. R. G. Eckert
University of Minnesota,
Department of Mechanical Engineering,
Knowledge of l t | | ^ § ^ l ^ Heat
Minneapolis, Minn.
Honorary Mem. ASME
Transfer
SEP 2 ^81
One Hundred Years of Heat Transfer Research
LIBRARY
This paper describes the development of our knowledge of heat transfer during the past
100 years. The foundations to a systematic and coherent body of knowledge were laid in
1880-1930. The contributions were applied to a variety of situations as they were encoun-
tered in new technical developments in 1930-1980.

Introduction
This year the American Society of Mechanical Engineers celebrates
its 100th birthday and the Heat Transfer Division is pleased to join
in this celebration. It is fitting at this occasion to take a look at the
development of our knowledge of heat transfer during the past 100
years. The field of convective heat transfer developed especially
rapidly during that period and will, therefore, be discussed in this
paper.
It appears that the 100-year period can be divided into two equal
parts. The foundations for a systematic and coherent body of
knowledge were laid in the years 1880 to 1930 by the efforts of scien-
tists and engineers in a number of countries and especially by the
pioneering contributions of a few of them. The insight gained through
these contributions were applied to an ever-widening variety of sit-
uations as they were encountered in new technical developments in
the years 1930 to 1980. The present lecture will primarily dwell on
early pioneering contributions which created the foundation on which
those working in the field of heat transfer still build their research.
Fig. 1 Osborn Reynolds' demonstration setup
1880-1930: Fluid Mechanics
Convective heat transfer is intimately connected with fluid me-
chanics and it is interesting to realize that our understanding of fluid sure that any disturbance in the water had died down. Sketches of
flow and, therefore, also of convective heat transfer has made a huge what he observed are again taken from his paper as Fig. 2. Figure 2(a)
step forward 100 years ago. The fundamental laws of mechanics were, shows that, at slow flow rate, the streak of colored water was com-
of course, already known since Isaac Newton and have been formu- pletely straight and steady, indicating an ordered laminar flow. He
lated for fluid flow in the Navier-Stokes equations. However, solutions observed that this persisted as the flow rate through the tube was
to these equations could only be obtained for a few simple situations increased until a certain value was reached where the colored streak
and even there they often did not agree with experimental results, for changed its appearance quite suddenly to that shown in Fig. 2(b). This
instance, with those describing the pressure drop connected with fluid indicates that the color band started fluctuating and dispersing into
flow through pipes. Neither did many of the experimental results the main flow. Reynolds also observed this flow with illumination by
correlate among themselves. an electric spark. Figure 2(c) indicates that, with such an instanta-
neous illumination, waves and vortices are observed downstream of
This was changed by a study by Osborne Reynolds, the results of
the location where the color band becomes unstable. Reynolds called
which he reported on the 22nd of February 1880 together with a
this flow sinuous. Today we refer to it as turbulent. He repeated the
demonstration. The apparatus which he used for this purpose is shown
experiments at different temperature levels and found that the
ip a figure from his paper [1] and is reproduced here in Fig. 1. Water
transition occurred at lower flow rates as the temperature in the water
was contained in a horizontal trough with glass walls. A glass tube was
increased.
located horizontally inside the trough. It had a bell mouth on its left
hand end and was connected on its right hand to a vertical pipe. Water The fact that the viscosity of water decreases with temperature may
could thus enter the tube from the trough and its discharge could be have led him to the conclusion which he reported in a paper published
regulated by a valve. The flow rate of water was measured by a float in 1883 in the Philosophical Transactions [1]. There he argued that
on the water surface in the trough and an indicator connected to it. the characteristics of the flow are expected to be determined by the
Reynolds explained that by normal viewing we cannot understand ratio of the inertia forces to the viscous forces in the fluid and he es-
the details of the flow because water is transparent so that we cannot tablished as the ratio of these forces the expression
observe what happens inside it. He, therefore, introduced a small
pvr
amount of colored fluid into the water at the tube entrance and ob- (1)
served what he called a color band as it moved through the tube after
he had waited several hours before the start of the experiment to make with p denoting the density of the water, v the mean velocity through
the tube, r the tube radius and JX the viscosity of the water. He,
Contributed by the Heat Transfer Division. This paper is based on an ASME therefore, concluded that the transition from laminar to turbulent
Centennial Lecture given at the 19th Joint ASME/AIChE National Heat flow should occur at a certain critical value of this dimensionless
Transfer Conference, Orlando, Fla., July 27-30,1980. Revised manuscript re-
ceived by the Heat Transfer Division February 25, 1981. Paper No. 80-HT- number. In the years between 1880 and 1883, he carried out a large
137. number of experiments with tubes of different sizes and with different

Journal of Heat Transfer AUGUST 1981, VOL. 103 / 409

Copyright © 1981 by ASME
Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 =-t:~~~====
liquids and established that the transition to turbulence occurred at
a critical number between 1900 and 2000, when it is based according
to [1) on the tube radius. He also found that a dimensionless param-
eter obtained by dividing the resistance R of the flow through the tube

..·m."'1ii!E:
by pu 2 is proportional to a certain power of the parameter listed under

=-t~;;;;;;;b~_
[1) according to the expression
~~ (pvr)n-2 (2)
pv 2 J1.

~~===~~~.e~'M~29~5====
His own experiments as well as those of previous investigators cor-
related with the parameter· n = 1 for laminar flow and n = 1.723 for
turbulent flow except for some of Darcy's experiments with an in-
crusted pipe which resulted in the value n = 2. Reynolds also realized
Fig. 2 Reynolds' sketches olllow observation through his studies that the value ofthe critical parameter depends
on the type of flow and in 1884 he published a table (see Table 1), in
which he lists those factors which delay and those which advance
transition to turbulence (2).
The importance of Reynold's contribution was readily recognized
and he was invited to give a lecture on it at the Royal Institution as
early as 1884. A reproduction of his portrait painted in 1904 and in
the possession of the University of Manchester is shown as Fig. 3. The
dimensionless parameter in equation (1) is referred to as Reynolds
number with ample justification. Reynolds' studies provided the
guideline for the correlation of experimental results in all the years
up to the present and in this way advanced our understanding of fluid
flow tremendously. Figure 4, as an example, presents sketches of the
flow characteristics around a cylinder with circular cross section, in-
dicating that they depend solely on Reynolds number and, therefore,
are valid for any fluid, be it a liquid or a gas and for any cylinder di-
ameter. Flow characteristics similar to those shown in Fig. 4 for flow
around a sphere describe, for instance, the details of the flow around
a small object like a tennis ball or a golf ball in the same way as flow
around a large balloon or a storage tank for gas or liquids. Osborne
Reynolds can, therefore, certainly be considered as a pioneer in the
field of fluid mechanics and convective heat transfer.

Convective Heat Transfer

It certainly could be expected that the Reynolds number is also an
important parameter for convective heat transfer processes. However,
here the situation is more involved. Local temperature differences
necessarily exist in the flow field and the fact that the properties in-
fluencing heat transfer depend on temperature complicates the sit-
Flg.3 Portrait 01 Osborn Reynolds (1904) uation. In addition, investigators probably were aware of the fact that
experimental results were found to depend on the nature of the fluid
beyond the influence expressed in the Reynolds number. The situa-
tion in convective heat transfer was, therefore, for many years similar
to the one existing before 1880 in fluid mechanics. More and more
experimental results became available but no general correlations
could be obtained. This was changed by a paper which Wilhelm
Nusselt published in 1915 and which was preceded by another paper
in 1909. In both of them he applied dimensional analysis to the flow
Re<IO 10<Re<60 60< Re< 5000 5000< Re and energy transport equations.
Fig. 4 Flow around a cylinder
In the paper published in 1909 (3), he reports the results of an an-
alytical and experimental study of heat transfer caused by turbulent
flow through a tube. He used a heat transfer coefficient, h, already
introduced by Newton and defined by the equation
Table 1 Q =hAO w (3)
Circumstances conducive to in which Q denotes the heat flux through the area A and Ow is a
Direct or steady motion: Sinuous or unsteady motion: characteristic temperature difference between the tube wall and the
1 Viscosity or fluid 5 Particular variation of velocity across
friction which the stream, as when a stream flows fluid. He then assumed that the heat transfer coefficient can be ex-
continually destroys through still water pressed as a product of the influencing parameters, each raised to a
disturbances (treacle characteristic power. Assuming additionally that the properties on
is steadier than which the heat transfer depends can be considered as constants, he
water)
2 A free surface 6 Solid boundary walls derived the following equation by dimensionalanalysis
3 Converging solid 7 Diverging solid boundaries
boundaries (4)
4 Curverature with the 8 Curvature with the velocity greatest
velocity greatest on on the inside
the outside with C denoting a constant, k the thermal conductivity, c the specific

410 / VOL. 103, AUGUST 1981 Transactions of the ASME

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 heat of the fluid, and d a characteristic length. The other parameters
have been defined before. He concludes from the energy equation that
the two exponents, m and n, should have the same value. From ex-
periments for turbulent heat transfer in a tube which he conducted
with air, CO 2 , and city gas and from experiments by M. Jakob with
superheated steam he obtained the value 0.786 for m and 0.85 for n
and he concluded that these values are sufficiently close to verify the
conclusion m = n, and thus he derived the final relation

h= c k; (P~~t (5)

Today we know that his value for n is too high, and that his con-
clusion, m = n, is not justified. He worried about the fact that the
properties in this relation vary actually with temperature and pro-
posed as an approximation to introduce the thermal conductivity at
the wall temperature, where it is indicated by the subscript w, as well
as the other properties at an average temperature in the bracketed
parameter.
His second paper entitled "The Basic Law of Heat Transfer" [4]
and published in 1915, is the more important one. In it he applies
dimensional analysis to natural convection heat transfer, dropping
the restriction that the heat flux is expressed as a product of the pa-
rameters raised to some power. He still considers the properties as
constant with the exception of the density, the variation of which is
expressed by a thermal expansion coefficient (Boussinesq assump- Flg.5 Wilhelm Nussell (-1940)
tion), and derives the following equation for the heat flow, Q, per unit
time
and the expression for forced convection can thus be written as
(6) Nu = f(Re, PI') (10)

The gravitational acceleration, g, and the thermal expansion coeffi- and that for natural convection
cient f3 appear in this equation in addition to the parameters defined
before; d denotes a characteristic length.
Nu = [(Gr, PI') (11)
This time he extends the thermal analysis to a gas in which the Heat transfer information has, in many cases, to be obtained by ex-
properties depend in the following way on the absolute temperature, periments and presentation of the results in the form of equations (10)
l' and (11) is the means to unify and generalize them.
It is my opinion that the most important step in the analysis by
p = Po
TO)
(T' IJ, = IJ,O
(T)a
To Nusselt was the introduction of an ideal constant property fluid. Only
in this way could he get to the simple general equations (10) and (11).

= ko (l~r cp = (~t
The equations describing heat transfer in dimensionless parameters
k cpo (7)
become more involved as soon as one attempts to consider the tem-
perature dependence of the properties. Equation (8), derived by
The index, 0, refers to a characteristic fluid temperature. He derives
Nusselt, actually holds only for those gases which have the same value
the following result for the heat transfer coefficient, h
of the exponents, a, b, and c. The equation changes to the form

h= '.!.. </J (d 3p2g, l:-, l'w) (8) Nu = f (Gr, PI', Tw

To' a, b, c ) (I2)
d IJ,2 IJ,c p To
It took a considerable number of years befOl'e the relations derived when it is used to describe generally heat transfer in gases the prop-
by Nusselt were generally accepted. This may have been caused by erties of which vary according to the relations (7). This is evidence for
the fact that heat transfer processes in engineering applications op- the fact that dimensionless relations describing heat transfer become
erate in general with fairly large temperature differences whereas the the more restricted the more accurately one tries to describe the fluid
derivation of equations (4) and (6) hold strictly for a constant property properties. The large value of relations of the form of equations (10)
fluid or for vanishingly small temperature difference. A book on heat and (11) lies in their generality. They hold for groups of fluids with
transfer by A. Schack [5], the second edition of which appeared in the same Prandtl number. They are exact in the limit of small tem-
1940, devotes a considerable portion to an attempted proof that di- perature difference and experience has shown that they describe with
mensional analysis does not work and the criticism raised in re~ent fair approximation conditions at larger temperature differences when
years against dimensionless relations of the type presented under the average values for the fluid properties are introduced. They are, today,
heading "The New Heat Transfer" [6] reflects the same objection. found in any textbook, and Wilhelm Nusselt, therefore, can be con-
Relations of the form as derived by Nusselt, however, have in the sidered as another pioneer in heat transfer (Fig. 5).
meantime been widely accepted and are, today, a tool used generally
by those working in heat transfer. The following names have been Heat/Mass Transfer
given to the dimensionless parameters appearing in the equations An important relation between heat transfer and mass transfer was
derived by W. K. Lewis in 1922 [7]. The mass transfer process which
hd he investigated was evaporation of a liquid into a gas stream. He de-
Nusselt number Nu = -
k fined a mass transfer coefficient, h m , by an equation analogous to
p
equation (3)
Prandtl number PI' = IJ,C (9)
k (I3)
gfJp 2d 3()
Grashof number Gr =- - - with W denoting the mass flux of vapor away from a liquid surface
IJ,2
of area A, c denotes the vapor concentration (mass per unit volume

Journal of Heat Transfer AUGUST 1981, VOL. 103 / 411

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 of the mixture), and'the indices, wand 0, indicate that the concen-
tration should be introduced in equation (13) at the liquid surface and
at a reference point in the gas stream,respectively. He established
the relation
h
hm = - (14)
pCp

between the mass transfer coefficient and the heat transfer coeffi-
cient.
In 1929 E. Schmidt presented a paper entitled "Evaporation and
Heat Transfer" [8] in which he investigated the restrictions under
which relation (14) is valid and in which he established the relations
for the heat and mass transfer analogy by a dimensional analysis of
the differential equations describing conservation of mass, momen-
tum, and energy. He considered forced convection as well as natural
convection. The thermodynamic and transport properties had again
to be considered as constants. The main results reported in the paper
by Schmidt can be formulated as the following rule for conversion of
heat transfer relations to those describing mass transfer with the
following dimensionless parameters
hmL
Sherwood number Sh = - - (15)
D
Schmidt number Sc = !!... (16)
D
The symbol D denotes the mass diffusion coefficient, L, characteristic
length. Flg.6 Ernst SchmIdt (-1942)
We will assume that a relation of the form (10) is available to de-
In a paper published independently in 1930, Nusselt came to the
scribe a heat transfer process in forced convection. The Nusselt
same conclusions as Schmidt. He pointed out additionally that, in the
number in the heat transfer relation has then simply to be replaced
dimensional analysis, one has to consider the boundary conditions
by the Sherwood number and the Prandtl number by the Schmidt
in addition to the differential equations. The boundary condition in
number for an analogous mass trasnfer process. This procedure is
evaporation of vapor from a liquid surface into a gas stream, however,
indicated in the following lines
is different from the boundary condition describing heat transfer on
NY. = </J(Re, Fr) a solid surface insofar as a convective mass flow occurs at the evapo-
(17) rating, liquid surface whereas such a mass flow is absent at the heat
Sh Sc
transfer surface. In the meantime, transpiration cooling has found
The rule gives the correct relations for the limiting situation in which
attention in the form of a cool gas ejected through a porous wall into
the differences in temperature and in concentration are vanishingly
a hot gas stream. In this case, a convective flow occurs at the porous
small, and it is valid for independent analogous heat and mass transfer
surface and similarity in the boundary conditions is established to
situations as well as for a combined heat and mass transfer process.
the mass transfer process described above.
Schmidt ,also showed that the Lewis relation (14) follows from di-
The analogy between heat and mass transfer has in the meantime
mensional analysis for the condition Sc = Pl'.
found wide application because more information is available on heat
Two situations are treated separately in considering natural con-
transfer than on mass transfer. It is also used in reverse direction to
vection and the discussion is restricted to gases. In the first one, the
'obtain heat transfer relations through mass transfer experiments
analogy is established between a heat transfer process without mass
which, in many situations, can be carried out more conveniently and
transfer and an isothermal mass transfer process. Natural convection
with more accuracy than heat transfer experiments. The main con-
flow in the second process is generated by local differences in the
tribution to this analogy is, in my opinion, the pioneer paper by E.
density or the molecular weight, M, of the two components involved
Schmidt; a photograph of him is shown in Fig. 6.
in mass transfer. Accordingly, a new mass transfer Grashofnumber
is defined by the equation
Boundary Layers
Grm = gL3 (MM o _ 1) (18) Dimensional analysis provides the parameters with which experi-
/I w mental results can be generalized and it is very useful in its way; but
The process through which the heat transfer relation can be trans- it does not lead to an understanding of the detailed mechanisms and
formed to a mass transfer relation for a analogous process is sketched interactions of flow and heat transfer processes. Only mathematical
in the following analysis combined with well-planned experimentation can accomplish
~ = <t> (Ck, Pi:) this. The analytical prediction of flow and heat transfer processes had
(19) to overcome two hurdles: A way had to be found to solve the Navier-
Sh Gr m Sc Stokes equations and a proper model had to be conceived which de-
If heat transfer and mass transfer occur in the same field, then the scribes the effect of turbulence.
Grashof number takes on the form Important contributions to both of these tasks are due to Ludwig
Prandtl (Fig. 7). In his famous paper entitled "The Movement of
Gr* = gL3 (MoTw _ 1) (20) Fluids with Very Small Friction" [10] and presented at the Third
/I MwT o International Congress of Mathematicians at Heidelberg in 1904, he
and the process connecting the relation describing heat transfer with argues that the influence of a small viscosity of the fluid becomes
the one for mass transfer is sketched in the following line noticeable only in regions with large transverse velocity gradients,
NY. = (Gr*, Fr, &) primarily in those close to surfaces exposed to the flow, whereas the
(21) main body of the fluid can with good approximation be considered
Sh Sc Pr as frictionless. For its analysis, well developed methods were already
which indicates that the Schmidt number and the Prandtl number available. The layers close to solid surfaces in which 'friction is im-
have to be interchanged in the process. portant have to be thin in order to produce large transverse velocity

412 / VOL. 103, AUGUST 1981 Transactions of the ASME

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 , /////,',.'//'//////// //.";-//,,///////; ////////////////~
Fig. 8 Flow separation In a boundary layer
.:

Flg.7 Ludwig Prandtl (-1938)

gradients. On this basis, Prandtl simplified the Navier-Stokes

equations to the differential equations describing laminar two-di-
mensional boundary layers. It took his student, H. Blasius, four more
years to solve the boundary layer equations for steady laminar flow
over a flat plate [11], but Prandtl presented a number of semi-quan- Fig. 9 Prandtl's water tunnel
titative flow fields obtained by quadrature. Among those was the one
reproduced in Fig. 8, which is found in every textbook on fluid flow. within the turbulent boundary layer. Prandtl's papers served as a
It shows the onset of separation of a boundary layer in a region where starting point for later developments on turbulent flow and heat
the pressure increases in flow direction. ' transfer.
The paper which contains just eight equations also presents a
number of photos obtained in a water tunnel. It is interesting to ob- 1930-1980
serve that the two contributions which advanced our understanding The second period, the years from 1930 to 1980, is characterized
of fluid flow enormously, namely the one by Reynolds on transition by a tremendous growth of research in all fields of heat transfer.
to turbulence and the one by Prandtl on boundary layers, were both Workers from more and more countries joined in this activity and
obtained by flow visualization using simple equipment. Figure 9 shows international cooperation developed strongly, furthered by Interna-
the water channel with a hand-operated paddle wheel in which tional Heat Transfer Conferences-the first of which was held in
Prandtl made his early studies. 1950-which are now organized every fourth year. More than 400
Many more solutions to Prandtl's boundary layer equations have papers were presented at the last conference, the Sixth International
been obtained in the meantime. The analysis was extended to con- Heat Transfer Conference held in Toronto, Canada. The Interna-
vective heat transfer by E. Pohlhausen, who introduced, in addition tional Journal of Heat Mass Transfer and the International Centre
to Prandtl's velocity boundary layer, a thermal boundary layer within for Heat and Mass Transfer, organized in Yugoslavia, also contributed
the temperature field adjacent to a heated surface. His paper (12) in to distribution of information on a worldwide scale.
1921 treated in this way heat transfer between a solid body and a fluid A vigorous activity developed in the United States where national
with small friction and heat conduction; and in 1930, Pohlhausen heat transfer conferences are held almost yearly with over 200 papers
solved the temperature and velocity boundary layer in free convection presented at the last one and where many other conferences provide
on a heated vertical plate in collaboration with Schmidt and Beck- sessions for the dissemination of heat transfer information. Two
mfinn [13J. names stand out among the early scholars in this development: Wil-
liam H. McAdams (Fig. 10) of the Massachusetts Institute of Tech-
Turbulence nology is the author of the book Heat Transmission [17] which first
Ludwig Prandtl also provided an important impetus to the analysis appeared in 1933. In this book McAdams collected, screened, and
of turbulent boundary layers with his mixing length hypothesis correlated the available information on heat transfer processes,
published in 1925 [14]. It served as a starting point for the develop- supplemented it by his own research, and presented through three
ment of the analysis of turbulent flow processes in the succeeding editions an up-to-date, concise, and unified picture of the state-of-
years. A contribution to the analysis ofconvective heat transfer had the-art. He created a standard text which served the heat transfer
already been provided by Osborne Reynolds in 1874 [15] in a paper community as a reference book through many years. Llewellyn M. K.
in which he postulated an analogy between turbulent transfer of Boelter (Fig. 11) started research in heat transfer in 1930 at the
momentum and heat. Ludwig Prandtl took up this idea in 1910 (16). University of California, Berkeley. He had the gift of attracting tal-
He recognized that Reynolds' analogy provides a satisfactory rela- ented students and implanting in them interest and love for engi-
tionship between momentum and heat transfer caused by turbulent neering research. His lecture notes were printed under the name Heat
convection; but for many fluids, it is not satisfactory for an analogy Transfer Notes in 1933, 1940, and 1943 and were republished in book
between viscous momentum transport and heat transfer by conduc- form in 1965 [18). His associates influenced heat transfer research
tion. To take care of this, he introduced a model with a sub-layer strongly through all the years and are still doing so.

Journal of Heat Transfer AUGUST 1981, VOL. 103 f 413

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 Fig. 11 Llewellyn M. K: Boeller (-1950)
Table 2
1. System:
Flg.10 William H. McAdams (-1950) one component: rare
two component, solid-fluid, without phase change,
Conclusion solid in bulk: tube, flat plate, cylinder, sphere
solid dispersed: x ) porous matrix, grains, fibres
Research activities in the last 50 years became so widespread that solid dispersed,x) moving: fluidized bed, dust, soot
it is impossible to discuss the important contributions in the frame two component,x) solid-fluid, with phase change
of this lecture. This growth is primarily due to the large variety of evaporation, boiling, condensation
two-component,x) fluid-fluid
parameters which determine a heat transfer process. Table 2, in which more component: x unsaturated porous media
parameters influencing a specific heat transfer process are listed, has 2. Forces
been prepared to illustrate this. Examples of such parameters are pressure
arranged in four groups starting with the system or geometry de- inertia
gravity
scribing a specific situation, followed by a second group listing the centrifugal, Coriolis force
forces which determine fluid flow, and by a third group containing electric, magnetic force
energy modes and energy transport mechanisms. The fourth group surface tension
lists the fluids the properties of which influence convective heat molecular force
3. Energy Mode and Energy Transport
transfer. Any combination using a parameter in each of the groups kinetic conduction
can determine a specific heat transfer situation and many of these chemical convection
combinations actually occur in engineering developments. In addition, internal diffusion
several of the parameters in the same group may become involved in, diffusion -thermo
4. Fluids
for instance, combined natural and forced convection. Newtonian-liquids, gases,
This raises the question of how a worker in the field can absorb, Non-Newtonian
digest, and organize the overwhelming amount of information to plasma
which he is continuously exposed. I think that the pioneer contribu- xl considered as continua or particulates
tions discussed in this paper show the way. All of them started out by 8 Schmidt, E., "Verdunstung und Warmeubergang," Gesundheitsinge-
simplifying the actual process, by neglecting parameters of minor nieur, Vol. 52, 1929, pp. 525-529.
impact, and by idealizing the fluids involved. Today electronic com- 9 Nusselt, W., "Wiirmeubergang, Diffusion und Verdunstung, Z. angnew,"
Mathematik und Mechanik, Vol. 10, 1930, pp. 105-121.
puters make it possible to describe a special heat transfer process in 10 Prandtl, L., "Grenzschichten und Widerstand" Verhandlungen des III,
great detail and accuracy, but, we should not forget to covel' in our Internationalen Mathematiker-Kongressses, Heidelberg, 1904, pp. 484-491,
research larger groups in a wide sweep and in this way to maintain a Teubner, Leipzig, 1905.
coherent and organic view of the whole field of heat transfer. 11 Blasius, H., "Grenzschichten in Flussigkeiten mit kleiner Reibung,"
Zeitschrift fuel' Mathematik und Physik, Vol. 56, 1908, pp. 1-37.
12 Pohlhausen, E., "Del' Warmeaustausch zwichen festen Karpel'll und
References Fliissigkeiten mit kleiner Reibung" Angewandte Mathematik und Mechanik
1 Reynolds, a., "On the Experimental Investigation of the Circumstances Vol. 1, 1921, p. 115.
which Determine Whether the Motion of Water Shall Be Direct or Sinuous, 13 Schmidt, E. and Beckmann, W., "Das Temperatur und Geschwindig-
and the Law of Resistance in Parallel Channels," Philosophical Transactions keitsfeld VOl' einer Warme abgebenden, senkrechten Platte bei naturlicher
of the Royal Society, Vol. 174,1883, pp. 935-982. Konvektion," Forschung im Ingenieurwesen Vol. 1, 1930, p. 391.
2 Reynolds, 0., "On the Two Manners of Motion of Water," Proceedings 14 Prandtl, L., "Dber die ausgebildete Turbulenz Zeitschr. angew,"
of the Royal Institute of Great Britain, 1884. Mathematik und Mechanik, Vol. 5, 1925, p. 136.
3 Nusselt, W., "Del' Wiirmeiibergang in Rohrleitungen," Zeitschrift Verein 15 Reynolds, 0., "On the Extent and Action of the Heating Surface of Steam
Deutscher Ingenieure, Vol. 53, 1909, pp. 1750-1808. Generators," Proceedings of Literary and Philosophical Society, Manchester,
4 Nusselt, W., "Das Grundgesetz des Warmeiiberganges," Gesundhe- Vol. 14, Session 1874-1875, pp. 81-85.
itsingenieur, Vol. 38, 1915, pp. 477-482. 16 Prandtl, L., "Eine Beziehung zwischen Warmeaustausch und
5 Schack, A., Del' IndustrieUe Warmeiibergang, 2nd ed. 1940, Stahleisen, Striimungswiderstand in Fliissigkeiten," Physik Zeitschrift Vol. 11, 1910, pp.
Diisseldorf. 1072-1078.
6 Adjutori, E. F., The New Heat Transfer, Ventuno Press, Cincinnati, 17 McAdams, W. H., Heat Transmission, McGraw-Hill, New York,
Ohio, Vol. 1, 1974. 1933.
7 Lewis, W. K., "The Evaporation of a Liquid into a Gas," Mechanical 18 Boelter, L. M. K., Heat Transfer Notes, McGraw-Hill, New York,
Engineering, Vol. 44, 1922, p. 445. 1965.

414 / VOL. 103, AUGUST 1981 Transactions of the ASME

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 Effect of Plenum Length and
S. C. Lau
E. M. Sparrow
Diameter on Turbulent Heat Transfer
in a Downstream Tube and on
Fellow ASME

J. W. Ramsey
Department ol Mechanical Engineering,
University of Minnesota,
Minneapolis, Minn 5545S
Plenum-Related Pressure Losses
A systematic experimental study was carried out to determine how the heat transfer char-
acteristics of a turbulent tube flow are affected by the length and diameter of a cylindrical
plenum chamber which deliuers fluid to the tube. The net pressure loss due to the pres-
ence of the plenum was also measured. The experimental arrangement was such that the
fluid experiences a consecutive expansion and contraction in the plenum before entering
the electrically heated test section. Air was the working fluid, and the Reynolds number
was varied over the range from 5,000 to 60,000. It was found that at axial stations in the
upstream portion of the tube, there are substantially higher heat transfer coefficients in
the presence of longer plenums. Thus, a longer plenum functions as an enhancement de-
vice. On the other hand, the plenum diameter appears to have only a minor influence in
the range investigated (i.e., plenum diameters equal to three and six times the tube diam-
eter). The fully developed Nusselt numbers are independent of the plenum length and
diameter. With longer plenums in place, the thermal entrance length showed increased
sensitivity to Reynolds number in the fully turbulent regime. The pressure loss coeffi-
cient, which compares the plenum-related pressure loss with the velocity head in the tube,
increases more or less linearly with the plenum length. With regard to experimental tech-
nique, it was demonstrated that guard heating/cooling of the electrical bus adjacent to
the tube inlet is necessary for accurate heat transfer results at low Reynolds numbers but,
although desirable, is less necessary at higher Reynolds numbers.

Introduction
Turbulent heat transfer in circular tubes has been the subject of test section tube, the investigated plenum configurations encom-
extensive study for many decades because the circular tube serves as passed the range \ < L/d < 10, respectively for D/d = 3 and 6. For
a basic component in a wide variety of heat exchange devices. In each fixed plenum configuration investigated, experiments were
general, fluid may be delivered to the tube from a broad range of conducted for Reynolds numbers ranging from 5000 to 60,000.
sources, encompassing, for example, plenums of various geometric The heated test section was heavily instrumented to enable local
forms, upstream pipelines (either heated or unheated), and fittings heat transfer coefficients to be determined and, thereby, to facilitate
of many types (e.g., tees, elbows). It may be expected that the tube identification of the effects of plenum geometry. Furthermore, the
heat transfer characteristics will be affected by the sources from which presentation of results is structured to highlight these effects. The
fluid is delivered, especially in the region of the tube adjacent to its total heat transfer presentation includes the thermal entrance region,
inlet. The present research was undertaken to experimentally de- the thermally developed regime, entrance length, and comparisons
termine how the turbulent heat transfer characteristics of a heated with the literature. With regard to fluid-flow aspects, the plenum-
tube respond to systematic variations of the length and diameter of related pressure drop is presented in terms of a dimensionless pressure
an upstream plenum which delivers fluid to the tube. The pressure loss coefficient for the same range of parameters as were employed
drop associated with the presence of the plenum was also measured, in the heat transfer experiments. The flow visualization observations
and flow visualization was employed for the observation of certain were of a qualitative nature and are reported as such.
characteristics of the flow field. The experimental apparatus was designed and fabricated to provide
In the heat transfer apparatus, a plenum chamber in the form of heat transfer results of impeccable quality which can serve both as
a circular cylinder is positioned immediately upstream of an electri- a guide to design and as a standard against which to compare candi-
cally heated tube such that the tube and the plenum share the same date analytical models. In recognition of the potential injurious role
axis. In turn, the working fluid, air, is delivered to the upstream end of extraneous heat losses in airflow systems, a number of measures
of the plenum via a tube which also shares the common axis of the were instituted to defend against such losses. These include (1)
plenum and the heated tube. Thus, the working fluid experiences a near-zero contact between the heated test section and the plenum,
consecutive expansion and contraction prior to its entry into the (2) guard heating and cooling of the bus bars which deliver electrical
heated test section tube. power to the system, (3) suspension of the test section with thin nylon
The apparatus used for the pressure-drop studies is similar to that line, and (4) use of an insulation having a thermal conductivity less
of the heat transfer experiments (and air is also the working fluid), than that of air. In a sensitivity study, it was demonstrated experi-
except that the heated test section is replaced by an unheated tube mentally that the absence of guard heating/cooling can have a sub-
instrumented with pressure taps. Flow visualization was conducted stantial impact on the results in the thermal entrance region.
in an all-plexiglass apparatus with water as the working fluid. Attention will now be turned to the literature relevant to the subject
Two plenums of different diameter D were employed during the under study. In general, a literature search failed to unearth any
experiments. Each plenum was designed and fabricated so that its published research of the type reported in the present paper. However,
streamwise length L could be varied. Relative to the diameter d of the to provide perspective, the general state of affairs with regard to up-
stream hydrodynamic conditions for turbulent tube-flow heat transfer
will be briefly reviewed. With regard to analysis, it has been customary
Contributed by the Heat Transfer Division for publication in the JOURNAL
OF HEAT TRANSFER. Manuscript received by the Heat Transfer Division to avoid involvement with the actualities of the flow field upstream
September 24,1980. of the heated tube by imposing hydrodynamic boundary conditions

Journal of Heat Transfer AUGUST 1981, VOL. 103 / 415

Copyright © 1981 by ASME
Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 at the tube inlet. Experiments have been reported in [1] and [2] for with the junctions placed in the mixing region of the incoming jet.
heated tubes to which flow is delivered via an abrupt or rounded At the downstream end of the plenum, special precautions were
contraction from a very large plenum, or via an adiabatic tube that taken to minimize possible extraneous heat conduction from the test
is either straight (hydrodynamic development length) or curved section tube to the plenum end-face. These precautions were moti-
(elbow or bend). In [2] and [3], a tee was used to convey the working vated by the concern that the plenum airflow, either impinging against
fluid to the heated tube. or recirculating adjacent to the end-face, would induce an extraneous
The heat transfer characteristics of a heated tube situated down- flow of heat out of the test section.
stream of either an abrupt expansion or an abrupt contraction of finite The first of the defenses against such extraneous losses is in the
diameter ratio were investigated in [4-8]. In all cases, the tube situated adopted structure of the end-face. The face consisted of two parts—an
upstream of the change of section was sufficiently long so that the inner and an outer annular disk, of which only the former is relevant
hydrodynamically developed conditions prevailed prior to the ex- to the heat loss issue because of its contact with the test section tube.
pansion or contraction. The case of a successive expansion and con- This disk was fabricated from rigid fiberglass circuitboard whose
traction (i.e., a plenum of finite length) upstream of a heated tube does thickness was only 0.043 cm (0.017 in.). The inner diameter of the disk
not appear to have been investigated. was made identical to the outer diameter of the test section tube, so
that the disk could just slip over the tube. They were positioned so
The Experiments that their upstream faces lay in a continuous plane. The only thermal
Heat Transfer Apparatus. The heat transfer apparatus will be path between the tube and the disk was through their 0.043 cm (0.017
described in sufficient detail to convey its main features and novel in.) interface. This is believed to be the very minimum that is con-
design aspects. A more detailed description, including design draw- sistent with structural integrity, especially when a nonmetallic, rel-
ings, is available in [9]. atively low thermal conductivity material such as fiberglass is em-
The apparatus is an open-loop airflow system which is supplied by ployed. A small fillet of silicone rubber was inserted just downstream
a dryer-equipped air compressor. Upon entering the loop, the air of the disk to seal against air leaks—other defenses against air leaks
passes through a set of control valves, a filter, and a regulator, and then are described in [9].
is ducted to a settling chamber which serves to damp the eddies in- Immediately downstream of the fiberglass disk, a copper ring was
duced by valves and elbows. From the settling chamber, the air enters soldered to the outer surface of the test section tube to serve as a
and traverses a 65-diameters-long hydrodynamic development tube, distribution bus for the electric current used for the ohmic heating
the efflux of which discharges into the plenum chamber whose di- of the tube. The ring was beveled on its upstream face in such a way
mensions (length and diameter) are to be varied throughout the ex- that it leaned away from the disk. To avoid ambiguity about the axial
periments. From the plenum, the air passes through the 76-diame- position where the ohmic heating began, the axial contact between
ters-long heated test section and the mixing chamber situated at its the tube and the bus ring was kept as short as possible (axial contact
downstream end. The discharge from the mixing chamber is ducted length ~ \ tube diameter). The ring also served as a hub against which
to a rotameter and then passes into the room. three sharp-tipped nylon screws were tightened to ensure the align-
Apparatus Components. As noted earlier, the hydrodynamic ment of the plenum and the test section tube.
development section, the plenum, and the test section all share a Electric current was supplied to the copper bus ring by six rod-like
common axis. In addition, the development and test section tubes copper conductors, each 0.476 cm ( ^ in.) in diameter and about 5.08
were cut from a single length of stainless steel tubing, the bore di- cm (2 in.) long. The rods, canted toward the downstream to avoid
ameter of which, after internal honing to attain a smooth satin finish, congestion adjacent to the plenum, emanated radially outward like
was 2.36 cm (0.930 in.). The post-honing wall thickness of the test the spokes of a wheel. The outer tip of each spoke was soldered to a
section tube (a parameter relevant to the ohmic heating) was 0.089 heavy copper ring, which served as a terminal for a stranded copper
cm (0.035 in.) with a circumferential variation of ±0.002 cm (±0.0008 cable from the power supply.
in.). The two plenum chambers used in the experiments were made Each spoke was fitted with a differential thermocuple which en-
of plexiglass, with respective internal diameters equal to 2.96 and 5.91 abled the existence of extraneous heat conduction along the spokes
times that of the test section tube (nominally, 3 and 6 times). to be detected. To null out such extraneous heat flows, the outer
To enable the streamwise length of the plenum chamber to be copper ring was fitted with a water-carrying copper tube. The water
varied during the course of the research, its upstream end-face was temperature could be precisely controlled by a thermostatically reg-
designed so that it could be moved axially within the cylindrical wall ulated heating/cooling bath. This guard heating/cooling capability
of the chamber. An air seal between the cylindrical wall and the constituted a very important defense against extraneous losses/gains
contacting edge of the end-face was achieved with a pair of O-rings. and, as will be demonstrated later, it served to insure the quality of
The end-face was rigidly attached to the downstream end of the hy- the thermal entrance region results.
drodynamic development tube, so that the axial movement of the face The test section tube itself was heavily instrumented with ther-
required a corresponding axial movement of the development tube. mocouples. In an axial length extending eight diameters from the
This movement was accommodated by a flexible coupling at the up- onset of heating, there were twelve thermocouple stations and, at each
stream end of the development tube. such station, four thermocouples were deployed at equal intervals
Two thermocouples threaded through the upstream end-face of around the circumference of the tube. Between the axial distances
the plenum served to measure the air temperature just downstream of 8 and 64 diameters from the inlet, there were 16 instrumented
of the plenum inlet. The thermocouples were positioned 180 deg apart, stations, each with two thermocouples. All thermocouples were of

.Nomenclature
Am = cross-sectional area of tube wall rh = mass flow rate of air x = axial coordinate measured from tube
cp = specific heat of air Nu = local Nusselt number, hd/k inlet
D = inside diameter of plenum chamber AP P = net plenum-related pressure drop xe = thermal entrance length
d = inside diameter of tube fj. = viscosity
Qi = rate of convective heat transfer from
h = local heat transfer coefficient, equation p = density
control volume i to the fluid
(4)
Re = Reynolds number, 4m/fnrd Subscript
K = pressure loss coefficient, equation (6)
k = thermal conductivity of air Tb = local bulk temperature
fd = thermally fully developed
kw = thermal conductivity of stainless steel T,„ = local tube wall temperature
wall T™ = temperature of surroundings Superscript
L = axial length of plenum chamber V = mean air velocity in tube * = at plenum inlet

416 / VOL. 103, AUGUST 1981 Transactions of the ASME

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 30-gage, calibrated copper and constantan wire, applied to the tube study was to detect the presence or absence of flow separation in the
via spot welding and oriented so that the lead wires lay along the ex- test section tube just downstream of the inlet cross section.
pected isothermal directions. The visualization apparatus was made entirely of plexiglass and
The final thermocouple station was situated twelve tube diameters was equipped with dye injection taps both along the cylindrical wall
from the downstream bus-bar assembly. That assembly was similar of the plenum and along the upstream portion of the test section tube.
in structure to that of the upstream bus but was not guard heated or Water was supplied to the apparatus from a constant-head tank
cooled because there was no need to do so (because it was too remote suspended from the ceiling of the laboratory. The discharge from the
from the nearest measurement station). apparatus was ducted to a weigh tank for flow metering.
The mixing chamber was mated to the downstream end of the test
section. It consisted of a length of pvc pipe which housed a series of Data Reduction
copper baffle plates. An array of holes was machined in each plate, Heat Transfer Results. The procedures employed in deducing
and the pattern and size of the holes in the respective plates were local heat transfer coefficients and Nusselt numbers will now be de-
designed to induce both radial and circumferential fluid motions. scribed. For the evaluation of the local heat transfer coefficient, the
Three thermocouples situated downstream of the last baffle were used needed inputs are the local heat flux, wall temperature, and bulk
to measure the temperature of the air. Flexible tubing served to couple temperature. Among these, only the local wall temperature is directly
the discharge end of the mixing chamber to the rotameter used for measured, so that the other quantities have to be obtained via data
flow metering. reduction.
As a further defense against extraneous heat losses (or gains), the The first step in the analysis is to envision the wall of the test section
plenum chamber, the test section, and the mixing chamber were sit- tube as being subdivided into axial segments of length Ax,. Each such
uated within an insulation enclosure. The enclosure consisted of a segment serves as a control volume for a heat balance. The axial
plywood box with a 30.5 cm (1 ft) square cross section whose inner boundaries of the respective control volumes are positioned midway
walls were lined with styrofoam sheet insulation. The styrofoam en- between the adjacent temperature measurement stations. Thus, if
closed a space having cross-sectional dimensions of 21.6 X 21.6 cm (8.5 #i-i, *i, and jcj-n denote the axial locations of successive wall tem-
X 8.5 in.). Into this space was poured silica aerogel powder insulation perature measurements, then the axial boundaries of the ith control
whose thermal conductivity is about 85 percent that of air. The test volume are at (x;_i + x,)/2 and at (x,+i + x,)/2 and, correspondingly,
section was suspended from the frame of the plywood box by mono- the control volume length Ax, = (x,+i — x,_i)/2.
filament nylon lines, and no other supports were employed. The energy transfers to and from a control volume include the
Power Supply and Instrumentation. Electric power was sup- following:
plied to the test section via a system which served to stabilize the 1 an internal heat source due to ohmic dissipation
voltage and to provide the high current and low voltage called for by 2 convective heat transfer from the inner surface of the control
the low-resistance test section. Test section voltage and current (both volume to the airflow
read as voltages) were measured by a true rms digital voltmeter 3 heat loss from the outer surface of the control volume through
(Keithly model 179). This type of meter was selected because of the insulation to the ambient air
concern that possible departures from a sinusoidal wave form would 4 conduction heat transfer into the downstream face of the control
introduce errors in the readings of other types of voltmeters. volume from the adjacent portion of the tube wall
All thermocouple emfs were read with a Fluke 2240B datalogger 5 conduction heat transfer out of the upstream face of the control
capable of programmable scanning. It provided a printed output with volume into the adjacent portion of the wall
a smallest digit corresponding to 1 /xV. The programming capabilities Under steady-state conditions, these inflows and outflows are in
of the datalogger enabled the lengthy approach of the apparatus to balance. As will now be described, numerical values can be supplied
thermal equilibrium to be monitored without human intervention. to the heat balance equation for items 1, 3, 4 and 5, enabling the de-
However, as a matter of course, the final attainment of equilibrium termination of the convective heat flow from the control volume to
and the collection of data were always monitored and controlled by the fluid (item 2).
the experimenter. Both the datalogger and the aforementioned true The ohmic dissipation for control volume i was determined by
rms voltmeter were new at the beginning of the research and were, prorating the overall dissipation by the ratio AX-JXL, where XL is the
therefore, under their calibration warranties. length of the test section tube. For the heat loss through the insulation
The rate of air flow through the apparatus was measured by one to the ambient, a conduction shape factor was found by setting up a
of three rotameters, depending on the magnitude of the flow. All ro- two-dimensional finite difference network that spanned the two types
tameters were calibrated against a secondary standard. of insulation (silica aerogel and styrofoam) and the surrounding
Pressure Measurement and Flow Visualization. A separate wooden containment structure. A total of 1600 grid points were used
set of experiments were performed to determine the pressure drop to resolve the complex geometry which included both curved and
due to the presence of the plenum chamber. For this purpose, the straight boundaries. The details of this work are presented in [9]. With
heated test section tube, its attached mixing chamber, and the insu- the shape factor available, the heat loss through the insulation was
lation enclosure were removed from the airflow loop. A stainless steel calculated by multiplying the shape factor with the temperature
tube (84 diameters long), that had been cut from the same length as difference (Tw(xi) — T„) and by the axial length Ax,, where T,„(x,)
the test section, was mated to the downstream end-face of the plenum is the measured wall temperature at x; and T„ is the ambient tem-
via a plexiglass flange (the fiberglass disk remained with the heated perature.
test section). The tube was internally honed for smoothness and was The axial conduction along the wall, items 4 and 5, was found by
equipped with pressure taps located sufficiently far downstream of applying a central difference representation of Fourier's law. If Aw
the plenum so as to lie in the redeveloped flow regime. There was also represents the cross-sectional area of the wall and kw denotes the
a series of taps deployed along the hydrodynamic development tube thermal conductivity of stainless steel (regarded as a function of
(upstream of the plenum) which were employed in the determination temperature), then items 4 and 5 may be expressed as
of the plenum-related pressure drop.
/e„,A,„(T,„(x,+i) - T,„(x,))/(x, + l - x,) (1)
The pressure signals were sensed by a Baratron solid-state capac-
itance-type pressure meter. The output of the Baratron was read with ku,Au,(Tw(xi) - Tw(xi-i))/(xi - xi-t) (2)
the aforementioned Keithly digital voltmeter. Pressure differences
as small as 10~4 mm Hg could be detected by the Baratron. It, too, was Equations (1) and (2) were evaluated using the measured wall tem-
new and under its warranty calibration during the duration of the peratures at stations x,_i, x,, and x,+ i.
experiments. The foregoing discussion has demonstrated that all contributions
The flow visualization study was made prior to the heat transfer to the energy balance for a typical tube-wall control volume i can be
and pressure drop experiment.' The specific focus of the visualization evaluated from the experimental data, aside from the convection term.

Journal of Meat Transfer AUGUST 1981, VOL. 103 / 417

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 Therefore, these operations yield the rate of convective heat transfer p* was obtained via extrapolation from the upstream pressure
Qi from the control volume to the fluid. The thus-determined Q; for data.
the successive control volumes are employed directly in the evaluation
of the heat transfer coefficient and also for the computation of the Results and Discussion
local bulk temperature, as will now be described. Basic Data. To illustrate the nature of the basic data collected
The axial distribution of the local bulk temperature was also found during the experiments, Fig. 1 has been prepared. The figure shows
from a control volume analysis. For this analysis, the control volumes wall and bulk temperature distributions for two data runs. The data
are cylinders of diameter d (equal to the tube diameter) and axial points for the wall temperature distributions represent direct mea-
length Ax(. The axial coordinates of the upstream and downstream surements. For the bulk temperature, only the inlet and exit values
faces of these control volumes are the same as those for the control were directly measured, and these are identified by blackened data
volumes used earlier in the determination of the convective heat symbols. The other plotted points for the bulk temperature distri-
transfer rates Q;. bution were computed as described in the preceding part of the paper.
For control volume i, let Tj n be the bulk temperature of the fluid Smooth curves are drawn through the data points for continuity; in
entering the control volume at its upstream face and T'oul denote the the case of the bulk temperature, the faired curve is virtually indis-
fluid bulk temperature at the downstream face. Then, tinguishable from a straight line.
The wall and bulk temperature distributions shown in the lower
TLt = T\n + Qi/mcp (3) part of the figure are typical of those for the overwhelming majority
Thus, with the already determined Q; and the measured inlet bulk of the data runs. They are, in fact, exemplary specimens of what is
temperature, equation (3) can be used in a marching fashion to expected for turbulent heat transfer in a uniformly heated tube. These
compute the fluid bulk temperature at all the control volume faces. distributions are exemplary from three standpoints: (1) they are re-
Then, the bulk temperature Tt,(xi) at any thermocouple measurement markably free of data scatter, (2) the wall and bulk temperature dis-
station x; was found by linear interpolation between the faces of the tributions are precisely parallel in the downstream (fully developed)
corresponding control volume. region, and (3) the computed bulk temperature distribution extra-
Then, with the information generated from the foregoing compu- polates almost exactly to the measured value at exit. With regard to
tations, the local heat transfer coefficient and Nusselt number at any the latter characteristic, it may be noted that in the majority of the
station i was evaluated from data runs, overall energy balances closed to within one percent.
The results shown in the upper portion of Fig. 1 are unusual in that
Qi/Aj hid the wall temperature distribution exhibits an overshoot before at-
hi = , Nu; = — (4)
Tw(xi)-Tb(xi)' ki taining the fully developed straight-line increase. This overshoot was
where A; = irdAxi and fc; is the thermal conductivity of air at T(, (*,-). encountered only at the largest Lid (=10) and for Reynolds numbers
Although hi may have the appearance of a "local average" heat of 7000 and lower. It can be attributed to the transition of the devel-
transfer coefficient, it is, for the most part, a truly local coefficient. oping boundary layer from laminar to turbulent flow.
In this regard, it can be verified by carefully re-examining the fore- To illustrate the nature of the heat transfer coefficients measured
going data reduction procedure that the only quantity that is actually here, a set of distribution curves for a given plenum geometry (D/d
related to the finite axial length of the control volume is the axial = 3, hid = 10) is plotted in Fig. 2 in the form of h versus x/d. Results
conduction in the wall. Since the net axial conduction is only of sig- are shown for nine distinct Reynolds numbers between 5200 and
nificance at the first few measurement stations, it is only there that 59,200. These results are distinguished by the general absence of
hi is a local average. At all other points, hi is strictly local. scatter and by the true constancy of h in the thermally developed
The results will be parameterized by the Reynolds number defined regime.
as The distribution curves for Re > 10,000 display the classical shape
that is characteristic of turbulent pipe flows, while the curves for Re
Re = imliiird (5) < 7000 are characterized by a slight undershoot prior to the attain-
ment of the fully developed value. This undershoot is a reflection of
where fi is the viscosity corresponding to the mean bulk temperature
the already-discussed wall temperature overshoot in Fig. 1.
of the air.
Variable fluid properties played only a minor role (e.g., maximum An important observation in Fig. 2 is the absence of any indication
wall-to-bulk temperature difference about 11°C (20°F)). Neverthe- of flow separation adjacent to the tube wall just downstream of the
less, for the reporting of fully developed Nusselt numbers, the fore-
going definitions of Nu and Re were altered to respectively include
kfd and ixfd, which correspond to the mean bulk temperature in the
fully developed regime.
Pressure Drop Results. The net pressure drop due to the pres-
ence of the plenum chamber will be reported as a dimensionless
pressure loss coefficient K whose evaluation will now be described.
First, it may be recalled that there are pressure taps in the tubes both
upstream and downstream of the plenum. These taps are so situated
that for both the upstream and downstream regions, the respective
p versus x distributions are straight lines. Envision next a graph where
the p versus x data for a given run are plotted. Within the accuracy
of the measurements, the aforementioned straight lines for the up-
stream and downstream regions are parallel but not colinear. Rather,
there is a vertical separation between them (to be denoted by AP P )
which is due entirely to the presence of the plenum.
In terms of AP P , the pressure loss coefficient is defined as

K = AP p /(ipV 2 )* (6)
The denominator of equation (6) is the velocity head corresponding
25L- 1 1 I 1 1 1 i_
to the density p* of the fluid in the tube at the inlet of the plenum. 0 10 20 30 40 50 60 70
For the evaluation of K, AP P was determined from least-squares X/d
straight lines fitted to the pressure data, and the pressure needed for
Fig. 1 Illustrative wall and bulk temperature distributions

418 / VOL. 103, AUGUST 1981 Transactions of the ASME

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 inlet. The characteristic signature signalling the presence of such a
separation region is an h versus x distribution which rises to a maxi-
mum at the point of flow reattachment and decreases thereafter.
Clearly, such a pattern is not in evidence in the present results. Fur-
thermore, the flow visualization studies did not reveal any separated
regions within the test section tube.
Another basic issue of the research is the role of guard heating/
cooling of the electrical bus adjacent to the tube inlet, and this issue
is addressed in Fig. 3. The figure contains two panels in which results
are respectively presented for a low and a high Reynolds number (Re
= 5000 and 50,000). In each panel there are three curves of Nu/Nu/y
versus x/d, where Nu and Nu/y respectively denote the local and fully
developed Nusselt numbers. These curves have been faired through
the experimental data, but the data have been omitted from the figure
to avoid confusion.
One of the curves (the solid line) corresponds to the case where the
guard heater/cooler is adjusted to null out extraneous heat transfer.
The other curves correspond to ±15 yiiV imbalances across the dif-
ferential thermocouples installed on the copper current-carrying
spokes. This imbalance is typical of that which would have occurred
in the absence of guard heating.
The results speak for themselves—that is, guard heating is strictly
necessary at low Reynolds numbers, but, while desirable, it is dis-
pensable at Re = 50,000 or greater.
Plenum-Related Heat Transfer Enhancement. Attention will
now be turned to the main focus of the research—the effect of plenum
Fig. 2 Illustrative axial distributions of the heat transfer coefficient for length and diameter on the heat transfer characteristics of a tube to
parametric values of Reynolds number, Did = 3 and Ud = 10 which fluid is delivered by the plenum. Results documenting the
plenum size effect will be presented in the next six figures, which will
now be discussed sequentially.
The effect of plenum length for a fixed plenum diameter Did = 3
is dealt with in Figs. 4 and 5. These figures encompass five individual
graphs which pertain respectively to fixed axial stations x/d = 1, 2,
3, 5, and 8. At each station, Nu/Nu/d is plotted as a function of
Reynolds number for three different plenum lengths characterized
by L/d = 1, 5, and 10. In this format, the effect of plenum length on
the local heat transfer coefficient is readily identified by comparing
the ordinates of the various L/d curves at a given Reynolds
number.
From an examination of Figs. 4 and 5, it is seen that the plenum
length has a decisive effect on the heat transfer coefficient when x/d
is small, but the extent of the effect decreases with increasing down-
stream distance. At x/d = 1 and at the larger Reynolds numbers, the
h value for L/d = 10 is about 40 percent greater than that for L/d =
1. The percentage increases are slightly smaller at lower Reynolds
numbers. Thus, a relatively long plenum serves as an enhancement
device.
At x/d = 2 and 3, heat transfer enhancement due to plenum length
Fig. 3 Effect of guard heating/cooling on the axial distribution of the Nusselt persists and is significant, but is smaller than that at x/d = 1. How-
number ever, at x/d = 5, there is no apparent difference between the results
for L/d = 5 and 10, and at this x/d the greatest enhancement in evi-
dence d u e to the plenum length is about six percent. At x/d = 8, the
2.6 - Ix/d = 11 curves tend to overlap, and while enhancement still persists at the
\ \. larger Reynolds numbers, it is only a few percent. For still larger x/d
2.2-- wx x TK 5
r-J°

1.4 ~A-,.A 5 °~ D- D-D-D D

- "A. —o—— HX1-0--0
L/d = I ~-A- L/d = I
*"--A_
-A-^--A-.AA
I.4
2.2 -
|x/d=2|
AVXI
- 1 _J0^
• ^ :
n o ^ n
~-A. —-D—- - 5 - - 0 — — M-o-—a
-A.
- L/d = I ~--A.
-A—A--A--A A
I I I I I I i I I I I
6 8 IO 20 40 60
Re x I0 J Re x 10°
Fig. 4 Effect of plenum length on the Nusselt numbers at x/d = 1 and 2 for Fig. 5 Effect of plenum length on the Nusselt numbers at x/d = 3, 5, and 8
D/d = 3 for Did = 3

Journal of Heat Transfer AUGUST 1981, VOL. 103 / 419

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 2.6,
2.61
-o- D/d= 3
2.2 o D/d = 6

2.2
1.8 ° ^^°^V^B
_i
1.8
v 1.4 —AA ~Q—Q- _D CD-^tl
^ 2.2 O o O O O O

1.8 5^

L/d = I .-O-O-Q—
1.4-
—Q—oo-cjoa
-A
- A — A . A-A
J I I l I I -J J_ J I I L
6 8 10 20 40 60
Re x I0"3
Fig. 6 Effect of plenum length on the Nusselt numbers at x/d = 1 and 2 for
D/d = 6

I x/d = 3 |
10
1.4 40 60
Re x 10"
L/d = I
Fig. 8 Effect of plenum diameter on the Nusselt numbers at x/d = 1, 2, 3,
Z I-4 Sc^; |x/d = 51 5, and 8 for Ud = 5 and 10

I.4-
Ix/d = 81
A-AA
I.O, J I III,, J I [_
4 6 8 10 20 40 60
Re x IO"3
Fig. 7 Effect of plenum length on the Nusselt numbers at x/d = 3, 5, and 8
for D/b = 6
(not shown), all three curves collapse to the line Nu/Nu/d = 1.
A presentation similar to that of Figs. 4 and 5, but for a fixed ple-
num diameter D/d = 6, is provided by Figs. 6 and 7. The format of
these figures is identical to those of the preceding figures. In actuality,
the similarities between the two sets of figures are deeper than mere
format. Aside from some minor differences in detail, the qualitative
and quantitative behavior already identified for the D/d = 3 case
appears to be reproduced for the D/d = 6 plenum.
To explore in greater detail the relationship between the results
for the two plenum diameters, Fig. 8 has been prepared. This figure
consists of two graphs, the upper of which conveys information for
a fixed plenum length L/d = 10 while the lower is for the fixed length
L/d = 5. In each graph, Nu/Nu/y is plotted versus Reynolds number
for both of the plenum diameters. The data for D/d = 3 are connected
by faired curves but those for D/d = 6 are not fitted with curves to Fig. 9 Effect of plenum length on the axial distributions of the Nusselt number
at Re = 6000, 15,000, and 50,000 for Did = 3
avoid the possible confusion of many closely positioned lines.
If attention is first given to the longest plenum L/d = 10, it is seen x/d for three different Reynolds numbers, respectively Re = 6000,
that the data at x/d = 1, 2, and 3 for the larger diameter plenum fall 15,000, and 50,000, in the three panels of the figure. Each panel con-
slightly below that for the smaller plenum, but the maximum devia- tains curves for three plenum lengths, L/d = 1, 5, and 10, and the
tions are only about five percent. At x/d = 5 and 8, the data for the objective of the figure is to show how the entirety of the h versus x
two cases are coincident. Next, turning to the results for the inter- distribution curve is affected by the length of the plenum. The figure
mediate length plenum L/d = 5 (lower graph), the same relationship is for D/d = 3, but a similar figure for D/d = 6 conveys nearly identical
between the Nusselt numbers for the two plenum diameters prevails results [9].
but the deviations are even smaller. A comparison similar to that of The figure shows that at small x/d, the curves are ordered in a
Fig. 8 was also made for the shortest plenum L/d = 1 [9], and only very regular way with L/d, with that for the longest plenum being highest.
small diameter-related deviations were observed. However, the larger the L/d, the more rapidly do the curves drop off,
If consideration is now given to the findings of Figs. 4-8, it may be • and this causes all curves to ultimately cross. However, in the regions
concluded that for the ranges investigated, plenum length has a much of the crossings, the Nu/Nu/y values do not differ appreciably. The
greater effect on the heat transfer coefficients than does plenum di- low-Reynolds-number curves (left-hand panel) are separated by as
ameter. Indeed, considering the modest effect of plenum diameter, much as ten percent at downstream stations x/d ~ 15 owing to the
it appears that the results presented here are valid for all plenums with absence or presence of the undershoot effect.
D/d > 3. Before leaving Figs. 4-9, a brief comment appears appropriate
An alternative perspective on the plenum-related heat transfer about another enhancement mechanism, namely, the boundary layer
enhancement is provided by Fig. 9. Here, Nu/Nu/d is plotted versus development that gives rise to Nu/Nu/^ values well in excess of one

420 / VOL. 103, AUGUST 1981 Transactions of the ASME

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 in the entrance region. The figures show that entrance-region en- Table 1 Thermal entrance lengths, xe/d
hancement is greater at low Reynolds numbers than at high Reynolds
numbers, a characteristic which does not appear to be well docu- L/d
mented in the literature. Did Re 1 5 10
Thermal Entrance Lengths. The results presented in the pre- 3 5000 16.5 11 9.5
ceding section, especially Figs. 2 and 9, suggest that the rapidity of 10,000 14 11 10
the thermal development is affected by both the plenum geometry 20,000 12.5 12.5 12
and the Reynolds number. To quantize the rate of thermal develop- 30,000 13 14 14
40,000 13 15 15
ment, a thermal entrance length x,, has been defined as the distance 50,000 13.5 16 16
from the test section inlet at which Nu/Nu/d = 1.05. Dimensionless 60,000 13.5 16.5 16.5
thermal entrance lengths xe/d based on this definition are listed in
6 5000 17 11 9.5
Table 1. 10,000 15 11.5 10
It is seen from the table that the overall range of the thermal en- 20,000 13 13.5 12
trance length extends from 10 to 17 tube diameters. The plenum 30,000 13.5 15 13.5
length appears to play a significant role in that the dependence of xe/d 40,000 14 16 14.5
50,000 14 16.5 15.5
on Re is affected by L/d. With a short plenum in place (i.e., Lid = 1), 60,000 14.5 17 16
the thermal entrance length increases only very slightly with Reynolds
number for Re > 20,000, but is quite sensitive to Reynolds number
at lower Re. These characteristics are similar to those expected when more from the data, which is entirely expected because it is based on
the plenum is absent (i.e., hydrodynamically developed flow at the a fully turbulent model. It might also be noted that equation (7) is a
inlet of the heated tube). On the other hand, for the two longer ple- slightly simplified version of the complete Petukhov-Popov equation
nums (L/d = 5 and 10), there is an appreciable increase of xe/d with (see equation (48) of ([10]). For airflow, the Nu/d from the complete
Reynolds number for Re > 10,000, while at lower Re the variation of Petukhov-Popov equation is 2j percent lower than that from equation
xe/d is slight. These marked differences in the characteristics of the (7) at Re = 10,000, while for Re = 60,000 the complete equation gives
thermal entrance length are reflective of flow field differences that a N u ^ value that is 1.3 percent higher than that of equation (7).
are caused by the presence of the plenum. Therefore, the agreement of the present data with the complete
Petukhov-Popov equation is even better than that in evidence in Fig.
Further inspection of Table 1 shows that the plenum diameter plays
10.
a secondary role. In general, the xe values for the two plenum diam-
eters do not differ by more than one tube diameter. With regard to other comparisons with the literature, it has already
been noted that there are no prior systematic studies of the effect of
Literature Comparisons. The first of the literature comparisons
plenum geometry on heat transfer in a tube fed by the plenum. Thus,
is for the fully developed Nusselt number Nufd- Among the presently
definitive comparisons of the present results with the literature are
available predictive equations for Nu/d, the widely quoted semi-
precluded. To enable limited comparisons to be made, a special set
empirical equation of Petukhov and Popov
of experiments was performed in which L/d was set at its minimum
M (//8)RePr
; possible value (L/d = 3) consistent with the thermocouple installation
NUH = 7}
' 1.07 + 12.7(//8) 1 / 2 (Pr 2/3 - 1) in the plenum and with the avoidance of extraneous conduction be-
/ = (1.821ogRe - 1.64)-2 (8) tween the upstream and downstream faces of the plenum. This ar-
is regarded by many as being the most accurate. Equation (7) is pur- rangement was thought to be a close approximation for the case in
ported [10] to be accurate to ±6 percent for Prandtl numbers between which an already hydrodynamically developed flow enters a heated
0.5 and 200 and for the Reynolds number range from 104 to 5 X 106. tube.
To avoid massive data overlap, the present results for D/d = 3 and The results of those experiments (Re = 10,000 and 50,000, D/d =
D/d = 6 have been plotted on separate graphs and respectively 6)are presented in Fig. 11 along with analytical predictions of [11] and
compared with equation (7). The two graphs are identical in all re- [12] which correspond to turbulent flow in a uniformly heated tube
spects, and only that for D/d = 3 is presented here in Fig. 10. with an axially unchanging velocity distribution. The predictions of
The figure conveys two main messages. The first is that the fully [12] fall well below the data, and this tendency has been noted by
developed Nusselt number is independent of L/d, and this is to be others (e.g., [2]). On the other hand, for Re = 50,000, the predictions
taken together with the aforementioned independence of Nu/d from of [11] are in remarkably good agreement with the data. The agree-
D/d. This finding confirms the expectation that given sufficient length ment at Re = 10,000 is not as good, but the deviation might well be
of run, the specific details of the inlet velocity field will ultimately due, at least in part, to the gap associated with L/d = 3 =•= 0. The lower
disappear so that the fully developed regime is truly universal. the Reynolds number, the more sensitive is the flow to the presence
The other noteworthy feature of Fig. 10 is the excellent agreement of such a gap.
of the present data with the Petukhov-Popov equation in the range The foregoing comparisons may be regarded both as a support of
Re 2: 10,000, with a maximum deviation of five percent. At lower the present experimental technique as well as of the predictions of
Reynolds numbers, the Petukhov-Popov equation departs more and [10] and[ll].

200 ANALYSIS DATA

O L / d « 10 tl [II] L/D =0.25
D L/d = 5 [12] D/d = 6
100' A L/d = 1
^ 1.2 -\\
80
^ Re = 5 0 , 0 0 0

, 60 :
- Petukhov a ,vp
X 1.0
1 1 1 1 1 1
20 25 30
40
Re = 10,000

20 POPOV N ^ ^
1 1 1 15 20 25
8 10 40 60 X/d
Fig. 11 Comparisons of measured Nusselt number distributions for Ud =
V 1 1 1 1 1 1 1 0.25 with analytical predictions corresponding to hydrodynamically developed
Fig. 10 Fully developed Nusselt numbers for Did = 3 flow in a uniformly heated tube

Journal of Heat Transfer AUGUST 1981, VOL. 103 / 421

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 Table 2 P r e s s u r e loss c o e f f i c i e n t K
(a) D/d = 3
L/d = 1 Lid = 5 Lid = 10
Re K Re K Re K
5700-6700 0.211 5900 0.722 5500 0.973
6300 0.658 6200 0.935
8100 0.183 6700 0.628
12,000 0.166 12,000 0.572 12,000 0.928
20,000 0.105 20,000 0.518 20,000 0.940
28,000 0.084 28,000 0.505 28,000 1.012
45,000 0.070 45,000 0.532 44,000 1.019
61,000 0.073 60,000 0.580 61,000 1.021
(b) Did = G
5700-6200 0.170 5600 0.680 5500 0.880
5900 0.607 5800 0.831
6400 0.106 6800 0.539 6400 0.782
8000 0.107 8000 0.512
12,000 0.101 12,000 0.460 12,000 0.797
21,000 0.084 21,000 0.423 21,000 0.819
28,000 0.077 28,000 0.420 29,000 0.832
45,000 0.074 45,000 0.430 45,000 0.921
53,000 0.091 59,000 0.461 59,000 0.941
59,000 0.110
Plenum-Related Pressure Losses. As defined in equation (6) agreement prevailed between the measured Nu/y values and those
and discussed in the related text, the net pressure loss AP P due to the predicted by the Petukhov-Popov equation for Re > 10,000, which
presence of the plenum will be reported as a dimensionless loss coef- is the range of applicability of the equation.
ficient K which is the ratio of AP P to the velocity head in the tube at The pressure loss coefficient, which compares the plenum-related
the plenum inlet. The experimentally determined values of K are pressure loss with the.velocity head in the tube, increased more or less
listed in Table 2. linearly with the length of the plenum.
The most significant message of the table is the tight correlation As a relevant adjunct to the main focus of the research, a study was
between the magnitude of K and Lid. As Lid varies from 1 to 5 to 10, performed to examine the role of guard heating/cooling at the elec-
K varies, in global terms, from 0.1 to 0.5 to 1. This trend is entirely trical bus adjacent to the tube inlet. It was found that such heating/
reasonable in that the two major contributions to K, the abrupt ex- cooling was necessary for accurate Nusselt number results at low
pansion at the plenum inlet and the abrupt contraction at the plenum Reynolds numbers (Re ~ 5000) and was desirable, but not strictly
exit, assert their separate identities more and more as L/d increases. necessary at higher Reynolds numbers (Re > 50,000).
However, even for Lid = 10, the K values listed in the table are less
than the sum of the separate expansion and contraction coeffi- Acknowledgment
cients. This research was performed under the auspices of the Office of
With regard to Reynolds number effects, K varies somewhat ir- Naval Research (contract no. N00014-79-C-0621).
regularly when Re > 12,000. At low Reynolds numbers, there is a
general tendency of K to increase as Re decreases, with the greatest References
sensitivity in evidence for L/d = 1. In general, K is not very sensitive 1 Boelter, L. M. K., Young, G., and Iverson, H. W., "An Investigation of
to D/d for the two values of D/d that were investigated. Aircraft Heaters XXVII—Distribution of Heat Transfer Rate in the Entrance
Section of a Circular Tube," NACA TN 1451, 1948.
Concluding Remarks 2 Mills, A. F., "Experimental Investigation of Turbulent Heat Transfer
A systematic experimental study has been performed to determine in the Entrance Region of a Circular Conduit," Journal of Mechanical Engi-
the effect of plenum geometry on the turbulent heat transfer char- neering Science, Vol. 4, 1962, pp. 63-77.
3 Wesley, D. W. and Sparrow, E. M., "Circumf'erentially Local and Average
acteristics of a tube fed by the plenum. Measurements were also made Turbulent Heat Transfer Coefficients in a Tube Downstream of a Tee," In-
of the plenum-related pressure drop. The experiments were designed ternational Journal of Heat and Mans Transfer, Vol. 19, 1976, pp. 1205-
and executed with painstaking care in order to provide a set of results 1214.
of impeccable quality which can serve both as a guide to design and 4 Ede, A. J., Hislop, C. I., and Morris, R., "Effect on the Local Heat
as a standard against which to compare candidate analytical Transfer Coefficient in a Pipe of an Abrupt Disturbance of the Fluid Flow:
Abrupt Convergence and Divergence of Diameter Ratio 2/1," Proceedings of
models. the Institution of Mechanical Engineers, London, Vol. 170, No. 38, 1956, pp.
It was found that at axial stations in the near neighborhood of the 1113-1126.
tube inlet, the local heat transfer coefficients were higher for longer 5 Ede, A. J., "Effect of an Abrupt Disturbance of the Flow on the Local
Heat Transfer Coefficient in a Pipe," HEAT 164, National Engineering Lab-
plenums. Thus, a longer plenum functions as an enhancement device. oratory, East Kilbride, Glasgow, Scotland, 1959.
As an example of the enhancement, it may be noted that at x/d - 1, 6 Ede, A. J., Morris, R. and Birch, E. S., "The Effect of Abrupt Changes
the heat transfer coefficient is 40 percent larger at a plenum length of Diameter on Heat Transfer in Pipes," NEL Report No. 73, National Engi-
L/d = 10 than when L/d = 1. The extent of the enhancement de- neering Laboratory, East Kilbride, Glasgow, Scotland, 1962.
creases as x/d increases and can be regarded as negligible for x/d > 7 Emerson, W. H., "Heat Transfer in a Duct in Regions of Separated
Flow," Proceedings of the Third International Heat Transfer Conference, Vol.
8. The plenum diameter appears to play a minor role with regard to 1,1966, pp. 267-275.
the enhancement, so that the present results may be regarded as being 8 Zemanick, P. P. and Dougall, R. S., "Local Heat Transfer Downstream
applicable for all D/d > 3. of Abrupt Circular Channel Expansion," ASME JOURNAL OP HEAT TRANS-
FER, Vol. 92,1970, pp. 53-60.
Thermal entrance lengths xc, based on a five percent approach to 9 Lau, S. C, "Effect of Plenum Length and Diameter on Turbulent Heat
complete thermal development, encompassed the range from 10 to Transfer in a Downstream Tube and on Plenum-Related Pressure Losses,"
17 tube diameters for the investigated operating conditions. Longer Ph.D. Thesis, Department of Mechanical Engineering, University of Minnesota,
plenums brought about a greater sensitivity of xc to the Reynolds Minneapolis, Minnesota, 1980.
10 Petukhov, B. S., "Heat Transfer and Friction in Turbulent Pipe Flow
number in the fully turbulent regime but gave rise to a lesser sensi- with Variable Physical Properties," in Advances in Heat Transfer, Academic
tivity to Re in the low-Reynolds-number turbulent regime. The en- Press, New York, 1970, pp. 503-564.
trance lengths were not found to be sensitive to the plenum diam- 11 Siegel, R. and Sparrow, E. M., "Comparison of Turbulent Heat Transfer
eter. Results for Uniform Wall Heat Flux and Uniform Wall Temperature," ASME
JOURNAL OF HEAT TRANSFER, Vol. 82,1960, pp. 152-153.
The fully developed Nusselt numbers Nu/j were independent of 12 Deissler, R. G., "Analysis of Turbulent Heat Transfer and Flow in the
both the length and diameter of the plenum chamber. Excellent Entrance Regions of Smooth Passages," NACA TN 3016, 1953.

422 / VOL 103, AUGUST 1981 Transactions of the ASME

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 Analytic Prediction of the Friction
M. J. Scott
Detroit Diesel Allison,
Detroit, Ml
Factor for Turbulent Flow in
R. L. Webb
Associate Professor,
Internally Finned Channels
Department of Mechanical Engineering,
The Pennsylvania State University, This work develops an analytical model for the friction factor with turbulent flow in inter-
University Park, PA 16801 nally finned channels. Such channels are an important class of enhanced heat transfer
surfaces. Until this work, no analytical models for the turbulent friction factor have been
proposed. The present model assumes the validity of the Law of the Wall and applies the
logarithmic velocity distribution to the interfin and core regions of the flow. Theoretically
based friction factor equations are developed for internally finned circular tubes and rec-
tangular channels. The model predicts Carnavos data for 21 internally finned tubes with-
in ±10 percent. Friction factor data were taken for five internally finned, rectangular
channels. The analytical model predicts these data within ±10 percent, except for the
case of a very high fin.

Introduction
This work is concerned with turbulent flow in internally finned helical internally finned tubes. The optimum geometric parameters
channels. Circular internally finned tubes illustrated in Fig. 1 are of require minimum tube material for specified duty and operating
interest for heat exchanger application. Although internally finned constraints.
tubes are commercially available in copper and aluminum, no ana- Ornatskii [5] recognized that the fins cause a reduced velocity in
lytical models are available for prediction of the heat transfer and the interfin region, and developed an empirical equation relating the
friction characteristics. Work conducted by Carnavos between 1977 interfin and core region velocities to the average velocity in the
and 1979 has led to the development of an empirical correlation for channel. This work was performed using a flow channel which simu-
the Nusselt number and friction factor as a function of the internal lated a single interfin channel and an adjacent core region flow
fin geometric parameters. Carnavos' 1977 paper [1] contains heat channel. Local and average velocities were measured in the two re-
transfer and friction data for turbulent flow in 21 axial and helical gions. The work was extended [6] to determine the resulting effects
finned tubes. Carnavos [1] initially attempted to correlate the data on heat transfer. Details of this work will be discussed in a later sec-
using the hydraulic diameter in the Dittus-Boelter heat transfer and tion.
Blasius friction factor equations. He found that the hydraulic diam- Watkinson, et al. tested internally finned tubes with turbulent air
eter overpredicted the Nusselt number and friction factor. Carnavos [7] and water [8] and attempted to develop heat transfer and friction
developed empirical correction factors to the hydraulic diameter correlations. Their correlation uses the hydraulic diameter with a
based correlation. The correction factor given by equations (1) and correction factor taken from Ornatskii, et al. [5]. They found different
(2) are multiplied by the hydraulic diameter based correlations. friction correlations for air and water. The correlations do not agree
with those of Carnavos. Vasil'chenko and Barbaritskaya [9,10] give
(1) empirical heat transfer and friction correlations for turbulent flow
of transformer oil in aluminum internally finned tubes. Other inter-
nally finned tube data are reported by Hilding and Coogan [11] and
(2) Lipets, etal. [12].
This literature survey shows that the empirical Carnavos correlation
Carnavos' 1979 study [2] used water and an ethylene glycol-water
[1, 2] represents the state-of-the-art, based on reliable data for air,
mixture as working fluids with eight of the original tubes and three
water and water-glycol mixtures. The works of Carnavos and Wat-
additional helical finned tubes. The 1979 study replaced the cos°-5a
kinson show that the hydraulic diameter concept overpredicts the
in equation (1) by sec acv. The air, water and glycol data were corre-
friction factor and heat transfer coefficients. Ornastskii's experiments
lated within ±10 percent for 10,000 < ReH < 100,000. The equation
show that the failure of the hydraulic diameter concept occurs because
(1) and (2) correlations yielded 0.80 <FF< 1.1 and 0.64 < FH < 0.96.
the velocity is slowed in the interfin region.
Thus, the hydraulic diameter correlations tend to overpredict the Nu
and /. A hydraulic diameter model assumes that the velocity distri-
bution is unaffected by the presence of the fins. The inaccuracy of the
hydraulic diameter correlation apparently occurs because the velocity
is reduced in the interfin region, relative to that in a smooth tube. The
heat transfer and friction characteristics should depend on the di-
mensionless fin height (e/D) and on the interfin region aspect ratio,
(p — t)/e. Equations (1) and (2) do not explicitly account for these
geometric parameters.
Patankar, et al. [3] used a numerical model to predict the heat
transfer and friction characteristics of axial internal fin tubes. This
model employs an empirical constant to fit the results with Carnavos
data [1].
Webb and Scott [4] used the equation (1) and (2) correlations to
predict "optimum" geometric parameters (e/D and pie) for axial and

Contributed by the Heat Transfer Division for publication in the JOURNAL

OF HEAT TRANSFER. Manuscript received by the Heat Transfer Division
August 11,1980. Fig. 1 Geometric variables of internally finned tube

Journal of Heat Transfer AUGUST 1981, VOL. 103 / 423

Copyright © 1981 by ASME
Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 This work seeks to develop a theoretically based model for pre-
diction of the friction factor with turbulent flow in internally finned
channels. The successful completion of this objective will provide the
basis for the development of a future analytical heat transfer
model.

Theoretical Development for Circular Tubes

The theoretical model is based on the Law of the Wall [13]. The Law
of the Wall assumes that the velocity distribution in the viscous in-
fluenced wall region is independent of the channel shape or pipe di-
ameter. For the turbulence dominated wall region (y + > 26) the al-
gebraic form of the Law of the Wall is
yu*
2.5 In — + 5.5 (3) Fig. 2 Definition of interfin and core region flow areas

The friction factor in circular channels is well-predicted assuming between theory and experiment for all channel aspect ratios, including
equation (3) applies over the entire pipe radius. Equation (3) may be the square duct.
integrated over the pipe radius giving u/u* = y/2/f which results in Due to the prior success in applying the Law of the Wall to rec-
tangular channels we have chosen to apply the concept to internally
the friction factor equation
finned tubes. However, substantial modifications in method are re-

Vr 2 - 51nRe V?
quired to account for the more complex flow geometry. Figure 2 shows
+ 1.75 (4)
that the cross-sectional flow area (S) of a circular finned channel is
divided into two flow areas: 1) the interfin region (Sf) for 0 < y < e
Hartnett, et al. [14] and Brundrett [15] have predicted the friction
and 2) the core region (Sc) for y > e. Thus,
factor in smooth square and rectangular channels by integrating
equation (3) over the cross section. As discussed by Hartnett, et al., S = Sf + Sc (5)
an approximation is introduced in this case, since secondary flows
occur in such noncircular channels. Schlichting [16] describes the local Writing the continuity equation for the two flow regions yields
flow structure in detail, based on Nikuradse's measurements. The
uS = UfSf + ucSc (6)
secondary motion consists of fluid velocity components near the wall
which flow toward the corner, and join to flow outward along the Dividing by u* and using u/u* = \f2/f allows equation (6) to be
corner bisetrix. Schlichting [16] shows figures of this secondary flow written as
field, based on Nikuradse's measurements in rectangular channels.
The secondary flows augment the velocity in the corner region and
cause the wall shear stress to vary from a maximum at the mid-point
and drop sharply to zero in the vicinity of the corners. Both Hartnett,
4. 2
f
iif Sf
u* S
uc Sc
u* S
(7)

Expressions must now be developed for the interfin region velocity

et al. and Brundrett ignore the effect of secondary flows in their (Uf/u*) and the core region velocity (uc/u*). Separate developments
analysis of rectangular channels. are necessary for each region. The development will be presented for
The analysis of Hartnett, et al., uses an iterative procedure to draw an internally finned circular tube. Then, the analysis will be extended
a flow field of orthogonal constant velocity and gradient lines. Then, to the case of internally finned rectangular channels.
equation (3) is graphically integrated over the flow field. Their pre- Interfin Region. The required assumptions are: (1) Secondary
dicted results were in good agreement with data for 10:1 aspect ratio flows are neglected. (2) The interfin region is an open rectangular
channels. But, the predicted values were 12 percent low for a 1:1 aspect channel. (3) The velocity distribution between each wall and the
ratio channel. corner bisectrix is given by equation (3).
Brundrett uses a simpler procedure which models the rectangular Figure 3 shows the regions over which equation (3) is integrated.
channel as the combination of an infinite aspect ratio channel with Different integrations are required for the three possible interfin re-
half of a square channel at each end. Brundrett shows good agreement gion aspect ratios: 1) e/b > 1, 2) e/b = 1 and 3) e/b < 1. Symmetry is

.Nomenclature^
A = total heat transfer surface area DHR = hydraulic diameter of a smooth rec- t = fin thickness
An = surface area based on nominal (smooth tangular channel T = temperature
channel) area Di = inside diameter of a smooth tube; di-
u = velocity
a = one-half channel spacing in a rectangular ameter to the base of the fins in a finned
channel tube ue = velocity at fin tip
b = half the spacing between fins, b = e = fin height u* = shear velocity, u* = -\/TW/P
(p-t) e' = distance defined by Fig. 3, Case 1 U = average velocity
/ = fanning friction factor uc = average velocity in the core region
2
n = number of fins in a tube Uf = average velocity in the inter-fin region
b' = distance defined on Fig. 3, Case 3
p = fin spacing Au e = shift in velocity profile for a finned
dfh = inter-fin hydraulic diameter, d;h = Pw = wetted perimeter channel, cf. equation (13)
Abe R = tube radius w = half-width of rectangular channel
(b + e) Ren = Reynolds number based on hydraulic y = distance from a wall
Dc = diameter of the core region diameter, Djju/v z = distance from a wall
DH = hydraulic diameter of a finned S = actual minimum free flow area a = helix angle in a spiral fin tube measured
channel, Sc = flow area of the core region from the tube axis
Sf = flow area of the inter-fin region v = kinematic viscosity
4A
Sn = flow area based on nominal (smooth p = density
channel) area TW = average wall shear stress

424 / VOL. 103, AUGUST 1981 Transactions of the ASME

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 -?I- E

E©
JS~
1
yg>§
2b
Case 2 e = b
3 4 5 6
REYNOLDS NUMBER

m
i.*• .i
2b
Fig. 4 Comparison of friction factor predicted by Table 1 equations with
Carnavos data [1]

T a b l e 1 E q u a t i o n s to b e u s e d i n e q u a t i o n (7) to
Case 3 e < b p r e d i c t t h e f r i c t i o n f a c t o r for i n t e r n a l l y f i n n e d
circular tubes
Fig. 3 Integration areas for interfin region

employed, which simplifies the integrations. The integrations to be if/ Aue

Case
performed for each of the three cases are given by equations 8 (a, b) u* U* u*
and (c). For simplicity, l e t B = 2.5 and A = 5.5.
e/b > 1 equation (11a) equation (16) (14a)
b e/b = l (lib) equation (16) (14b)
if/=l( C (e - e') B In + A dz e/b < 1 (He) equation (16) (14c)
u* eb [Jo

dy J Bin + A\dz at y — e above a smooth surface. This velocity shift is given by

Aue = ues — uef (13)
+ f dz f* |Bln — + A\dy for(e > 6) (8a)
Jo Jo \ v I The uef is the average velocity at the fin-tip and is obtained by in-
b
1 e
f~ey I ZU* \ tegrating equation (3) to find the average velocity at y = e. The ues
if/ =
u* eb X dy J Bin
e
+ A\dz
is calculated from equation (3) a t y = e. The resulting expressions for
the three cases shown by Fig. 3 are:
b for (e = b) (8b)
+ C dz C" B i n — + A\dy Aue
: 2.5 In - + 2 . 5 f o r ( e > 6 ) (14a)
\b
1
if/ = j ; d y j j ( B l n ^ + A)d Z Aue
u* eb • = 2.5 for (e = b) (14b)
e .
w
X *'

dz j
/»S'» /

Bin — +
yu* I

A\dy ^
u*
= 2.5-
b
for(e<b) (14c)

X" ( b - b ' ) |B In — + A|dy for(e < b) (8c) The existence of the velocity shift has been well-documented for
flow over roughened surfaces and is discussed by Hinze [17].
where The average core region velocity is found by integrating equation
(12) over e < y < R
b=-(p-t) (9)
uc l rtt I yu* AuPe \
~Z = — \ 2w(R-y) 2.5 In ^— + 5 . 5 -\dy (15)
irDi "* Ac Jo \ v u* I
(10)
The result of the integration
ition is
rc is the number of fins and D; is the diameter to the base of the fins, uc + Aue Di2
and e' and b ' are defined in Fig. 3. By geometry, e' = b and b ' = e. u* ~ (Di - 2e) 2 '
Equations (11a, b) and (c) are the integrated values of equations
4(£>;e - e 2 )
(8a, b) and (c). The equations are written in terms of the hydraulic
diameters (DH) and Reynolds number (Re//) of the interfin region,
which are defined in the Nomenclature.
„ 2.5 In — R e H
(Dt - 2e)2 \D„ VD
( i D ; 2 - 4 A e + 2e 2 )
2.5 + 5.5 (16)
(Di - 2e) 2
UJ_
u*
2.5 In | — R e H
PH VI) 1 . 2 5 - + 3.0 for (e > b) (11a)
The term Aue in equation (16) is given by equations (14) for the
interfin aspect ratio of interest.
i£/ ; 2.5 In I — R e H W - J + 1.75 for (e = b) (lib)
u* Predicted Friction Factor. The theoretical friction factor is
obtained by substituting equations (11) and (16) in equation (7). Table
• ^ = 2.5 In I — Re H A / - ) - 1.25 - + 3.0 for (e < b) (lie) 1 summarizes the equations to be used for each of the three possible
u* I \DH V 2/ b J interfin aspect ratios.
The Core Region. Equation (3) also applies to the core region (y Figure 4 shows the ratio of the predicted-to-measured friction factor
> e). However, the velocity a t y = e is less than that which would occur based on the Table 1 equations and the data of Carnavos [1]. Carna-
at the same wall distance in a smooth tube due to the increased surface vos' data are for 11 different axial internal-fin geometries. The
shear of the interfin region. Thus, the velocity in the core region is agreement is within ±10 percent except for two data points. These
points correspond to the low Reynolds number data for Carnavos
uc , yu* (12) tubes number 10 and 13. These tubes have thick fins and a small
—
If* = 2.5 In —V + 5.5 • U* spacing between fins. The existence of a relatively thick viscous
where Aue is the shift of the velocity profile relative to the velocity dominated flow may exist in the interfin regions. This flow condition

Journal of Heat Transfer AUGUST 1981, VOL 103 / 425

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 would cause the present model to over-predict the friction factor.
The 11 tubes tested by Carnavos have 0.36 < e/b < 0.60, and 0.02 T u r
< e/D < 0.29. Therefore, the ability of the model to predict the e/b
> 1 cases is not tested in Fig. 3. The assumption of rectangular shaped jfzm
interfin channels is of questionable validity when e/b > 1 with large
e/D. For this situation, a more precise formulation of equation (8a) Fig. 5 Geometric variables of Internally finned rectangular channel
should account for the decreasing channel width with distance from
the wall. Wall curvature would not affect the core region equation TRANSITION
(equation (16)) or the velocity shift equation (equation (14a)). The
TRANSITION
SECTION
^ 1 SECTION

effect of wall curvature on Uf/u* was examined by writing equation

(8a) to account for wall curvature effects; this integral yields values [SHOP! FINNED TEST SECTION' 'SMOOTH ENTRANCE SECTION
ITION

oiUf/u* 4 to 10 percent smaller than from equation (11a). When these AIR SUPPLY.

smaller Uf/u* values are substituted in equation (7), the friction factor C><3
will be reduced 2 to 5 percent. This correction would yield a very ^
FLOW ORIFICE
modest improvement for tubes 10 and 13. Fortunately, the case of -^-
AIR SUPPLY
large e/D; and e/b > 1 is not a practical geometry (for circular tubes), (COMPRESSOR) $ REGULATOR T THERMOCOUPLE
and should not be encountered in practice. WA FILTER P STATIC PRESSURE
iP PRESSURE DROP

Fig. 6 Schematic of test apparatus

Theory for Internally Finned Rectangular Channels
A procedure similar to the circular channel model may be applied T a b l e 2 E q u a t i o n s t o b e u s e d i n e q u a t i o n (7) f o r a n
to rectangular channels. The geometry of the rectangular channel is internally finned rectangular channel
shown in Fig. 5. Equations (11a, b) and (c) apply equally for both
channel shapes, since the interfin region is assumed to be of rectan- Aue
gular cross section. However, the core region velocity expression Case
u* u* u*
(uju*) is different. In this case, equation (12) must be integrated over
a rectangular channel, rather than a circular cross section. The result e/6>l equation (11a) equation (17) equation (18a)
of the integration is e/6 = l (116) equation (17) (186)
e/6<l (He) equation (17) (18c)
u^+Au, _ 0 ^ | o ) a ^ |_a_ ff\
u* w(a - e) \DH \ 2)
Table 3 Rectangular test sections tested
+ JiZf)2.61n(gJJlReH ff_\_(w + ie-a)e
Test Section Number
2w \DH V 2] w(a - e)
1 2 3 4 5 6
X 2.5 In — R e H \ / - - 11 + — I 2.5 + 5.5 (17) — smooth 0.115 0.211 0.211 0.211 0.405
\DH V 2] \ 2w) 2a
The equations for the velocity shift (Aw,,) are obtained following ^ — 1.690 0.918 1.836 3.673 1.915
e
a similar procedure for the rectangular shaped channel. The results
- — 5.65 10.40 10.40 10.40 19.95
of the integration are
t
P g ( m m ) 35.70 13.40 7.75 14.38 24.53 14.89
— = 2.5-] + - for(e>6) (18a)
u* \e) a
test section was connected to the rectangular entrance section by
— = 2.5 + - for (e = 6) (186) common side plates. The channel width and spacing of the entrance
u* a section were the same as those of the test section. The 2a and 2w di-
Aue lb\ b mensions used for data reduction were based on average values of 20
— - = 2.5 - + - f o r ( e < 6 ) (18c) measurements of the channel width and 50 channel spacing mea-
u* \e) a
surements. The mean deviation of the channel width and spacing
Table 2 summarizes the equations to be used for the friction factor dimensions were 0.20 mm (0.1 percent) and 0.06 mm (0.3 percent),
in an internally finned rectangular channel. respectively. Static pressures were sensed by 1.6 mm dia holes drilled
The equations for the circular and rectangular channels are dif- 95 mm from each end of the test section. At each axial location, three
ferent primarily because the average velocity in the core region (u c ) taps were drilled in the top plate and two were drilled in the bottom
is different for the two channel shapes. plate. The taps at each axial location were manifolded to yield the
average static pressure at each axial location. A traveling-well mi-
cromanometer measured the difference between static pressure over
Experimental Program the 1650 mm flow length between static taps.
Friction data were taken in a finned rectangular flow channel in The initial tests were performed on a smooth rectangular channel
order to provide further validation of the theoretical model. The test
(test section 1). The measured friction factor is compared with the
section was made from flat aluminum plates in which axial grooves
data of others [14,17,18] and our theoretical prediction, in Fig. 7.
were machined to insert 0.40 mm thick fins made of aluminum strip.
The predicted friction factor for the smooth channel is obtained
A cross-sectional view of the test section is shown in Fig. 5 and the
by integrating equation (3) over the regions illustrated in Fig. 8. The
apparatus schematic is illustrated in Fig. 6. For all tests, the channel
resulting analytical prediction for the smooth rectangular channel
dimensions were 2w = 190.75 mm and 2a = 19.69 mm which gives an
is
aspect ratio of 9.7:1. The range of test section dimensions tested is
listed in Table 3.
Air was used as the working fluid and the flow rate spanned 5000
< Ren < 90,000. The air flow rate was measured by a calibrated
ASME orifice in a 52.5 mm dia pipe installed upstream of the test Equation (19) is in very close agreement with the model developed
section. The air passed through a 760 mm long transition section by Brundrett [15]. Equation (19) underpredicts the reference [14,18]
having a 5.2 deg divergence angle before entering a 1840 mm length and [19] data 3 to 5 percent. The present data agree well with the re-
smooth rectangular entrance section. The 1840 mm length rectangular ferenced data for Rejf < 20,000 but are about 15 percent high at Re/f

426 / VOL. 103, AUGUST 1981 Transactions of the ASME

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 .02 ' » ThisSludy '(9.7:0 ' ' ' f e/2o p/e
D Huebschsr (8:1) 0 Smoolh
A Washington i Marts (B.89:l) 0 0115 1.690
OHarlnstt st at.(IO:l) a 0.211 0.918
— «an 19 (9.71) A02IJ 1.836
A 0.211 3,673
D 0.405 1.915
So.o
: ca
°» S o
f? 8
AA :

.003
10 100
" ' 1 1 1
3 10 100 200
REYNOLDS NUMBER ReuxlO~
REYNOLDS NUMBER Re »I0" 3
Fig. 7 Predicted and measured friction factors for smooth rectangular
Fig. 9 Friction factors for smooth and internally finned rectangular chan-
channel. Numbers in parentheses define channel aspect ratio
nels

1.2

^UE

yr®\
H g 0.9 I- **\ »*
s O Smooth
Fig. 8 Integration areas for smooth rectangular channel
I 0.8 O 0.115 1.690
V) A 0.211 0.91 B
A 0.211 1.836
a 0.7
= 80,000. The deviation at high Reynolds numbers suggests possible A 0.211
• 0.405
3.673
1.915
roughness effects.
0.6
Figure 9 shows the data obtained for the test sections listed in Table
3. It is interesting to note that the friction factors of the finned
channels are higher than that of the smooth channel for some cases
Fig. 10 Comparison of Internally finned rectangular channel data predicted
and lower in other cases. Figure 10 compares the predicted and by Table 2 equations with experimental data
measured friction factors. The predicted friction factors are based
on the Table 2 equations. The theory predicts the present data within
10 percent for four of the five finned channels. The fifth case (test
w
section 6) has very high fins and the friction factor is underpredicted
25 to 30 percent. In this case, the fin height occupies 81 percent of the i
channel spacing and the core region flow area is very small. Therefore,
the model, which assumes distinctly different velocities in the interfin
and core flow regions may become questionable. When the fins span
the entire flow channel, the flow should be modeled as distinct non-
L
circular flow channels. - -
The present theoretical equations predict the friction factor for
internally finned circular and rectangular channels within ± 10 per-
cent for typical geometries of interest.

Comparison with Ornatskii Model

Ornatskii [5] developed an empirical equation to relate the velocity
in the interfin region (uf) to the average channel velocity {u). His e
relation was derived from velocity distribution measurements in the
special channel shape shown in Fig. 11. Air flow data were taken for '
0.25 < 2b/e < 1.9 with 7 < Re H X lO"4 < 25. A similar channel was
"*2 b
used for water flow data with 0.09 < 2b/e < 1.2 with 3 < Re# X 10~4
< 10. Velocity distributions were obtained for 25 channel configura- Fig. 11 Cross section of flow channel used by Ornatskii [5, 6]

tions. Their experimental results were empirically correlated by

equation (20).

^ = L22 (B°- 26 1 -exp

e,
dfh
J
(20) _
- "tt,_E
I-* I.2 " O
o
o
-
°o o o
I.I o
Dividing by u* and using u/u* = y/2/f, equation (20) is o 8 o
Uf n (2b\0-25 _o Q
—
'W;@"
I.O
• exp dfh (20a) o o °o o
o o
The predicted friction factor for an internally finned tube is ob-
0.9 - 0
° o -
o
tained by substituting Uf/u* from equations (11) into equation (20a). 0.8
o
Figure 12 shows the ability of equation (20a) to predict the friction o
factor for Carnavos' internally finned tubes. The predicted values 0.7
differ from the measured values by —25 percent to +19 percent. Or-
natskii's semi-empirical model yields significantly poorer results than
the present theoretical models defined by the equations in Tables 1 Fig. 12 Comparison of friction factor predicted by Ornatskii theory (equations
and 2. ( 2 0 a ) and ( 1 1 ) ) with Carnavos data [1]

Journal of Heat Transfer AUGUST 1981, VOL. 103 / 427

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 Designer Correlation
Although the theoretical models summarized in Tables 1 and 2
provide quite accurate predictions of the friction factor, they may be
cumbersome for design purposes. We have derived simple power law
correlations, which may be used for design purposes. These correla-
tions apply a geometric correction factor to the Blasuis friction cor-
relation defined in terms of the hydraulic diameter. The correlation
is derived by multiple regression analysis of the dimensionless vari-
ables. 0 1 2 3 4 5 6 7 8 9
For finned circular channels, the correlation is REYNOLDS NUMBER Re^lO"4

_ 0.056 Mo.067 / M0.124 /_e\-0-i7 Fig. 13 Comparison of multiple regression curve fit (equation ( 2 2 ) ) with
f internally finned rectangular channel data
~Re„°™\tl \DiJ W
Equation (21) correlates Carnavos data with ±10 percent. 4 Webb, R. L., and Scott, M. J.," A Parametric Analysis of the Performance
The comparable correlation for finned rectangular channels is of Internally Finned Tubes for Heat Exchanger Application," ASME JOURNAL
OF HEAT TRANSFER, Vol. 102,1979, pp. 38-44.
5 Ornatskii, A. P., Shcherbakuv, V. K., and Semena, M. G., "Investigation
0 26 of Velocity Distribution in Tubes with Internal Longitudinal Fins," Thermal
(Rew) - U W
Engineering, Vol. 17(1), 1970, pp. 108-111.
Equation (22) correlates all of the finned rectangular channel data 6 Ornatskii, A. P., Shcherbakov, V. K., and Semena, M. G., "Investigation
of the present study within ±10 percent, as shown in Fig. 13. of Heat Transfer in Channels Between Fins of Tubes with Internal Longitudinal
Fins," Thermal Engineering, Vol. 17(11), 1970, pp. 96-100.
7 Watkinson, A. P., Miletti, D. L., and Tarasoff, P., "Turbulent Heat
Conclusions Transfer and Pressure Drop of Internally Finned Tubes," AIChE Symposium
1 An analytical model was developed to predict the friction factor Series, Vol. 69(131), 1973, pp. 94-103.
for turbulent flow in internally finned circular and rectangular 8 Watkinson, A. P., Miletti, D. L., and Kubanek, G. R., "Heat Transfer
channels. The model applies the logarithmic velocity distribution to and Pressure Drop of Internally Finned Tubes in Turbulent Air Flow,"
ASHRAE Transactions, Vol. 81(1), 1975, pp. 330-349.
the interfin and core regions of the flow. 9 Vasil'chenko, Y. A., and Barbaritskaya, M. S., "Resistance with Non-
2 T h e analytical model predicts the friction factors in both isothermal Fluid Flow in Tubes with Longitudinal Fins," Thermal Engineering,
channel geometries within ±10 percent, except for very high fins. Vol. 16(1), 1969, pp. 28-35.
3 The algebraic form of the analytic model is different for circular 10 Vasil'chenko, Y. A., and Barbaritskaya, M. S., "Heat Transfer in Tubes
with Longitudinal Fins," Thermal Engineering, Vol. 16(5), 1969, pp. 105-
and rectangular channels due to the effect of channel shape on the 109.
average velocity in the core region. 11 Hilding, W. E., and Coogan, C. H., Jr., "Heat Transfer and Pressure Loss
4 The present theoretical model yields a more accurate prediction Measurements in Internally Finned Tubes," Symposium on Air-Cooled Heat
of the data than the semi-empirical model of Ornatskii, et al. Exchangers, ASME, New York, 1964, pp. 57-85.
12 Lipets, A. U., Zholudov, Y. S., Lokshin, V. A., and Gromov, G. V., "The
5 Simple power law friction correlations are also provided for Temperature Regime and Hydraulic Resistance of Tubes with Internal Lon-
design purposes. gitudinal Fins," Heat Transfer-Soviet Research, Vol. 1(5), 1969, pp. 86-94.
6 The friction factor model should provide a basis for the future 13 Hinze, J. O., Turbulence 2nd Ed., McGraw-Hill, New York, 1975, pp.
development of a theoretically based heat transfer equation. 614-619.
14 Hartnett, J. P., Koh, J. C. Y., and McComas, S. T., "A Comparison of
Predicted and Measured Friction Factors for Turbulent Flow through Rec-
Acknowledgments tangular Ducts," ASME JOURNAL OF HEAT TRANSFER, Vol. 84,1962, pp.
This study was supported by Department of Energy Contract 82-88.
EG-78-S-02-4699. 15 Brundrett, E., "Modified Hydraulic Diameter for Turbulent Flows,"
in Turbulent Forced Convection in Channels and Bundles, S. Kakac and D.
B. Spalding, eds., Hemisphere, Washington, D.C., 1979, pp. 361-367.
References 16 Schlichting, H., Boundary Layer Theory, 7th Ed., McGraw-Hill, New
1 Carnavos, T. C, "Cooling Air in Turbulent Flow with Internally Finned York, 1979, pp. 612-615.
Tubes," Heat Transfer Engineering, Vol. 1(2), 1979, pp. 41-46. 17 Hinze, J. 0., Turbulence, 2nd Ed., McGraw-Hill, New York, 1975, pp.
2 Carnavos, T. C., "Heat Transfer Performance of Internally Finned Tubes 635-637.
in Turbulent Flow," Advances in Enhanced Heat Transfer, J. M. Chenoweth, 18 Huebscher, R. G., "Friction Equivalents for Round, Square and Rec-
J. Kaellis, J. W. Michel, and S. Shenkman, eds., ASME, New York, 1979, pp. tangular Ducts," Heating, Piping and Air Conditioning, Vol. 19(12), 1974, pp.
61-67. 127-135.
3 Patankar, S. V., Ivanovic, M., and Sparrow, E. M., "Analysis of Turbulent 19 Washington, L., and Marks, W. M., "Heat Transfer and Pressure Drop
Flow and Heat Transfer in Internally Finned Tubes and Annuli," ASME in Rectangular Air Passages," Industrial and Engineering Chemistry, Vol.
JOURNAL OF HEAT TRANSFER, Vol. 101,1979, pp. 29-37. 29(3), 1937, pp. 337-334.

428 / VOL. 103, AUGUST 1981 Transactions of the ASME

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 Heat Transfer in a Capillary Flow
Elefiherios Papoutsakis 1
Graduate Student.
Emerging from a Reserwoir
Doraiswami Ramkrishna
Professor. An extended Graetz problem is analyzed, with a semi-infinite axial domain and the Robin
boundary condition on the heat-transfer wall. The heat-transfer problem examines a vis-
cous fluid entering a cylindrical capillary from a reservoir. The capillary fluid is exchang-
School of Chemical Engineering,
ing heat with the surrounding environment of prescribed temperature and thus the Robin
Purdue University,
West Lafayette, Ind. 47907 boundary condition is employed on the wall. Since axial heat conduction is included in
the analysis, a generalized Danckwerts boundary condition is shown to be most appropri-
ate for the tube entrance. The energy equation is decomposed into a system of first-order
partial differential equations, as in [4, 8, 9], to obtain a selfadjoint formalism. The Gram-
Schmidt orthonormalization process is finally used to obtain what is technically an ana-
lytical solution, which is computationally simple and efficient. Other entrance boundary
conditions are also discussed and analyzed.

1 Introduction of the entrance region. Usual Reynolds numbers in viscous-fluid

This work pertains to the temperature or concentration profile processing and many biological flows are of the order of 10~3 and the
development in a fluid exchanging energy or mass with its sur- approximation of a fully developed flow at 2 = 0 is a typical one [2,
roundings as it flows into a long circular tube. Our most specific ob- 6]. However, small Reynolds numbers would result also in small Peclet
jective is the analysis of low Peclet number problems with semi-infi- numbers even for Schmidt or Prandtl numbers of the order of 104.
nite heat- or mass-transfer sections, where the hydrodynamic problem Thus the validity of the model is not just restricted to the heat transfer
may be decoupled from the heat- or mass-transfer problem. On the problem in liquid metals. For the very low Peclet number regime,
exchange wall we wish to specify the boundary condition of the third where the conductive terms dominate the heat transfer mechanism,
kind (Robin), where a linear combination of the heat flux and the the velocity profile approximation has a minimal effect on the solution
temperature on the wall is prescribed. This boundary condition on of the problem. Although here we treat Newtonian flows only, ex-
the wall is for the most general and physically important case, where tensions to fluids with any velocity profile may be realized in a
the wall containing the fluid conduit has a finite resistance to heat or straightforward manner from the analysis presented herein. The only
mass transfer. The external fluid may be of constant temperature or difference will be computational, in the integration of the resulting
may have a prescribed temperature distribution. eigenvalue problem (equation (21)). Thus, the methodology may apply
to problems in experimental rheology, viscous fluid processing, heat
The specification of a semi-infinite axial domain derives from the
transfer in liquid metals and mass transfer in biological systems. In
fact that in a large number of problems of practical importance, what
the discussion and analysis that follow we discuss the heat transfer
precedes the heat- or mass-exchange section is either not well defined
problem only, the analysis of the mass transfer problem being entirely
physically (e.g., in biological systems [1-3]) or hydrodynamically very
analogous.
complex. For example, the fluid may enter the exchange section from
an adiabatic reservoir or an adiabatic conduit of larger cross section, The problem under discussion may be looked upon as an extended
as is the case in viscous-fluid processing. A preceding adiabatic section Graetz problem, in that axial heat conduction is included in the
of the same tubular cross section is one particular case that does not analysis, so that it can be applied to low Peclet number flows. We
provide always the best description of the physical picture. The latter consider as classical the problem without axial conduction but with
case leads to a boundary value problem with mixed boundary condi- any kind of a wall boundary condition. Past efforts on the extended
tions (Neumann-Robin) 2 on the wall. That problem is discussed in Graetz problems, which have been discussed in detail by Shah and
[4], its solution culminating into solving an infinite set of algebraic London [7], Papoutsakis [4] and Papoutsakis, et al. [8,9], were largely
equations. In view of the above discussion, in the present problem the approximate. Truly analytical (and computationally efficient) solu-
fluid will be taken to flow into the circular tube from an adiabatic tions to extended Graetz problems with infinite axial domains were
reservoir of large heat or mass capacity. obtained in [8,9]. The Dirichlet-Dirichlet problem with a jump in the
The corresponding hydrodynamic problem has been widely studied wall temperature at z = 0 was treated in [8]. The Neumann-Neumann
in the literature [5], its importance deriving from its applications in problem was solved in [9]. The solutions to both the above problems
almost every industrial process involving transport of fluids. For flows were based on a selfadjoint formalism obtained by decomposing the
of high Reynolds numbers, the velocity profile develops very slowly energy equation into a pair of first order partial differential equations
[5], so that the heat- or mass-transfer problem may not be decoupled in the spirit of [10]. The decomposition was accomplished by defining
from the hydrodynamic problem. However, in the case of low Reyn- an axial-energy flow function [8,9] based on physical arguments alone.
olds numbers, the velocity profile develops rapidly; from the results This technique allowed analytical solutions to be obtained for the
of Vrentas and Duda [5], and possibly extrapolating to large values apparently non-selfadjoint extended Graetz problems.
of their parameter /?, even at z = 0, the velocity profile has essentially Unfortunately, all extended Graetz problems with infinite axial
attained its fully-developed shape, except for a small core region domains and with the Robin wall boundary condition lead to
around the centerline of the tube. It is therefore a good approximation mixed-boundary-value problems, which elude analytical solutions
to assume the fully developed velocity profile from the very beginning [4]. However, the extended Graetz problem with the Robin boundary
condition on the wall and a semi-infinite axial domain admits a
1
Currently Asst. Professor in the Department of Chemical Engineering, Rice technically analytical solution, as it will be presently shown.
University, Houston, Tex. 77001 (to whom all correspondence may be ad- Yet, the treatment of the Robin wall boundary condition with a
2
The Neumann boundary condition specifies the flux on the conduit wall semi-infinite axial domain poses nontrivial difficulties, thus requiring
(zero for adiabatic wall). In mixed boundary conditions, the first name refers further analysis. Specifically, the Robin boundary condition demands
to the upstream section and the second one to the heat- or mass-exchange a nontrivial extension of the matrix differential operator that was used
section. in our recent work [8, 9]; the semi-infinite axial domain introduces
Contributed by the Heat Transfer Division for publication in the JOURNAL
OF HEAT TRANSFER. Manuscript received by the Heat Transfer Division May further difficulties whose resolution is even more demanding. The
27,1980. latter will become evident from the following analysis. This is indeed

Journal of H@at Transfer AUGUST 1981, VOL. 103 / 429

Copyright © 1981 by ASME
Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 the case where an apparently simpler problem (the present) is actually we ask that the reservoir be well stirred, have a large heat capacity,
more complex and computationally demanding than some apparently and be maintained at a uniform temperature To everywhere, by which
more complex problems (the problems in [8, 9]). we imply that the mixing and its heat capacity are enough to overcome
The heat transfer problem for flow out of a reservoir and with the axial gradient in temperature that will be induced by heat con-
Neumann wall boundary conditions (prescribed wall flux) has been duction at the outlet of the reservoir. Thus, if 2 = 0 is regarded as the
addressed by Pirkle and Sigillito [11]; a flat temperature profile was capillary inlet, then for z < 0 the essentially uniform temperature field
employed as an entrance boundary condition. Variational techniques, makes for negligible conduction whereas for z > 0 the heat transfer
leading to the solution of an infinite set of algebraic equations, were at the periphery will result in axial gradients near there and hence
employed for the computation of the coefficients of expansion of the in axial conduction.
solution in terms of non-orthogonal eigenfunctions [11]. Hsu [12] Summarizing, we have the situation of a virtually fully developed
treated the problem with a flat entrance temperature profile and the velocity profile at the outlet of the reservoir and a uniform tempera-
Robin wall boundary condition. The incorrectness and questionable ture of To with an energy efflux substantially by convection. On the
computability of that solution have been discussed by Pirkle and capillary side, however, there is some axial conduction especially near
Sigillito [13] and Michelsen and Villadsen [14], respectively. Finally, the periphery. We have now two choices for the entrance boundary
Sorensen and Stewart [15] solved the semi-infinite-domain problem condition. Either we insist that the temperature be continuous so that
with a Dirichlet wall boundary condition (prescribed wall tempera- the fluid will be entering at a uniform temperature To, in which case
ture) and a generalized Danckwerts entrance condition (see below). the situation is one of the total energy flux entering the capillary being
That problem was meant as an approximation to the Neumann- different from that leaving the reservoir (implying a flux disconti-
Dirichlet problem and was treated by means of a collocation technique nuity), or we admit a temperature discontinuity by allowing the total
[15]. flux to be continuous. Between these two options, the latter appears
to us to be more reasonable since the procedure is in essence collapsing
2 Analysis a finite but small region, where the temperature suffers a rapid change
2(a) Entrance Boundary Condition. In view of the discussion axially. Besides, the overall energy balance is intact. Indeed such
in the preceding section we consider the problem where a viscous fluid idealizations are not foreign to analysis. We thus converge on the
in a cylindrical capillary exchanges energy with its environment of boundary condition
constant temperature T/ through a wall of finite heat resistance. The
viscous fluid is entering the capillary at low velocity from a reservoir dT
of large heat capacity and constant temperature To. lim •k h pvz(r)cpT •• pvz(r)cpTo (1)
2^0 + dz
In the formulation of an entrance boundary condition, the presence
of a hydrodynamic entrance region represents an extraordinary degree Boundary condition (1) (which is a generalized Danckwerts type
of complexity even for an extensive numerical computation scheme. condition [16]) stands all by itself as an entrance condition for the type
Consequently, we focus on situations that admit some important of physical applications that we discussed in the introduction. It has
simplifications. Thus, we impose the restrictions of low Reynolds also been examined earlier [15] as a possible approximation to a
numbers for which the hydrodynamic entrance region is short and mixed-boundary problem (Neumann-Dirichlet). In this context, it
a Prandtl number large enough to preserve the uniformity of the is in order to compare the present problem to the Neumann-Robin
temperature profile in this entrance region (except for peripheral problem with regard to the temperature at z = 0. For the Neumann-
distortions due to small amounts of heat transfer). The Peclet number, Robin problem, the temperature at z = 0 will be lower than the tem-
however, is assumed to be small enough to render axial conduction perature of the present problem at the same point. This is because
important (which is true for low enough Reynolds numbers). Next of the high heat capacity of the reservoir. In other words, the fluid

.Nomenclature-
Ay = expansion coefficients, equations (30) Nu = Nusselt number Qe ~ dimensionless T e , equation (43)
and (36) /_ k(dT/dr)r=R 2R A;, \j+, Ay~ = eigenvalues, positive and
aj~ = first argument of Kummer's functions; negative eigenvalues, respectively
a constant, equation (46) T Tb-Tw 'k p. J = the square root of —A; - , equation
/ hR (47)
B = Biot number = — Pe = Peclet number (= pcpVR/k) 2 = dimensionless axial-energy-flow func-
t* R = radius of the tube (capillary) tion, equation (7)
Bj = a constant defined by equation (37) 1i = the set of real numbers p = fluid density
Cj+, Cj~ = integration constants corre- r = radial variable (cylindrical coordinates) 0 = any function of r\ in "Hi
sponding to Xj~ and A ; + , respectively, T = fluid temperature 0i> 02, 03 = the first, second and third com-
equations (26) and (27), respectively. Tb = fluid bulk temperature ponent, respectively of 0
cp = specific heat of a fluid at constant Te = a given entrance-temperature distri- 0 = any three component vector in Ji
pressure bution, equation (43) 0j> 0j + > 0 j ~ = the three component eigen-
D = the domain of operator Li, equation Tf = ambient temperature vectors corresponding to Ay, Xj+, and A/~,
(17) To = fluid temperature in the reservoir respectively
Ek = the integral defined by equation (37) Ta = fluid temperature on the tubular wall 0;'i> 0j'2 = the first and second components,
F = the three-component-vector solution, V = some characteristic velocity (a con- respectively, of 0y-
equations (14) and (15) stant) 0/'i~i 0j'2~ = the first and second components,
f = a function in Ji, equation (22) v = dimensionless axial velocity of the fluid respectively, of <t>j ~
ft, Jix, Jfi = Hilbert spaces vz = axial velocity of the fluid \j/ = any function of r\ in "Hi
h = heat transfer coefficient for the tubular z = axial variable \j/ = any three component vector in ft
wall resistance. tjk = a constant, equations (34) and (42) \j/j = a function of r), equation (34)
k = thermal conductivity of the fluid \pj' = non-normalized \pj
Li = the linear matrix differential operator ( , ) = the inner product of equation (16)
of equation (14); see also equation (15) i] = dimensionless r = ( , ) = the inner product of equation (33)
M( ,,) = Kummer's function; the three blank R, || • II = the norm that corresponds to < , )
spaces are for its first, second and third 9 = dimensionless T(= (T - Tf)/(T0 - | -| = the norm that corresponds to ( , )
arguments, respectively, equation (45) Tf))

430 / VOL 103, AUGUST 1981 Transactions of the ASME

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 temperature at z = 0 for this problem will be lower than To but while from equations (7) and (6a) we obtain,
somewhat higher than the fluid temperature of the Neumann-Robin
problem. Also, the higher the Peclet number the lower the above limS(f,?/) = 0, all?; (11)
temperature differences. Thus, for high Peclet numbers, the solutions
of the three problems (i.e., the Neumann-Robin, the Danckwerts- 2 ( f , 0 ) = 0, allf (12)
entrance-condition and the uniform-temperature-entrance problems)
will be the same. We have now recast the problem into the pair of partial differential
equations (8) and (10) in the two functions 9(f, TJ) and 2(f, 71) with
Although entrance boundary condition (1) is used for the calcula-
boundary conditions given by (4,5,11) and (12). Note that (12) implies
tions presented in this paper, the generality of our solution method-
(3) 3 and (11) implies (6a) so that equations (3) and (6a) need not be
ology is demonstrated by solving the problem with other possible
included. Since it is important for the mathematical formalism to
entrance conditions. These entrance conditions may include a spec-
follow, we observe from (4) and (8) that
ification of the axial flux or the temperature at z = 0.
The selfadjoint formalism presented here is useful for establishing
- S ( f , 1) = 2 — (f, 1) = - 2 B 9 ( f , 1) (13)
the fact that the eigenvalues involved are real and an orthonormal df ot)
basis is available in the space in which the differential operator is
defined; yet, the solution to the boundary value problem is obtained Equations (8), (10), and (13) now may be written in matrix notation
as an expansion in terms of a part only of the above complete set of as
functions. The coefficients of expansion are however obtained by an Pe 2 d
application of the Gram-Schmidt orthogonalization process, which ~9(f,r>f Pe 2 t>(77) 0 9(f, 7,)
277 d77
makes it unnecessary to solve an infinite set of algebraic equations.
The selfadjoint formalism presented here naturally accommodates d
the entrance boundary condition, whether we use the Danckwerts S(f, 1?) •= 21,— 0 0 S(f,i/) (14)
df or)
boundary condition or the temperature or flux specification.
2(h) Formalism. In dimensionless form, the energy equation
and the boundary conditions, assuming constant physical properties, _2(f, D_ 2Blim 0 0 _2(f, D„
r-l
negligible viscous dissipation, axisymmetry and in view of the above
discussion, are given by, Since equation (13) has already accounted for boundary condition
1 d29 1 d / d9\ , d9 0 < i) < 1 (4), it is now possible to state the boundary conditions for (14) as given
by (5,11) and (12). Some notational brevity is achieved by denoting
— h — \ + v(.v) — = o (2)
by Li the operator in matrix form in equation (14). Also we denote
Pe2df2 jidvVdvl df 0<f<o
d9 by F(f) the three component vector appearing on both sides of
lim — = 0 , all f (3)
?—D or/ equation (14). Note that for each f this vector may be looked upon
as an element of the linear space ft = fi\ ® ft-i ®Jl, which consists
— (f, 1) + B9(f, 1) = 0, allf (4) of ordered triplets of elements from the respective spaces fii, "Hz and
31, the first two representing the spaces of functions, /(??) and g(rj)
defined in [0,1], such that the Lebesque integrals §\ prjdr) < «> and
• — ; — 9 ( 0 + , T J ) + v(v)Q(0+, J?) = U(J?), all 7/ (5)
fog2dr)/ri < <*>, respectively, while 31 represents the space of the real
Pe*1 df
numbers. Equation (14) is now written as
9 = bounded, f —*• °°, all 17 (6)
Since axial and radial fluxes become zero for z —• °°, boundary con- F(f) = LjF(f) (15)
dition (6) implies that, df
The solution of the problem is now reduced to that of solving (15)
lim 9 = 0, all ?? (6a) subject to boundary conditions (5,11) and (12).
The most interesting aspect of the operator Li is that it gives rise
The dimensionless quantities are defined in the Nomenclature. to a selfadjoint problem although the original convective diffusion
The motivation for the decomposition of the energy equation and operator is nonselfadjoint. Indeed, if we define the inner product
the physical arguments used for the definition of the axial energy flow between two vectors
function may be found elsewhere [4, 8,9j. In dimensionless form, the
axial energy flow function is defined as,
<t>i(n)~ Hv)
2(f )s 1 d9
*" JT 2ri'dti' (7) fabi), * = foiy)
Now the convective-diffusion process may be looked upon as a pair _ 03 _ _ f3 _
of first order differential equations describing how the axial energy 'i f 4 r
flow varies with f and r\. Thus, we have J r—r0iO;)tf'i(i/)»7 + 02(1)^2(1)-
0 [Pe2 77
dy + — fats

d2 d9 (16)
(8) and the following domain for Li,
df di?
D(Li) = | 0 e # : L i 0 (exists and) Oi, <t>3 = <fe(l), <fe(0) = 0]
which expresses the fact that the axial change of energy flow results
from radial conduction and (17)

d2 it is shown in [4] that Li is a symmetric operator in the Hilbert space

2r? (9)
dr; Pe2df
which is obtained by differentiation from (7) and expresses the fact
3
that the radial change of energy flow results from both axial convec- Indeed, from (8), 2 lim — = lim , the right-hand side of which, using
tion and conduction of heat. Equation (9) may be rearranged to read
as L'Hospital's rule, equals to: lim lim= 1— — which by virtue of (7) and
,-.o37)df„-.o dfdT)
d9 Pe 2 d2
— = Pe 2 u(i,)9 (10) thus of (10) and (12), equals to lim — 2T? I - —-2 —- + u(r))9 = 0.
df 27) dV ,-0 af I \ Pe Q? I\

Journal of Heat Transfer AUGUST 1981, VOL. 103 / 431

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 J4 of interest. Thus the selfadjoint (or Sturm-Liouville) eigenvalue while from equation (5) we obtain,
problem is given by
2(0+,)/)= C" 2v'v(v')dri' (32)
Li<l>j = \j<t>j (18)

from which we may obtain, Since 2 ( 0 + , rj) is known from (32), all that is required is to expand
Pe 2 it in terms of the set \(l>j2~(v)) for the determination of the Aj~'s. This
Pe2v(v)<t>ji(v) -—-4>jAv) = \<l>ji(v) (19) can be done by generating from \(j>j2~(r])\ an orthonormal set \i/j(r])}
by the Gram-Schmidt procedure (see, for example, Courant and
(20) Hilbert [17], pp. 3, 50). Note that [<l>j2~(v)} is a linearly independent
set, as may be readily inferred from the expansion theorem (22) and
Therefore, the eigenvalue problem (18) implies, through (19,20), and
the linear independence of the set \<t>j~(t})). The natural inner product
(17), that to use for the Gram-Schmidt process is

- v(n) 'l 1
Xy^l = 0 (21)
J
:
(ritji'y +
Pe2 (</#) 0^-djj (33)
o n
subject to the boundary conditions $yi(0) = 0, </>ji'(l) + B0yi(l) = 0. that corresponds to Hilbert space fi2- It is a simple computational
Li is neither positive definite nor negative definite and in fact pos- procedure to determine constants Ujk,j, k = 1, 2, 3 , . .| such that
sesses both positive eigenvalues (Xy+| with corresponding eigenvectors
{<j>j+\ and negative eigenvalues {Ay~| with eigenvectors |0y~). The two $j(v) = £ (jk4>k2 (v) (34)
sets of eigenvectors normalized according to (16), together constitute k-i
an orthonormal basis in ft. It is worth noting that the eigenvalue as it will be seen subsequently. Since the set \\pj(r])\ is an orthonormal
problem (21) is the same as the eigenvalue problem obtained in the basis in 3i2, we have
past [12] without recognizing the symmetric structure. The analysis
herein justifies rigorously the implicit assumption made in [12] that 2(o+, v) = E (2(o+, „), ^)<M») = E By £ ^ f e r w
the X;- 's are real since they are in fact the eigenvalues of a selfadjoint y-i ;=i *-i
problem.
Before we proceed to the solution of the problem, the following = E 4>h2-(ri) E ByeyA (35)
k=i j=k
expansion theorem is obtained from the eigenvalue problem, for any
vector ieJt, Hence, from (31) and (35) we have
+
(Ui ) <f,*y> Ak' = E Bye;* (36)
t- E (22)

where
where
(23) By = (2(0+, n), i/'y) = E 6,7,(2(0+, n), <t>m-{r\)) = E ijkEk
'•^(<t,<t>) k=l h=l

2(c) The Solution. The solution of the problem, F(f), is ob- (37)
tained in the form of series (22). To that effect the inner products
The evaluation of Ek is discussed in detail in [4].
indicated in the expansion coefficients of (22) must be obtained. Thus,
since Fe-D(Li) we have, Now we turn to the evaluation of the set jey/,). From the Gram-
Schmidt procedure we obtain for the non-normalized i/'y,
( L x F , ^ ) = (F,L 1 0 J > (24)
W = 4>j2 - E (0/2 , \Mi/v (38)
and now using (15) and (18), r-l
and then
— <F,0y) = Ay<F,0y> (25)
fj = (39)
from which we obtain, separately for the positive and negative ei-
genvalues, where the norm in (39) derives from the inner product (33). Now, from
<F,*,-+> = C ; + e V - f (26) (38) and (34) we obtain
j-i r
<F,0y"> =Cj-e*rt (27) fj' = <t>j2 - E E fr S (*j2 _'a2 ) E £rk<Ph2 (40)
r=l s = l k=l
but since the solution is bounded for all f,
The summation range covered by r and k in (40) may be modified to
Cj+ = 0, all; (28) read as,
j-i _j-l r
and also
W =
<t>j2 ~ E 0*2 E frk E ers(<t>j2~ (41)
k=l r=k s=l
Cj- = < F ( O + ) , 0 y - ) s A J - | | 0 y - | | 2 (29)
Thus, by virtue of (34) and (41), we obtain
Equation (28) expresses the fact that the eigenvectors that correspond 1
to the positive eigenvalues are orthogonal to the solution, and thus ,k = j , where i/'y' is given by equation (41)
to F(0+). From (22) and (26-29) now we have <jk my-i r
~ E irk E e«(0;2 , <Ps2 )l\$j'\,k = 1, 2, . . . ,;' - 1
(30) r=k s=l

(42)
where the expansion coefficients Ay must be determined. From (30)
for f = 0 and for the second vector components, we have Equation (42) is a recursive formula for the computation of the set
\(jk\ because we know en = 1/jtf>i2—|. Considered as an infinite di-
mensional matrix, jey^j is a lower triangular matrix in which the cal-
2 ( 0 + , v)=Z Aj-0j2-(7j) (31)
culation of each row requires all the rows above it.

432 / VOL. 103, AUGUST 1981 Transactions of the ASME

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 The computation of the integral {<j>j2~, <t>S2~) is discussed in [4]. The i i r
defining equations for the bulk temperature, the Nusselt number and
the asymptotic Nusselt number may be found in [4] or [8]. Expres-
sions, suitable for computations, for the above quantities may be
readily derived from their defining equations and equation (30)
[4].
2(d) Other Entrance Conditions. If instead of the Danckwerts
entrance condition (1), the entrance temperature Te(r) is prescribed
at 2 = 0, the analysis up to equation (30) remains the same. From the
first vector components of (30) at f = 0 we would have

ee(v)s 9(o+, v)=t Aj-^riv) (43)

where Qe(r]) is the dimensionless Te(r). For the determination of the
Aj~'s from (43) the Gram-Schmidt process must now be performed
on the set |0yi~(jj)) with the inner product that corresponds to Ji\.
If the axial flux is specified at the entrance, from equation (30) we
obtain,

^ ( 0 + , 7 7 ) = EX;-Ay-^i-(7)) (44)

and again the Aj~'s are determined from (44) with the Gram-Schmidt
process performed on the set |$yi~(?/)}.

3 Computational Aspects
For the parabolic velocity profile the eigenvectors are computed I i i i i
using Kummer's functions [18], after transforming equation (21) into 0 0.2 0.4 0.6 0.8 1.0
Kummer's equation [4, 8, 9, 18]. Thus, we have,
1
2 2 Fig. 1 Radial profiles of dimensionless temperature for various axial dis-
0;i-(7)) = exp ( - ^ - 7, j M(aj-, 1, W - r / ) (45) tances; Pe = 5

where
P®=20
e.-Ali-HC-J^) (46) PARAMETERS
;
2I 2 2Pe 2 /
nr = ( - * n 1 / 2 (47)
while the eigenvalues are obtained from the characteristic equa-
tion,

(B - nj-)M(aj-, 1, jxj-) + 2fij-M'(aj-, 1, w ~ ) = 0 (48)

obtained from boundary condition 0yi'(l) + B$ ; i(l) = 0, the real

positive roots of which provide the /ij ~ 's. The computation of the
eigenvalues has already been widely discussed in [4, 8, 9]. The ex-
pansion coefficients Aj~ of the problem are computed according to
equation (36). It may be seen from the process of computing the set
je/fc), that the computational effort increases much faster than linearly
with the number of expansion coefficients to be computed. Thus, 11
expansion coefficients may be computed within 3-4 CP s on a CDC
6600 computer, 13 of them within 4—5 CP s and 15 of them within 8-12
CP 6600. Finally, the computation of 168 temperature points, typically
used for either of the Figs. 1 and 2, required 1-1.5 CP 6600 s. It will
be seen that the use of 11 terms for solution (30) provides an accuracy
of better than 0.05 percent. It should be evident then that the ana-
lytical solution presented herein is also computationally very effi-
cient. I i i I I I
0 0 . 2 0.4 0.6 0 . 8 I.O
4 Results V
A number of negative eigenvalues has been presented in [12] for
Fig. 2 Radial Profiles of dimensionless temperature for various axial dis-
three different Biot numbers and various Peclet numbers. Further tances; Pe = 20
eigenvalues with their corresponding expansion coefficients may be
found in [4]. Table 1 presents the expansion coefficients (computed percent. The corresponding maximum difference between 11 and 15
with ;' = 13) for Pe = 5 and B = 1. term computations was found to be less than 0.03 percent. It may be
A careful study of the construction of the expansion coefficients concluded then that a solution with 11 series terms provides final
may reveal that, although the addition in the construction process of accuracy of better than 0.05 percent with minimal computation time.
one or two terms may alter all the previous expansion coefficients, the The fast convergence of the series solution is primarily due to the fast
changes are such that the net effect on the expansion of a quantity decrease of the expansion coefficients; however, the decrease of the
is minimal beyond a number of terms. Indeed, the maximum differ- non-normalized eigenfunctions themselves [4] contribute to the fast
ence (usually in the entrance and the center of the tube) on the final series convergence in view of equation (30). An interesting point to
computations with 14 and 18 terms was found to be less than 0.008 note is that, unlike the Dirichlet [8] and the Neumann [9] problems,

Journal of Heat Transfer AUGUST 1981, VOL. 103 / 433

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 Table 1 Expansion coefficient (Aj ~) corresponding
to the eigenvalues Xy - for the Robin problem with
Danckwerts entrance boundary condition;
Pe = 5, j = 13
;' Aj~
l 1.03234
2 -0.06927
3 -0.02683
4 0.02526
5 -0.01625
6 0.01017
7 -0.00656
8 0.00440
9 -0.00305
10 0.00218
11 -0.00158
12 0.00117
13 -0.00087

there exist two consecutive expansion coefficients with the same sign,
except for Pe = °°. Finally, note that for Pe = °° the expansion coef-
ficients [4] are the same with those of the classical Graetz problem
with Robin boundary conditions (see, for example [19]). Indeed, as
discussed already in the introduction, for Pe = <» the Danckwerts
boundary condition is equivalent to the uniform temperature
boundary condition.
Figures 1 and 2 show radial temperature profiles for various axial
distances, computed according to equation (30) for Pe = 5 and 20
respectively, both with B = 1. It can be shown, from further compu-
tational results in [4], that axial heat conduction may be neglected
for Pe > 30.
Unlike the problem with Neumann boundary conditions [9], the
length of the heating section becomes less important a factor in the
analysis, since for dimensionless axial distance 1, the dimensionless
temperature is almost zero. A heat exchange section of length 5-20 Tl.OO 0.20 0.40 0.60 0.80 1.00 1.20
times the tube diameter is neither long nor unusual indeed particu- C
larly for the small Biot number used here. This feature derives from Fig. 3 Dimensionless bulk temperature Qb as a function of the axial distance
for Pe = 5, 10, and oo
the nature of the boundary condition and it also exists for the problem
with Dirichlet boundary conditions [8]. 2 X 2 operators in [8, 9]; in the latter, the wall boundary conditions
Figure 3 shows the dimensionless bulk temperature 9b as a function were accounted for by the proper definition of the domains of the
of the dimensionless axial variable, for three Peclet numbers, 5,10 operators. This non-trivial extension of the operators in [8,9] to obtain
and <». The bulk temperature behavior shows a remarkable averaging Li we found most instructive for the definition of the operators in the
of the axial conduction effect that may lead one to believe axial con- conjugated Graetz problems [21].
duction not to be important for Peclet numbers higher than 10; the The Dirichlet-Dirichlet [8] and Neumann-Neumann [9] problems
latter is not true as it was discussed above. Figure 4 shows a plot of had the unique feature that the wall-boundary conditions of the two
the Nusselt number as a function of the dimensionless axial distance sections (z > 0, z < 0) of the tube although physically different, were
for three Peclet numbers, 5,10 and 20. As it has been discussed in [20], of the same type mathematically. This allowed analytical solutions
the Nusselt number at f = 0 becomes unbounded for Pe -» <». How- to be obtained in terms of orthogonal functions, which resulted in the
ever, the Nusselt numbers at f = 0 are finite for all finite Peclet coefficients of expansion to be computed in the usual simple form of
numbers, decreasing with decreasing Peclet numbers. In the contrary, the classical Graetz problems. On the contrary, the Robin-Robin
the asymptotic Nusselt numbers increase with decreasing Peclet problem (and all problems with a different-type boundary condition
numbers. This creates the crossing of the Nusselt-number curves on the wall of each section of the tube) lead to mixed-boundary
shown in Fig. 4. Note also that for Pe < 10 the Nusselt number curve problems, where the convenience of an analytical solution is lost. Yet,
shows a maximum occurring at an axial distance which increases with the semi-infinite problems, like the Robin problem presently dis-
decreasing Peclet numbers. This fact is observed for the first time in cussed, lend themselves to technically analytical solutions, albeit not
a Newtonian Graetz problem. as efficient as those of [8, 9]. The nicety of these solutions derives from
Finally, Fig. 5 presents a plot of the asymptotic Nusselt number the fact that no variational techniques and/or infinite sets of algebraic
(NUQ S ) as a function of the Peclet number. Thus, for Pe ->- 0, Nu o s -» equations must be used as in the past [11,12]. The implications of the
5.53 and for Pe -> °>, Nu a s - * 4.125. It is shown in [4] that the as- above is a solution with an improved accuracy and computational
ymptotic Nusselt number is independent of the entrance boundary efficiency. Avoiding the use of variational techniques in handling the
condition. pointwise Danckwerts entrance boundary condition must be viewed
as a particular accomplishment of the presented formalism. This
5 Discussion derives from the fact that the pointwise Danckwerts condition is a
The extended Graetz problems with a semi-infinite axial domain linear combination of the heat flux and the temperature (equation
and either the Dirichlet or the Neumann boundary condition on the (5)), with a coefficient of combination (v(rj)) which is a function of the
wall may be treated in an entirely analogous manner. The definitions radial distance.
of the proper matrix differential operators that correspond to Li, may
be found in [8] and [9], respectively. What distinguishes Li from the Acknowledgment
operators in [8, 9] is that the wall boundary condition (4) has been One of the authors (E.P.) is indebted to Purdue University, School
incorporated into the differential expression of Li (equation (14)) by of Chemical Engineering, for the financial assistance for the duration
virtue of equation (13). This incorporation is necessary for a selfad- of this research. The computer facilities were generously provided by
joint formalism. L,! is thus a 3 X 3 matrix operator in contrast to the Purdue and Rice Universities.

434 / VOL 103, AUGUST 1981 Transactions of the ASME

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 10 20 30 40 50
PECLET NUMBER
Nil Fig. 5 Asymptotic Nusselt number as a function of Peclet number

Ducts. A Source Book for Compact Heat Exchanger Analytical Data. Advances
in Heat Transfer, Irvine, T. F. Jr. and Hartnett, J. P., Eds., Academic, New
York, 1978.
8 Papoutsakis, E., Ramkrishna, D., and Lim, H. C, "The Extended Graetz
Problem with Dirichlet Wall Boundary Conditions," Applied Scientific Re-
search, Vol. 36,1980, p. 13.
9 Papoutsakis, E., Ramkrishna, D., and Lim, H. C , "The Extended Graetz
Problem with Prescribed Wall Flux," AICHE Journal, Vol. 26,1980, p. 779.
10 Ramkrishna, D., and Amundson, N. R., "Boundary Value Problems in
Transport with Mixed and Oblique Derivative Boundary Conditions. II. Re-
duction to First Order Systems," Chemical Engineering Science, Vol. 34,1979,
p. 309.
11 Pirkle, J. D. and Sigillito, V. G., "A Variational Approach to Low Peclet
Number Heat Transfer in Laminar Flow," Journal of Computational Physics,
Vol. 9,1972, p. 207.
12 Hsu, C. J., "Exact Solution to Entry-Region Laminar Heat Transfer with
Axial Conduction and the Boundary Condition of the Third Kind," Chemical
Engineering Science, Vol. 23,1968, p. 457.
13 Pirkle, J. C , and Sigillito, V. G., "Calculation of Coefficients in Certain
Fig. 4 Nusselt number as a function of the axial distance for Pe = 5,10, and Eigenfunction Expansions," Applied Science Research, Vol. 26, 1972, p.
20 105.
14 Michelsen, M. L., and Villadsen, J., "The Graetz Problem with Axial
Heat Conduction," International Journal of Heat Mass Transfer, Vol. 17,1974,
References p. 1391.
1 Middleman, S-, Transport Phenomena in the Cardiovascular System, 15 Sorensen, J. P., and Stewart, W. E., "Computation of Forced Convection
Wiley-Interscience, New York, 1972. in Slow Flow Through Ducts and Packed Beds.—I. Extensions of the Graetz
2 Leonard, E. F. and Jorgensen, S. B., "The Analysis of Convection and Problem," Chemical Engineering Science, Vol. 29,1974, p. 811.
Diffusion in Capillary Beds," Annual Review of Biophysics and Bioengineering, 16 Wehner, J. F., and Wilhelm, R. H., "Boundary Conditions of Flow Re-
Vol. 3,1974, pp. 293. actor," Chemical Engineering Science, Vol. 6,1956, p. 89.
3 Davis, E. J., Cooney, D. 0., and Chang, R., "Mass Transfer Between 17 Courant, R., and Hilbert, D., Methods of Mathematical Physics, Vol.
Capillary Blood and Tissues," Chemical Engineering Journal, Vol. 7,1974, I, Interscience, New York, 1953.
p. 213. 18 Slater, S. L., Confluent Hypergeometric Functions, Cambridge Univ.
4 Papoutsakis, E., "The Solution of Boundary Value Problems in Diffusive Press, Cambridge, 1960.
Heat and Mass Transfer with Convection. Functional Analytic Methods," Ph.D. 19 Cooney, D. O., Kim, S. S., and Davis, E. J., "Analysis of Mass Transfer
dissertation, Purdue University, W. Lafayette, Ind., 1979. in Hemodialyzers for Laminar Blood Flow and Homogeneous Dialysate,"
5 Vrentas, J. S., and Duda, J. L., "Flow of a Newtonian Fluid Through a Chemical Engineering Science, Vol. 29,1974, p. 1731.
Sudden Contraction," Applied Scientific Research, Vol. 28,1973, p. 241. 20 Papoutsakis, E., "Nusselt Numbers near the Entrance of the Heat-
6 Ybarra, R. M., and Eckert, R. E., "Viscous Heat Generation in Slit Flow," Exchange Section in Flow Systems," AICHE Journal, 1981, in press.
AICHE Journal, 1980, Vol. 26, 751. 21 Papoutsakis, E., and Ramrishna, D., "Conjugated Graetz Problems. II.
7 Shah, R. K., and London, A. L., Laminar Flow Forced Convection in Fluid-Fluid Problems," Chemical Engineering Science, 1981, in press.

Journal of Heat Transfer AUGUST 1981, VOL 103 / 435

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 Effects of Axial Conduction and
B. Vick
M. N. Ozifik
Conwectiwe Boundary Conditions in
Mechanical and Aerospace Engineering
Department,
North Carolina State University,
Slug Flow inside a Circular Tube
Raleigh, N. C. 27650

An exact solution is presented for the temperature distribution and the local Nusselt
number in slug flow of a low Peclet number fluid inside an infinitely long tube by includ-
ing in the analysis the axial conduction in the fluid and convection from the walls for both
the upstream and downstream regions. It is shown that the Peclet number and the relative
magnitudes of the upstream and downstream Biot numbers significantly effect the local
Nusselt number and the Wall and mean fluid temperatures in the thermal entrance re-
gion. A particularly interesting result of the present investigation is the crossover of the
wall and the mean fluid temperatures in such a manner as to give rise to a negative Nus-
selt number in the thermal entry region. Depending on the Peclet number and the relative
magnitudes of the upstream and downstream Biot numbers, the negative Nusselt number
can occur in the upstream or downstream region.

Introduction 0l(£,-°°) = O,02(£,°°) = O dd,e)

The relative importance of axial heat conduction in flow inside a
conduit depends on the magnitude of the Peclet number. For example, where i = 1,2, and throughout the analysis the subscript i = 1 will refer
the axial conduction is negligible in comparison with radial conduction to the upstream region (?) < 0) and i = 2 to the downstream region (??
for flow inside a tube if the Peclet number is above 100. However, for >0).
liquid metals the Peclet number can be very low, hence the axial The energy equations (la) are coupled at t? = 0 with the require-
conduction can be important. The effects of axial conduction in heat ment that the temperatures and heat fluxes are continuous. These
transfer for liquids flowing inside tubes have been studied by several coupling conditions are given by
investigators [1-17]; but in all these investigations the problem was 1 - fli(fcO) = fl2(f ,0) (2a)
solved for very simple boundary conditions, such as a prescribed wall
heat flux for the downstream region and insulated wall condition for d0x(£,O) d02(£,O)
(2b)
the upstream region. However, when the upstream and downstream d); dr)
regions are subjected to convection with external environments at
In the foregoing formulation, various dimensionless quantities are
different temperatures, the relative magnitude of the Biot numbers
defined as
for the upstream and downstream regions and the Peclet number are
expected to influence the temperature distribution, hence the local a a- N •^1" _ T\{r,z)
Nusselt number, in the entrance region. Therefore, the objective of flitt.fl)"— ,02«,i?)
J 1» - J2-
this work is to investigate the effects of the Biot and Peclet numbers
T2(r,z) - T2~ a 2r0U
on heat transfer at the entrance region of a circular tube. . Pe ,
Ti„ - TV a
Analysis
Hi — ,H2 = ——,£-—,v-— (3)
We consider incompressible, constant property fluid in slug flow k k 7"o Pe
inside a circular tube of radius r0, lying in the axial domain — °° < z The method of separation of variables is now applied to solve the
< <=°. The tube has negligible wall thickness so that the axial conduc- coupled heat transfer problem defined by equations (1) and (2). Let
tion along the tube walls can be ignored. In the upstream region (z < Rimd) and Zjm(ij) be the separated functions in the radial variable
0), the environment is at a temperature Ti„, with a heat transfer £ and the axial variable r], respectively. The solution for the temper-
coefficient hi. In the downstream region (z > 0), the environment is atures 6i (£,??) are constructed in the form
at a temperature T2a, with a heat transfer coefficient h2. Figure 1 il-
lustrates the geometry and coordinates. We assume a slug flow model i = 1 for i) < 0
(4)
for the velocity field. For liquids having very low Prandtl number, the i = 2 for j) > 0
temperature profile develops much faster than the velocity profile,
Here the functions Rim(i), (m = 1,2,...), satisfy the following ei-
hence the slug flow model is meaningful for the thermal entrance re-
genvalue problem
gion problems. The mathematical formulation of this heat transfer
problem, for the upstream (?j < 0) and downstream (?j > 0) regions, 1 _d_ L dR,
is now given in the dimensionless form as + ftm2fl,-m(£) = 0 , i n 0 < £ < l (5a)
fd£\ df
dRim(0)
= t — ^ in0<f <1 (la) =0 (5b)
2 dv £d£\ d£/ Pe 2 d7j 2
d6i(0,ri) dRim(l)
0 (lb) + Hi Rim(l) = 0 (5c)
d*
where i = 1,2. The orthogonality condition for the eigenfunctions
+ Hi 0i{i,v) = o (lc) ^im(?)isgivenby
a*
0 for m ^ n
Contributed by the Heat Transfer Division for publication in the JOURNAL
CkRim(l;)Rin(&dt = (6a)
Jo Nifiim) for m = n
OP HEAT TRANSFER. Manuscript received by the Heat Transfer Division
January 16,1981. where the normalization integral N(Pim) is defined by

436 / VOL. 103, AUGUST 1981 Transactions of the ASME

Copyright © 1981 by ASME
Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 w(ftm)= C k\Rim(m2 da (66) CONVECTION INTO MEDIUM CONVECTION INTO MEDIUM
AT T I C 0 WITH HEAT TRANSFER AT T 2a ^VITH HEAT TRANSFER
Jo COEFFICIENT h . COEFFICIENT h 2
The sets of functions, R\m(Q and fl2m(^), are further related by the
following relation (see appendix for proof)
Ht~H2
I1**"® R2 (H)dt m =
•02,
Rin(DRim(l) (la) (DOWNSTREAM)

which is valid for all m and re if H i 5= H 2 . For the special case of H i

= H 2 , the eigenvalue problems given by equations (5), for i = 1 and
2, become identical. Then we have Fig. 1 Geometry and coordinates

X' {fll»(f)K2m«)df

0 for m ^ re
(7ft)
UV(0im) = W(fen) = N ( f t J for rei = re (4), take the form
which is valid for H i = H2 = H. • 1 for 7) < 0
The eigenvalue problem given by equations (5) is Bessel's equation 0.-(£,u)= E Cto.eW'.Mftma (11)
•• 2 for ij > 0
of order zero and its solution satisfying the boundary condition (56)
is taken as If these solutions should satisfy the coupling conditions at ?? = 0,
equations (2a,6) lead to
RimiO = MPimt) (8a)
If this solution should satisfy the boundary condition (5c), the ei-
E E c,-mJ0(ftm£) = i (12a)
i=l m= l
genvalues ftm, (m = 1,2,3,...), should be the solutions of the following
transcendental equation E E C imTim Jo(ft m ?) = o (126)
( = 1 771 = 1

- A m Jl(Pim) + Ht J o ( f t m ) = 0 (8ft) The expansion coefficients, C im >

The normalization integral, iV(0, m ), is readily determined as can be determined from equations
(12), by utilizing the orthogonality property of the eigenfunctions,
iV(ftm)=-[J02(ftm)+Ji2(ftm)] (8c) as now described. We consider the analysis for the cases H i = H 2 =
H and Hi ^ H 2 separately.
For the special case of H i = 0, 0m = 0 is also an allowable eigenvalue, The Case Hi = H% = H. For this special case, the eigenvalue
corresponding to the eigenfunction flio(£) = 1 with the resulting problems, given by equations (5), are identical for i = 1 and 2, hence
normalization integral N(0iO) = i- However, the possibility H 2 = 0 we have
01 : 02m = Prr, (13)
need not be considered as it leads to a trivial solution regardless of the
value of H i . We now operate on both sides of equations (12a) and (126) with the
We now turn our attention to the separated functions Z,m(?j), (m operator
= 1,2,. . .), in the axial variable T). It can be shown that the functions
Zimiv) satisfy the differential equation J^€«/o(ft,|)d$,
1 d2Zim 1 dZim 1 for r\ < 0
+ -—. + Pirn Zimim) = 0 (9a) and utilize the orthogonality relations (6a, 76) to obtain, respec-
Pe 2 a V 2 dv i = 2 for r) > 0 tively,
subject to the conditions
(14a)
Zim(-°°)=0,Z 2 m (<») = o (96,c) i=l Pn

The solution of equations (9a) satisfying the conditions (9ft ,c) can be E C;„ yin = 0 (146)
expressed as 1=1

Zim(y) = eyi»>\i = 1,2 (10a) Here we made use of the transcendental equation (86) to obtain the
result given by equation (14a).
where the exponents 7; m are given by A simultaneous solution of equations (14) yields the expansion
1L coefficients as
Jim : Pe 2 (-1)
V'•£«-•
Having established the functional forms of the separated variables
(106)
Cin = ( - 1 ) '
73- HMPn)
lTin-72jlfti2AT(0re),
, a = 1,2) (15)

Rim(& and Zimir]), as given by equations (8a) and (10a), respectively, which is valid for Hi = H 2 = H. The normalization integral, N(j3n)
the dimensionless temperature functions #((£,77), given by equation is given by equation (8c).

••Nomenclature.
Cim = expansion coefficients r0 = tube radius 81, (77) = dimensionless bulk temperature de-
Hi = Biqt numbers Ti(r,z) = temperature functions fined by equation (25)
hi = heat transfer coefficients Ti,b(y) = bulk temperatures Si,b(v) = dimensionless bulk temperatures
k = thermal conductivity T;«, = environment temperatures 6i (rj) = dimensionless temperatures
JV(ftm) = normalization integrals U = fluid velocity 8w (v) ~ dimensionless wall temperature de-
NU(JJ) = local Nusselt number Zim (v) - separated axial functions fined by equation (24)
z = axial variable
Pe =^ = Peclet number a = thermal diffusivity ri = = dimensionless axial variable
a Pe
Pim = eigenvalues
Rim (£) = separated radial functions 7,m = exponents given by equation (106)
r
£ = — = dimensionless radial variable
r = radial variable ro

Journal of Heat Transfer AUGUST 1981, VOL. 103 / 437

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 The case H = 0 need not be considered, since it leads to a trivial for n = 0, is rearranged so that the term Hi//3io 2 appears on the
solution as discussed previously. right-hand side and H i is replaced by /5iot/i(|8io)M}(Ao) according
The Case H i ^ Hi. For this more general situation, we operate to the transcendental equation (8b). The application of L'HospitaPs
on equations (12a) and (12b) with the operator. rule reduces equation (18b), for H i -* 0 and fto - • 0, to

£ SJoiPlnt) dl; E C2m\i-^H-^-ja(hm) = ~ (21)

and utilize the relations (6a, 7a) and (8b) to obtain, respectively, The Nusselt Number. Once the temperature distributions
Sii^tV), (i = 1>2), in the fluid are known, the local Nusselt number,
Hi — Hi
ClnN(Pln)+ E C2m MPln)MP2m) Nu(?7), for the upstream or downstream region, is determined from
L8in2-/32m2 its definition, given by
HiJaWm)
(16a)
H 2Hi 0;(1,TJ)
Nu(rj) = - - (22a)
Hi-H2 BiMv) ~ Oi(l,v) 8ilb(v) ~ Odhv)
Cl„ N(Pln) 7 l „ + E C 2 m 72m Jo(ftn) ^o(ftm) = 0
m=l L8ln2 - / W where the subscript i = 1 for 7) < 0 and i = 2 for t] > 0; and equation
(16b) (lc) is utilized to obtain the alternative form of the numerator. The
dimensionless bulk or mean temperature, 0;,6(»!), is defined as
Similarly, we operate on equations (12a) and (12b) with the oper-
ator
0;,t(») = 2 J^€flitt,u)d£
J^tJotfm&dt
(22b)
and utilize the relations (6a, 7a) and (8b) to obtain, respectively, m=l Pirn

H i — Hi and the dimensionless wall temperature, 0,-(1,J/) is given by

E Ci Jo(Pim)J0(P2n) + C2nN(fcn)
m=l
Bdhv) = E Q, e ™ J 0 ( f t m ) (22c)
H2=/o(/?2n)
(17a)
For the special case of H x = 0, we have fto = 0 as an eigenvalue.
Hi-Hz Then, equation (22b), for the limiting case of H i —• 0, fto —>- 0, is
E Cim 7l ff • W l m ) </0(/32n) + C2„ iV(/?2„) 7 2 „ = 0 evaluated by replacing H i Jo(fto) by fto J1W10) according to the
transcendental equation (8b), and applying L'Hospital's rule. We
(17b)
find
We eliminate C2nN(fiin) between equations (17a,b) and rearrange
«i,6(j?) = C 10 eT«"' = C 1 oc P e V2 (23)
the result in the form •
2
J2n Ho where we utilized equation (10b) to relate 710 to Pe /2.
E Clm|i-2i2t MPlm) •• (18a)
m-l V 7 2 n / V l m ^ - f<2,i^ Hi - H2 R e s u l t s and Discussion
Similarly, the elimination of C l n Af(/3i„) between equations (16a,b) Before presenting numerical results for the temperature distribu-
yields tion and the local Nusselt number, we discuss briefly the physical
Hi significance of the present mathematical modelling in relation to heat
E C 2 m |i-^, Mfam) •• (18b). exchanger applications. Consider a heat exchanger tube having an
ro=l \ 7 l n / \ P l n " ~ P2m^ Hi - H2
unheated (or uncooled) hydrodynamic entrance region confined to
Equations (18a) and (18b), for n = 1,2,3,.. ., provide an infinite set the domain ri < 0, followed by an applied heat transfer region confined
of simultaneous equations for the determination of the expansion to r\ > 0. Considerable amount of heat transfer takes place between
coefficients Cim and C 2 m , for m = 1,2,3 These equations can be these two regions by axial conduction in the fluid through the cross-
written in matrix form as section at i] = 0 if the Peclet number is low. Therefore, the tempera-
ture of the fluid at r\ = 0 is not known a priori. In the present model-
A; C; = B;, i = 1,2 (19) ling, it is assumed that the fluid enters the tube with a temperature
where A; is a square matrix whose elements are given by Ti„ at an axial location sufficiently away from the origin of the axial
coordinate, r\ = 0, where the applied heat transfer section begins. The
7» (20a) boundary condition of the present analysis, for the upstream region,
Ai,rr 1- Jo(ftm)
7 3 - j ' , n / \Plm P2n" 7/ < 0, implies that the temperature of the external ambient is the same
with subscript m designating the column and n designating the as the fluid inlet temperature, Ti„, and that the external Biot number
row. is Hi. At the beginning of the heat transfer section, ?j = 0, a step
B, and C; are the column vectors whose elements are given by change occurs in the external ambient temperature from T\„ to T 2 »,
while the external Biot number for the downstream region, r\ > 0, is
H„-i taken as H 2 . Clearly, several special cases are obtainable from such
Bi,, (20b)
H i - H2 general boundary conditions. For example, H i = 0 represents an ad-
iabatic upstream condition; H2-* 0 0 corresponds to a prescribed uni-
C = C- (20c) form wall temperature for the downstream region; and
where i = 1 or 2. H i = H 2 = H = f i n i t e represents equal Biot numbers for the upstream
and downstream regions while Ti«, ^ T%„.
Once the expansion coefficients Ci„ and C2n are determined from
the solution of the system of algebraic equations (19), the tempera- Numerical calculations are performed to determine the dimen-
tures 0,(£>'?)> (i - 1,2), for the upstream and downstream regions, are sionless bulk temperatures, 8itb(ri), the wall temperatures, 0;(l,?j), and
computed according to equation (11). the local Nusselt number, Nu(?;), for both the upstream (?) < 0) and
A Special Case. When H x = 0, corresponding to an insulated the downstream (7/ > 0) regions, for several different values of the
upstream region, ,810 = 0 is also an eigenvalue. Then equation (18b), Peclet number and the Biot numbers, H i and H2. The temperatures
for n = 0, becomes indeterminate and must be evaluated by L'Hos- 0,-,f,(?j) and 6i(l,r}), given by equations (22b) and (22c) converge faster
pital's rule for H i —>- 0, /?io -»• 0. To evaluate this limit, equation (18b), for longer values of Peclet number and 1i?|; but the convergence be-

438 / VOL. 103, AUGUST 1981 Transactions of the ASME

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 comes slower for larger values of H i and Hi- Except for extreme cases, Nusselt number at the thermal entry region, but this effect is reversed
20 to 30 terms in the series give sufficiently accurate results. as the Nusselt number approaches its fully developed value.
For convenience in the discussion of the numerical results, we Figures 7 and 8 show the effects of the variation of the upstream
consider the external ambient for the upstream region (?) < 0) hot and Biot number on the dimensionless wall and mean fluid temperatures
that for the downstream region (?) > 0) cold; that is T"i„ > Ti*>. for both the upstream and the downstream regions. We note that in
In the presentation of the results for temperatures in graphical Fig. 7, the wall temperature and the mean fluid temperature cross
form, we defined a new dimensionless wall temperature, 6w(ri)t as each other and, by the definition of the Nusselt number, this crossover
point corresponds to an infinite Nusselt number. The region after the
. , ^ (1 - 0i(l,»j) for r, < 0
(24) crossing point, where &bbi) > 8w(y) while the heat flow is still from
l82(l,i?) for»)>0
and a new dimensionless bulk fluid temperature, dt,(r]), as
L - oi.bOj) for ?) < 0
Mu) : (25)
h.b(v) for?/>0 Pe = 2 0 0 ^ ~ ^ " ~ ^ 9b(ij)
9„(,)
The notations 6W (?)) and 8b (?/) are used in all graphical presentations.
The physical significance of the definitions given by equations (24)
H, =0
and (25) is envisioned better by recalling the definitions of 0i(£,?)) and ——^-——_ H2=20
02(f,?)) given by equation (3).
~ ^ ^ - - - P e 3 200
Figure 2 shows the dimensionless wall and bulk fluid temperatures
plotted as a function of the dimensionless axial distance, ?), for Hi = _^-Pe=5
--..^ ^ - - ^
0, H2 = 20, at Peclet numbers 1, 5 and 200 for the upstream region.
- -~ _ —. = = >

Figure 3 is a similar plot for the downstream region. As expected, the ->»>< • ~ ^ - ^ _ ^ ^

effects of axial conduction are strongly pronounced at the thermal 1,, 1 , 1 1 iTTrr-
entry region where lowering the Peclet number lowers the tempera- Z/r.
tures. At high Pficlet number, Pe = 200, the effects of axial conduction Pe

are unnoticeable. Fig. 3 Dimensionless wall and bulk temperatures for the downstream region,
Figures 4 and 5 are prepared to illustrate the effects of the variation ?) > 0 (effects of Peclet number)
of the Peclet number on the local Nusselt number for the downstream
region. Figure 4 is for Hi = 0,H2= 1, while Fig. 5 is for Hi = 0, Hi =
20. We note that in both of these figures, the curves for different Peclet
numbers have crossover regions. These crossing regions are shifted
toward r\ = 0 as the Biot number for the downstream region, H2, is -\ =0
increased. These curves show that at the beginning of the heat transfer
section, r\ = 0, decreasing the Peclet number decreases the local
- V -Pe=IOO
Nusselt number. The reason for such a trend is that as the Peclet
number is reduced, the effects of axial conduction in the fluid become Pt*20K
more pronounced, which in turn causes the development of the tem-
perature profile to begin further into the upstream region. As a result, Pe=5-7

the fluid entering the heat transfer region becomes more thermally
Pe = l - ^
developed, thus reducing the Nusselt number for fluids with a smaller
Peclet number. These figures also show a crossover of the curves, thus
indicating an apparent reversal of the dependence of the Nusselt
number on the Peclet number in the region between the crossover area
and the thermally developed zone. This trend is not representative
of the actual physical situation in the real 2-coordinates, since the true
1
axial distance 2 has been scaled off by the Peclet number. If the curves 10"'
Z/r.
were not stretched in proportion to the Peclet number, such a cross- Pe
over would not have occurred in the graphs. It is also to be noted that
Fig. 4 Local Nusselt number for the downstream region, j) > 0
the crossover of the curves in these figures takes place over a finite
region rather than at a specific point.
Figure 6 shows the effects of the variation of the downstream Biot
number, H2, on the local Nusselt number for the downstream region.
H, • 0
Increasing the downstream Biot number increases the downstream
30 H2 •20
N. X^-—Pe * l
*"-*, "~"^~^-Pe = 200

25
V " b '•/"'
0.8 ""*>. v \ ^ tf.(l)»
V\ \^-—P« s 5 >v

20
V\--Pe=2o\ ^^
0.6 -\
\ ^ — 15
^-Pe=lOO\\ \v
^~- Vpe-S
0.4 \
10
0.2
—^^— ST""
5
, , ! 1

, , 1 1 I • 1 .1 1 I I

,=2Ai
Fig. 2 Dimensionless wall and bulk temperature for the upstream region,
if < 0 (effects of Peclet number) Fig. 5 Local Nusselt number for the downstream region, j) > 0

Journal of Heat Transfer AUGUST 1981, VOL 103 / 439

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 References
1 Singh, S. N., "Heat Transfer by Laminar Flow in a Cylindrical Tube,"
Pe Applied Scientific Research, Section A, Vol. 7, 1957-1958, pp. 325-340.
60 - ^ v H, = 500 H, = 0 2 Agrawal, H. C , "Heat Transfer in Laminar Flow Between Parallel Plates
at Small Peclet Numbers," Applied Scientific Research, Section A, Vol. 9,1960,
50 pp. 177-189.
HZ = K X > ^ 3 Hennecke, D. K., "Heat Transfer by Hagen-Poiseuille Flow in the
Thermal Development Region with Axial Conduction," Wdrme-Stoffube-
40 tragungBd., Vol. 1,1968, pp. 177-184.
H,.20 4 Verhoff, F. H., and Fisher, D. P., "A Numerical Solution of the Graetz
Problem with Axial Conduction," ASME JOURNAL OF HEAT TRANSFER, Vol.
J-30 95.1973, pp. 132-134.
z H2-5
5 Michelsen, M. L., and Villadsen, J., "The Graetz Problem with Axial
Conduction," International Journal Heat Mass Transfer, Vol. 17, 1974, pp.
IV 1 1391-1402.
6 Campo, A., and Auguste, J. C , "Axial Conduction in Laminar Pipe Flows
10
with Nonlinear Wall Heat Fluxes," International Journal Heat Mass Transfer,
Vol. 21,1978, pp. 1597-1607.
7 Hsu, Chia-Jung, "Theoretical Solutions for Low-Peclet-Number
1 1 Thermal-Entry-Region Heat Transfer in Laminar Flow Through Concentric
Annuli," International Journal Heat Mass Transfer, Vol. 13, 1970, pp.
1907-1924.
8 Hsu, Chia-Jung, "An Exact Analysis of Low-Peclet-Number Thermal-
Fig. 6 Local Nusselt number for the downstream region, j? > 0 Entry-Region Heat Transfer in Transversely Nonuniform Velocity Fields,"
AIChE Journal, Vol. 17, No. 3, May 1971, pp. 732-740.
9 Davies, E. J., "Exact Solutions for a Class of Heat and Mass Transfer,"
Canadian Journal of Chemical Engineering, Vol. 51,1973, pp. 562-572.
10 Smith, C. E., Faghri, M., and Welty, J. R., "On the Determination of
Temperature Distribution in Laminar Pipe Flow with a Step Change in Wall
Heat Flux," ASME J O U R N A L OF H E A T T R A N S F E R , Vol. 97,1975, pp. 137-
139.
11 Jones, A. S., "Extensions to the Solutions of the Graetz Problem," In-
ternational Journal of Heat Mass Transfer, Volume 14,1971, pp. 619-623.
12 Jones, A. S., "Laminar Forced Convection at Low Peclet Number,"
Bulletin of the Australian Mathematical Society, Vol. 6,1972, pp. 83-105.
13 Jones, A. S., "Two-Dimensional Adiabatic Forced Convection at Low-
Peclet Number," Applied Scientific Research, Vol. 25,1972, pp. 337-348.
14 Deavours, C. A., "An Exact Solution for Temperature Distribution in
Parallel Plate Poiseuille Flow," ASME JOURNAL OF HEAT TRANSFER, Vol.
96.1974, pp. 489-495.
15 Bayazitoglu, Y., and Ozisik, M. N., "On the Solution of Graetz Type
Problems and Axial Conduction," International Journal of Heat Mass
Fig. 7 Dimensionless wall and bulk temperatures for the upstream region. Transfer, Vol. 23,1980, pp. 1399-1402.
i) < 0 (effects of Hi) 16 Vick, B., Ozisik, M. N., and Bayazitoglu, Y., "A Method of Analysis of
Low Peclet Number Thermal Entry Region Problems with Axial Conduction,"
Letters in Heat Mass Transfer, Vol. 7,1980, pp. 235-248.
bill 17 Vick, B., and Ozijik, M. N., "An Exact Analysis of Low Peclet Number
Heat Transfer in Laminar Flow with Axial Conduction," Letters in Heat Mass
Transfer, Vol. 8,1981, pp. 1-10.

-.0.4
APPENDIX
P r o o f of E q u a t i o n ( 7 a ) . W e w r i t e e q u a t i o n (5a) for Rin(£) and
RlmW
l_d_ .dRin
+ Pm2 Rm(0 = 0 (la)
' d£ .
Fig. 8 Dimensionless wall and bulk temperature for the downstream region, 1 d dRim
»7 > 0 (effects of H,) „_,,., _,„ +/?2m2.R2m(£)=0 < lfe )

t h e h o t s u r r o u n d i n g i n t o t h e fluid, c o r r e s p o n d s t o a negative N u s s e l t W e o p e r a t e on e q u a t i o n ( l a ) w i t h t h e o p e r a t o r Jo £ #2m(£) d £ a n d

n u m b e r . T h e crossover p o i n t is shifted t o w a r d s t h e origin, i\ = 0, as e q u a t i o n (16) w i t h t h e o p e r a t o r So £ Rin(0 d!;, s u b t r a c t t h e r e s u l t s
Hi is increased. Clearly, t h e negative N u s s e l t n u m b e r resulting from and integrate by parts to obtain
such a peculiar t e m p e r a t u r e distribution is caused by axial conduction . dRin\
in t h e fluid. T h e crossover p o i n t m a y occur in t h e u p s t r e a m or (ft*2 - fen2) ^ $RlnRlm d£ = - ^ R
^ J t * d£
d o w n s t r e a m regions d e p e n d i n g on t h e relative m a g n i t u d e s of H i a n d
d$\ d£
H%. W h e n H\ = Hi = <*>, t h e crossover p o i n t s h o u l d occur a t ?) = 0. d j.dRzm.
Since e x p e r i m e n t a l d a t a on l a m i n a r h e a t transfer for liquid m e t a l s
in t h e t h e r m a l e n t r a n c e region a r e meager, no c o m p a r i s o n of t h e
dRim „ dR\n
present results has been m a d e with the experiments. T h e analytical Rln R2 (2)
results p r e s e n t e d here m a y help t o assess t h e i m p o r t a n c e of axial h e a t d£ d£ 1=1
c o n d u c t i o n on h e a t t r a n s f e r in a fluid a t t h e t h e r m a l e n t r a n c e r e - If Hi ^ Hi, t h e r i g h t - h a n d side of e q u a t i o n (2) is e v a l u a t e d b y u t i
gion. lizing t h e b o u n d a r y c o n d i t i o n (5c). T h e n e q u a t i o n (2) b e c o m e s

Hi-Hi
Acknowledgment f^RiniORimi&dt fll„(l)fl2m(l) (3)
T h i s w o r k was s u p p o r t e d in p a r t b y t h e N a t i o n a l Science F o u n -
d a t i o n t h r o u g h t h e G r a n t E N G . 77-12949. w h i c h is t h e r e s u l t given b y e q u a t i o n ( 7 a ) .

440 / VOL. 103, AUGUST 1981 Transactions of the ASME

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 Heat Transfer from a Plate Elevated
abowe a Host Surface and Washed
E. M. Sparrow by a Separated Flow Induced by
Fellow ASME

F. Samie
the Elevation Step
S. C. Lai)
Department ol Mechanical Engineering, Wind tunnel experiments were performed to determine heat transfer coefficients and
University of Minnesota, fluid flow patterns for a thermally active surface elevated above a parallel host surface.
Minneapolis, Minn. 55455
The step-like blockage associated with the elevation causes flow separation and recircula-
tion on the forward portion of the thermally active surface. Four parameters were varied
during the course of the experiments, including the angle of attack of the oncoming air-
flow relative to the surface, the step height, the extent of the host surface which frames
the active surface (i.e., the skirt width), and the Reynolds number. Flow visualization
studies, performed with the oil-lampblack technique, showed that the streamwise extent
of the separation zone increases with decreasing angle of attack, with larger step heights
and skirt widths, and at higher Reynolds numbers. At larger angles of attack, separation
does not occur. The experimentally determined heat transfer coefficients were found to
increase markedly due to the flow separation, and separation-related enhancements as
large as a factor of two were encountered. The enhancement was accentuated at small an-
gles of attack, at large step heights and skirt widths, and at high Reynolds numbers. A
main finding of the study is that the separation-affected heat transfer coefficients are
generally greater than those for no separation, so that the use of the latter may underesti-
mate the heat transfer rates. For an application such as a retrofit solar collector, such an
underestimation of the wind-related heat loss would yield an optimistic prediction of the
collector efficiency.

Introduction
This paper is concerned with the heat transfer characteristics and collector will be mounted atop the roof, so that its upper surface is
related patterns of fluid flow for a plate that is inclined to an air- elevated above the plane of the roof. With respect to the oncoming
stream, with part of the plate surface being washed by a separated, airflow, the thus-mounted collector presents a step-like blockage to
recirculating flow. The physical situation under consideration is de- the flow.
picted schematically in the upper diagram of Fig. 1. As shown there, In light of the foregoing discussion, it is evident that the configu-
a heat transfer surface is elevated above a host surface such that the ration to be studied here serves to model the retrofit installation of
two surfaces are parallel to each other and are inclined to an airflow. a solar collector. Thus, in addition to their basic research content, the
The mounting arrangement of the elevated surface presents a step- present results should have application to solar collector design.
like blockage to the oncoming flow. The flow attempts to climb the There is no information in the solar literature about wind-related
step but, because of its inability to turn sharp corners, separated re- heat transfer coefficients which takes account of the elevation of the
gions, containing recirculating fluid, occur. One such separated region collector above a host surface, nor are there related results in the basic
is situated on the forward portion of the heat transfer surface, while heat transfer literature. Until the recent past, solar texts and com-
a second zone of separation occurs at the intersection of the step and pendiums (e.g., [1-3]) recommended the equation hw = a + b V for
the host surface. In general, literature information on separated, re- the wind-related heat transfer coefficient, where V is the wind velocity
circulating flows indicates that the corresponding heat transfer and a, b, and n are empirical constants. This equation, based on ex-
coefficients are decisively different from those for flows which move periments for airflow parallel to a heated plate, is actually inappro-
in the mainstream direction. priate for solar applications. It does not take account of either the
While the study of the aforementioned physical situation can be orientation of the wind vector with respect to the collector plate or
regarded as basic research on a complex external flow, it also bears of the streamwise length of the collector plate. Recent experiments
a clear relationship to the design of flat plate solar collectors, as will [4-6] have dealt with these issues and have yielded a dimensionless
now be elaborated. Flat plate collectors are highly susceptible to heat HEAT TRANSFER SURFACE
losses, as witnessed by the fact that their efficiencies are on the order
of 50 percent or lower. One of the significant heat loss paths is via
forced convection from the upper cover plate of the collector to the
wind which passes over the collector. Consequently, heat transfer
coefficients for airflow over surfaces which are geometrically similar
to collector surface configurations are needed for a proper appraisal
of wind-related heat losses.
The situation is particularly complex in the case of a collector that
is installed on the roof of an existing building (i.e., retrofit installa-
tion). In that case, there is little likelihood that the upper surface of
the collector will be flush with the surface of the roof. Rather, the

Contributed by the Heat Transfer Division for publication in the JOURNAL

OF HEAT TRANSFER. Manuscript received by The Heat Transfer Division Fig. 1 Diagrams showing the physical situation, nomenclature, and per-
September 12,1980. ceptions of the separation zones

Journal of Heat Transfer AUGUST 1981, VOL 103 / 441

Copyright © 1981 by ASME
Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 correlation for hw. However, neither the aforementioned inappro- NAPHTHALENE
SURFACE
priate equation nor the new dimensionless correlation take account
of the recirculating flow which washes the upper surface of an elevated
collector.
Very recently [7], it was demonstrated that the use of an inappro-
priate wind-related heat transfer coefficient could have a marked
effect on the predicted performance of a flat plate solar collector.
The present research encompassed two complementary portions,
both of which dealt with the configuration pictured in Fig. 1. One of
these was the determination of heat transfer coefficients for the ele-
vated surface. The other was a flow visualization study which was
^BEVELED LEADING
focused on revealing the pattern of fluid flow adjacent to the elevated EDGE
surface.
Fig. 2 Pictorial view of the mass transfer model
The heat transfer coefficients were determined indirectly from mass
transfer measurements and by subsequent application of the well-
established analogy between heat and mass transfer [8]. For the mass degree beveled leading edge of the model (i.e., the edge which thrusts
transfer measurements, the naphthalene sublimation technique was forward into the airflow); the purpose of the beveling was to avoid
used. Relative to direct heat transfer measurements, the advantages bluntness effects.
of the naphthalene technique lie in the absence of extraneous losses The dimensions of the model were chosen so as to avoid excessive
via structure and supports, easier attainment of boundary condition blockage of the 61 X 30.5 cm ( 2 X 1 ft) cross section of the wind tunnel.
uniformity, and generally higher accuracy. The heat transfer Experience has indicated that when the blockage is small (~ three
boundary condition analogous to that of the mass transfer experi- percent or less), the results of wind tunnel experiments can be applied
ments is uniform wall temperature. The flow visualization studies without correction for flows in large unbounded domains. To fulfill
were performed by employing the oil-lampblack technique. this constraint, the base plate, the projected area of which is the
The experiments were carried out in a low-turbulence, subsonic blockage area, was initially fabricated as a 8.89 X 8.89 cm (3.5 X 3.5
wind tunnel, with a plate of square planform being used as the active in.) square. The elevated surface (i.e., the naphthalene surface plus
mass transfer surface. Four parameters were varied during the course the thickness of the framing walls of the cavity) was also a square, with
of the experiments. Three of these are geometrical parameters and dimensions 6.35 X 6.35 cm (2.5 X 2.5 in.) thereby leaving a 1.27 cm
are illustrated in the lower right of Fig. 1. They include: (1) the angle (0.500 in.) skirt all around. As initially fabricated, the step height was
of attack a of the oncoming airflow relative to the plate surface, (2) 1.27 cm (0.500 in.). The thickness of the base plate, selected to be the
the height H of the step, and (3) the width W of the portion of the host minimum consistent with strength and structural considerations, was
surface which frames the active mass transfer surface. Both H and 0.159 cm (ro in.).
W will be reported in terms of dimensionless ratios, H/L and W/L As was noted in the Introduction, the step height and the skirt
respectively, where L is the streamwise length of the elevated mass width were varied during the course of the investigation. With regard
transfer surface (L was held fixed for the experiments). The fourth to the step height, it was reduced to 0.846 cm (0.333 in.) after the
parameter is the Reynolds number Re. The experiments were con- completion of the data runs for the initial height. Subsequently, when
ducted so that for fixed values otH/L, W/L, and Re, the angle of at- the runs for the second step had been completed, the height was re-
tack a was varied systematically and served as the independent duced once again, this time to 0.423 cm (0.167 in.). These height re-
variable against which the heat (mass) transfer results are plotted. ductions were accomplished by machining the edges of the cavity wall
The format chosen for the presentation of the heat transfer results with a milling machine.
enables ready identification of the range of operating conditions where With regard to the skirt width, the visualization studies revealed
flow separation plays a significant role; the deviations of the separa- that only the portion of the skirtwhich thrusts forward into the on-
tion-affected coefficients from those for separation-free flow are also coming stream has a significant effect on the flow pattern on the mass
clearly portrayed. The visualization studies revealed the boundary transfer surface. Therefore, changes in width were made only for that
of the separated region on the active surface as well as the streamline part of the skirt. On completion of the planned data runs with the
patterns of the flow adjacent to the surface. original skirt, it was removed (by milling) and replaced by a 2.54 cm
(1.000 in.) skirt which was held in place by rivets which were anchored
The Experiments within the naphthalene cavity. A final set of runs was made with the
Experimental Apparatus. The mass transfer model used in the forward skirt removed altogether.
experiments is shown in pictorial view in Fig. 2. In essence, the model The geometric characteristics of all the variants of the mass transfer
consists of a naphthalene-filled cavity situated atop a flat plate. The model are listed in Table 1 in terms of the quantities H, W, and L that
metallic portion of the model (i.e., all parts of the model aside from are defined in the lower diagram in Fig. 1. The range of the angle of
the naphthalene) is a one-piece unit which was machined from a single attack a is also indicated in the table.
block of aluminum. As seen in the figure, the walls of the cavity frame The model was mounted in the test section of the wind tunnel by
the exposed surface of the naphthalene. The 0.117-cm (0.046-in.) means of a sting which was suspended from the upper wall of the
thickness of the frame was the minimum that could be machined tunnel. A swivel-type coupling situated at the forward end of the sting
consistent with the attainment of straight walls, uniform thickness, enabled the angle of incidence a to be set and maintained at any value
and a smooth exposed edge. Also shown in the diagram is the 45- between 0 and 90 deg, with the angle being read with a protractor. The

-Nomenclature.
cp = specific heat K* = value of K for nonseparated flow W = width of forward-thrusting skirt
'£) = diffusion coefficient L = streamwise length of active surface a = angle of attack, Fig. 1
H = step height, Fig. 1 til = rate of mass transfer per unit area
Pr = Prandtl number v - kinematic viscosity
h = average heat transfer coefficient
h* = value of h for nonseparated flow Re = Reynolds number, U„L/v p = density
j = Colburn;-factor, equations (2) and (3) Sc = Schmidt number Pnw = naphthalene vapor density at active
j * = ./-factor value for nonseparated flow t/co = free stream velocity of oncoming flow surface
K = average mass transfer coefficient, Xr = distance of reattachment line from
Pn<= = naphthalene vapor density in free
equation (1) leading edge of active surface
stream

442 / VOL. 103, AUGUST 1981 Transactions of the ASME

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 Table 1 Geometric characteristics of the mass As noted earlier, a new naphthalene test surface was cast for each
transfer models data run. To prepare for a run, airflow was initiated in the wind tunnel,
and the controls were set to yield the desired Reynolds number. The
H/L W/L a (deg) tunnel was operated in the open-circuit mode, with air being drawn
1/5 1/5 15-60 from the building and exhausted to the outdoors. This mode of op-
1/7.5 1/5 15-60 eration ensured that the air approaching the model would be free of
1/15 1/5 15-60 naphthalene vapor and, since the fan is positioned downstream of the
1/15 1/2.5 15-60 test section, there was no preheating of the air.
1/15 0 15-60
The tunnel was allowed to operate for about an hour to attain
coupling mated with a socket-like fixture that had been machined into steady state. At that point of time, the model was weighed and the
the underside of the base surface of the model. The sting supported cover applied and clamped in place. The thus-capped model was then
the model from the rear—the rear support being chosen to avoid affixed on the sting in the wind tunnel and the angle of attack a was
disturbing the flow over the windward surfaces of the model. set. Previous experience had demonstrated that the capped model
Mass Transfer Surface. The naphthalene test surface was achieved a steady temperature equal to that of the flowing air within
produced by a casting process utilizing a two-part metallic mold. One a half hour. During the equilibration period, the air velocity and
of the parts is the aluminum model itself, while the other is a thick, temperature were carefully monitored for constancy. Then, after the
highly polished stainless steel plate. To initiate the casting process, passage of a half hour (or longer) the impact tube was retracted, and
the solid naphthalene contained in the model cavity from a prior data the data run proper was initiated by uncapping the model and starting
run is removed by melting and subsequent evaporation. Then, the the stop watch.
model is placed face down on the stainless steel plate so that the rear Temperatures were read and recorded at frequent intervals during
surface of the base plate is up while the open end of the cavity is down the data run. The duration of the run was selected so that the amount
(and blocked by the steel plate). A funnel is implanted in an aperture of mass transfer could be measured accurately while limiting the av-
in the upfacing surface to facilitate the pouring of molten naphthalene erage recession of the surface to 0.0025 cm (0.001 in.). Depending on
into the cavity. Risers in the form of small diameter tubes are also the operating conditions, 100 to 200 mg were sublimed, and run times
implanted in holes in the upfacing surface to facilitate the escape of ranged from 20 to 75 min.
air displaced from the cavity by the molten naphthalene. To conclude the run, the wind tunnel hatch was opened, and the
Once the naphthalene had solidified, the two parts of the mold were model cap was applied and clamped. The mass of the model was im-
separated, exposing the naphthalene surface that had been cast mediately measured. Before the tunnel was turned off, a calibration
against the highly polished stainless steel plate. The quality of the cast procedure was executed to correct for any ambiguities in mass transfer
naphthalene surface was uniformly excellent—equal to that of the during the model's installation in and removal from the wind tunnel.
plate against which it was cast. The airspeed was also measured prior to shutdown.
A new mass transfer surface was cast for each data run using fresh Flow Visualization. As noted earlier, the oil-lampblack technique
(previously unused) reagent-grade naphthalene. Once the casting had was employed to enable visualization of the flow field adjacent to the
been made, the pouring aperture and the air-escape holes were care- elevated surface. Lampblack is a very fine black powder now primarily
fully sealed with a pressure-sensitive tape, thereby ensuring that the available as an ingredient of paint. It mixes readily with oil, and the
test surface would be the only naphthalene surface exposed to the mixture, when brushed on a surface, produces a smooth, glossy-black
airflow in the wind tunnel. coating. The fluidity of the mixture can be regulated by the selection
A cover, made of plexiglass, was designed to fit over the entire ele- of the viscosity of the oil and by the proportions of the oil and the
vated portion of the model (i.e., both the test surface and the step). lampblack powder. These proportions also have a significant effect
When locked in place by two spring-loaded clamps which pressed the on the degree of definition of the flow pattern that can be achieved
flange of the cover against the base plate, the cover served to provide with the method. The degree of fluidity of the mixture is an especially
an air-tight seal around the test surface. This seal served to prevent relevant factor when the flow adjacent to an inclined surface is to be
sublimation during the preparatory stage of the data run, as will be visualized, as is the case in the present investigation. In such a situa-
described shortly. tion, the mixture will sag under the action of gravity, unless it is suf-
Instrumentation and Experimental Procedure. For the de- ficiently stiff.
termination of the mass (heat) transfer coefficients and their corre- The general procedure for using the technique is to brush the oil-
lation with the airflow, six quantities were measured for each data run. lampblack mixture on a surface and then to expose the surface to the
The amount of mass transferred during the course of the run was airflow whose characteristics are to be studied. Ideally, under the
obtained from weighings of the model immediately before and im- action of the shear stresses exerted by the flow, the mixture will move
mediately after the run. For the weighing, a vernier-equipped Mettler along the surface, following the paths of the fluid particles that pass
analytical balance was employed; the balance had a smallest scale adjacent to the surface. These path lines will appear as streaks on the
division of 0.1 mg and a total weighing capacity of 200 g. The tem- surface. In regions of low shear stress (e.g., at separation, reattach-
perature of the airflow adjacent to the model was measured with an ment, or stagnation), the mixture will remain stationary, so that such
ASTM-certified 0.1°F thermometer positioned just downstream of regions show themselves as black streak-free zones on the surface.
the model. The accuracy of the thermometer was verified by com- For all of the visualization runs, the elevated surface was coated
parisons with a secondary standard. with white, plasticized contact paper in order to obtain the highest
To determine the free stream velocity of the air approaching the possible contrast for the black streak lines and the black low-shear
model, a retractable impact tube was used in conjunction with a wall zones. In order to attain the clearest possible patterns, the bulk of the
static tap. The impact and static pressures were conveyed via plastic visualization work was performed at the highest Reynolds number
tubing to the ports of a Baratron solid-state capacitance-type pressure investigated (~70,000), which also corresponded to the highest shear
meter capable of discriminating pressure differences to 10~4 mm Hg. stress. Some observations made at lower Reynolds numbers will, also
The output of the pressure meter was read with a digital voltmeter. be reported.
The impact tube-wall tap combination was checked against a Pitot
tube with integral pressure taps, with a resulting agreement within Results and Discussion
one percent. For the static pressure in the test section of the wind The presentation of results will begin with the information obtained
tunnel, the Baratron was used to measure the tunnel static relative from the flow visualization studies, with the heat (mass) transfer re-
to ambient, and the ambient pressure was read from a precision ba- sults to follow.
rometer. The duration of the data run was read with a stop watch to Flow Visualization Results. To provide guidance for the inter-
within one second, while the angle of attack was set with the aid of a pretation of the visualization results, attention will be directed to the
protractor. diagram at the lower left of Fig. 1. This diagram shows a perception

Journal of Heat Transfer AUGUST 1981, VOL. 103 / 443

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 of the expected separation zones associated with the presence of the of attack increases. This decrease in the size of the separated region
step. These separation zones are created because the flow is unable with angle is altogether reasonable and reflects the decrease of the
to turn the sharp corners at both the bottom and top of the step. projected step height as seen by the oncoming flow. The reattachment
Within each separation zone, there is a recirculating flow. line is more or less parallel to the leading edge, indicating a nearly
Of particular interest is the point C where the mainflow reattaches two-dimensional flow pattern in the neighborhood of reattachment.
to the elevated surface. Downstream of C, the fluid adjacent to the Indeed, in the recirculation region, the streamlines are straight and
surface moves in the streamwise direction, whereas upstream of C parallel. Even downstream of reattachment, the flow is nearly two-
(within the separation bubble), the fluid adjacent to the surface moves dimensional except near the side edges of the surface. As the size of
upstream. Thus, C is a point of zero shear, and it should appear as a the separation zone shrinks with increasing angle, three-dimension-
dark spot in the oil-lampblack visualization (actually as a line when ality becomes more evident.
the plate is viewed head-on rather than from the side). Similarly, As noted earlier, the bulk of the flow visualization runs were made
points A and B, which respectively correspond to separation and at the highest Reynolds number (~70,000) because the relatively high
reattachment, should also appear as black lines. shear stresses provided the most sharply etched flow pattern traces.
These expectations were actually realized during the visualization For this Reynolds number and for all step heights and skirt widths
runs. All three lines were observed, but primary attention was focused investigated, the reattachment distance (i.e., distance between the
on the reattachment at C since the heat (mass) transfer is directly leading edge of the elevated surface and the reattachment line) was
affected by the extent of the separated zone which washes the forward measured. The reattachment distance, to be termed X" is a direct
part of the elevated surface. index of the size of the separated region. It was measured directly from
Illustrative flow-visualization photographs showing head-on views the visualization patterns on the model surface (not from photo-
of the elevated surface are presented in Fig. 3. These photographs graphs).
correspond to the highest Reynolds number of the experiments Dimensionless reattachment distances XrlL are plotted in Figs.
(70,000), to the intermediate step height (H/L= 1/7.5), and to the 4 and 5 as a function of the angle of attack ct. Figure 4 conveys results
intermediate skirt width (WIL = 1/5). In the upper row, the successive for parametric values of the step height HIL at a fixed skirt width WIL
photographs (left to right) are for angles of attack ct of 15, 20 and 30 = 1/5, whereas Fig. 5 is for a fixed step height HIL = 1/15 with the
deg, while the lower photographs are for ct values of 40,45, and 50 deg. skirt width serving as the parameter. Turning first to Fig. 4, it is seen
These photographs show the traces of the streamlines adjacent to the that the trend of reattachment distance decreasing with angle, as
surface as well as the reattachment line. In some of the photographs, evidenced earlier by the photographs of Fig. 3, is reaffirmed. At the
extraneous black regions caused by sagging of the oil-lampblack smaller angles of attack, the reattachment distance is quite sensitive
mixture are visible. to the step height, increasing as the step height increases. On the other
By inspection of Fig. 3, it is seen that the reattachment line moves hand, for angles of 30 deg and larger, the reattachment distance is
forward toward the leading edge and finally disappears as the angle more or less independent of the height of the step. At angles of 50 deg

Flg.3 illustrative flow-visualization photographs; Re- 70,000, HlL = 117.5,

and WIL = 1/5. Upper row (left-to-right), (l( = 15,20, and 30 0 ; lower row
(left-to-right), (l( = 40, 45, and 50 0

444 I VOL. 103, AUGUST 1981 Transactions of the ASME

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 H/L ••(K/U„)Sc2' (2)
O 1/15
D 1/7.5 in which Ua is the free stream velocity and Sc is the Schmidt number
O 1/5 (analogous to the Prandtl number). There is also a /-factor for heat
transfer given by
/ = (hi pcpU JVTW (3)

According to the Colburn Analogy, the / values for heat and mass
transfer are equal at any given value of Reynolds number.
The exponent 2/3 of Pr and Sc in equations (2) and (3) bears some
20 30 40 50 discussion in view of the fact that for certain operating conditions,
a(DEGREES)
a portion of the plate surface is covered by a zone of flow separation.
Within the knowledge of the authors, the effect of Pr (or Sc) on heat
Fig. 4 Reattachment distances for various step heights at a skirt width W/L (mass) transfer coefficients in separated flows has not yet been de-
= 1/5 and Re ~ 70,000
finitively correlated. There is evidence, however, that the 2/3 power
W/L
may not be too far off the mark. In [9], the Nusselt number for a cyl-
0.3 O 0 inder in crossflow (a situation where about half the surface is washed
D I/5 by a separated flow) was correlated with Pr 0 - 37 , which corresponds
0 I /2.5 to (h/pcp [/„ )Pr 0 - 63 = constant at a given value of Reynolds number.
0.2
n ^-<x. In the range of Pr or Sc between 0.7 and 2.5, which encompasses the
Prandtl number for heat transfer in air and the Schmidt number for
O.I -

^xTx
^ ^ ~ naphthalene diffusion in air, there is little difference between a Pr 2 / 3
dependence or a Pr 0 - 63 dependence. However, to minimize the issue
0 —o—o—o—o—-b—o—o o
of which exponent is appropriate, most of the results are presented
I > i i i
in a ratio form which eliminates the presence of the thermophysical
20 30 40 60 properties.
a (DEGREES)
The /-factor for mass transfer was evaluated from equations (1) and
Fig. 5 Reattachment distances for various skirt widths at a step height H/L (2) by using the experimental data. To determine tu, the change in
= 1/15 and Re ~ 70,000 mass during a data run was divided by the duration of the run and the
or larger, separation could not be detected. mass transfer surface area. The determination of pnw involves a
As a supplement to the data presented in Fig. 4, additional visual- two-step process. First, the naphthalene vapor pressure pnw at the
ization runs were performed over a range of Reynolds numbers at wall was evaluated from the Sogin relation [10] with the measured
various fixed angles of attack. In general, there was a tendency for the temperature as input. Then, with pnw and with the temperature, pnw
reattachment distance to increase with the Reynolds number, rather was calculated from the perfect gas law. With regard to pn„, the
rapidly at first and then very slowly. The Xr/L values of Fig. 4 are complete absence of naphthalene vapor in the free stream gives pn „
more or less applicable for Reynolds numbers in the range 45,000 to = 0.
70,000. The Schmidt number appearing in equation (2) was evaluated from
The results of Fig. 5 show that the reattachment distance is quite its definition
sensitive to the width of the skirt that thrusts forward from the base
of the step. In general, the greater the skirt width, the larger the Sc = v/'D (4)
reattachment distance. In addition, separation persists at larger angles Since the naphthalene vapor concentration is minute, the kinematic
of attack for wider skirts. For the no-skirt case, the reattachment viscosity can be taken as that of pure air. The diffusion coefficient »
distance is generally small, and separation does not occur at angles was calculated from Skelland's formula [11]. For the operating con-
of 25 deg or greater. ditions of the experiments, Sc ~2.55.
Brief observations will be made about points A and B (lower left, The Reynolds number, which served to parameterize the results,
Fig. 1), which are the anchor points of the separation zone at the foot was defined as
of the step. Point A moves slowly forward toward the step as the angle
of attack increases. At angles of 40 deg and larger, a stagnation line Re = UMv (5)
first appears on the skirt and, as the angle increases, the stagnation
in which L is the streamwise length of the test surface.
line and the separation line markedly interact. Point B, which is sit-
Heat (Mass) Transfer Results. To help identify a format which
uated at mid-height at low angles, moves up the step as the angle in-
would be most effective in displaying the effects of flow separation,
creases and finally disappears. This disappearance coincides with the
working graphs were prepared in which/ was plotted as function of
condition that no separation occurs on the elevated surface.
the angle of attack a for a fixed Reynolds number and for fixed ge-
With regard to the relation of the visualization results to heat (mass) ometry. It was observed from these graphs t h a t ; varied with a in the
transfer, it may be expected that the larger the separation zone, the range of small and intermediate a, but t h a t / was essentially a con-
greater the transfer. This expectation is based on the belief that at stant at larger a. This latter behavior confirms a prior finding [4, 6]
the Reynolds numbers investigated, the separated region will contain that / is a very weak function of a in separation-free flow. It was fur-
a vigorously recirculating fluid. Consequently, high rates of transfer ther observed that in the range where / was nearly independent of a,
are expected at low angles of attack, high steps, wide skirts, and high it was also relatively insensitive to the specifics of the model geometry
Reynolds numbers. (i.e., step height, skirt width).
Dimensionless Heat (Mass) Transfer Parameters. The in- These characteristics suggested a presentation format that is ex-
terchangeability of heat and mass transfer results, which has been emplified by any one of Figs. 6 to 9. Consider, for example, Fig. 8. This
noted at various points throughout the paper, will now be formalized. figure corresponds to a fixed Reynolds number, Re = 50,000, and to
First, a mass transfer coefficient K is defined as a fixed skirt width, W/L = 1/5. In the figure, ///* is plotted as a
K = rhKpnu, - pn-) function of the angle of attack a for parametric values of the step
(1)
height H/L = 1/5,1/7.5, and 1/15. The quantity/* is equal to the av-
where m is the rate of mass transfer per unit surface area, and pnw and erage of the / values that fall in the a-independent portion of the data
p„» are the densities of naphthalene vapor at the mass transfer surface range. Aside from the secondary effect of geometry in the a-inde-
and in the free stream. A convenient dimensionless version of the mass pendent regime,/* corresponds to t h e / value for separation-free flow
transfer coefficient is given by the Colburn/-factor at the Reynolds number in question.

Journal of Heat Transfer AUGUST 1981, VOL 103 / 445

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 Re =70,000

H/L
» 1.6-
O 1/15
• 1/7.5
o 1/5

30 40
a (DEGREES)

Fig. 6 Heat (mass) transfer coefficients for various step heights at a skirt 30 40
width W/L = 1/5; Re = 20,000 and / * = 0.00603 a(DEGREES)

Fig. 9 Heat (mass) transfer coefficients for various step heights at a skirt
Re = 30,000 width W/L = 1/5; Re = 70,000 and j' = 0.00329
H/L
O I/I5 Re = 3 0 , 0 0 0
D 1/7.5 W/L
\ O I/5
0 0
n I/5
0 I/2.5

I I I 1 1
20 30 40 50
a(DEGREES) I.O
Fig. 7 Heat (mass) transfer coefficients for various step heights at a skirt
width W/L = 1/5; Re = 30,000 and / * = 0.00492
0.8
20 30 40 50 60
l.U
a (DEGREES)
Re = 50,000
Fig. 10 Heat (mass) transfer coefficients for various skirl widths at a step
1.8 height H/L = 1/15; Re = 30,000 and / * = 0.00495
H/L
O 1/15
D 1/7.5 2.0r
1.6
\ o 1/5

1.8
1.4

I.6-
1.2 -

I.O
I I 1 1 I
20 30 40 50 60 I.2
a (DEGREES)
Fig. 8 Heat (mass) transfer coefficients for various step heights at a skirt
width W/L = 1/5; Re = 50,000 and / * = 0.00379 I.O

In light of the foregoing, it is evident that deviations of j/j* from

unity are indicative of the effect of flow separation on the heat transfer 0.8
20 30 40 50 60
coefficient. Furthermore, since j and j * correspond to the same
a (DEGREES)
Reynolds number and to the same thermophysical properties, it fol-
lows that j/j* = K/K* = h/h*. The ordinates of the figures have been Fig. 11 Heat (mass) transfer coefficients for various skirt widths at a step
labeled to reflect these equalities. height H/L - 1/15; Re = 66,000 and / * = 0.00337

Figures 6 to 9 convey results for j/j* versus a for a fixed skirt width and levels off to the separation-free value, with the leveling-off taking
W/L = 1/5, with each successive figure corresponding to a fixed place in the a range 35-45 deg, depending on the operating condi-
Reynolds number equal to 20,000,30,000, 50,000, and 70,000. The data tions.
in each figure are parameterized by the step height H/L, and curves With regard to step height, higher steps give rise to progressively
have been faired through the data to provide continuity. greater enhancement of the heat transfer coefficient at those a values
The striking feature of these figures is the large values oij/j* that where there is significant enhancement. However, at lower en-
are encountered as the angle of attack decreases. These increases in hancements, there is some crossing of the curves for the various step
j/j* can be directly attributed to the enlargement of the separated heights, but in that range the enhancements caused by the different
region that was displayed in Figs. 3 to 5. For the operating conditions step heights are not very different. These trends with step height tend
of the experiments, separation-induced increases in heat transfer to reflect the variation of the reattachment distance with step height
coefficient of almost a factor of two are in evidence. as portrayed in Figs. 4 and 5.
As shown in the figures, the highest heat transfer coefficients are From a careful examination of Figs. 6 to 9, a number of significant
attained at the smallest of the investigated angles of attack, a = 15 Reynolds-number-related effects can be identified. First of all, the
deg. As a increases, the transfer coefficient decreases monotonically general level of enhancement increases as the Reynolds number in-

446 / VOL. 103, AUGUST 1981 Transactions of the ASME

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 creases. This is due to the increased vigor of the recirculating flow in height and skirt width), the heat transfer coefficient increases
the separation zone. Second, there is a diminished sensitivity to monotonically with decreasing angle of attack. Separation-related
variations of step height at higher Reynolds numbers. Finally, en- enhancements of the heat transfer coefficient as large as a factor of
hancement persists to higher angles of attack at larger Reynolds two were encountered. At larger angles of attack, where there is no
numbers. Thus, for example, at Re = 70,000, enhancement persists flow separation, the heat transfer coefficient is essentially indepen-
to a = 45 deg, while at Re = 20,000, enhancement has effectively dent of angle and of geometry.
ceased at 35 deg, for the higher steps and at 25-30 deg for the lowest Higher steps give rise to progressively greater heat transfer en-
step. hancement at angles where there is significant enhancement. The
Attention may now be turned to the effect of skirt width on the heat degree of enhancement is markedly increased as the skirt width in-
transfer coefficient, with the step height being held fixed at H/L - creases and, in addition, the enhancement persists to larger angles
1/15. These results are presented in Figs. 10 and 11, which correspond of attack. In the absence of a skirt, there is very little enhance-
respectively to Re = 30,000 and 66,000. In each figure, j'/j* is plotted ment.
as a function of a for parametric values of the forward-thrusting skirt The Reynolds number plays a significant role in the separation-
width W/L = 1/2.5, 1/5, and 0. The results shown in these figures induced enhancement. The degree of enhancement is greater and
convey a clear message, namely, that the separation-related en- enhancement persists to larger angles at higher Reynolds numbers.
hancement of the heat transfer coefficient is markedly affected by the On the other hand, with increasing Reynolds number the heat transfer
skirt width. coefficients become less sensitive to variations in step height.
Not only is the magnitude of j/j* significantly affected by skirt With regard to applications, it may be concluded that the use of
width, but also is the range of angles of attack at which enhancement literature information for non-separated-flow heat transfer coeffi-
occurs. Both the magnitude of the enhancement and the angles to cients will generally provide a low estimate of the surface heat transfer
which enhancement persists are increased as the skirt width becomes rate. For solar collectors, such a low estimate will yield an optimistic
greater. prediction of the collector efficiency.
The curves for the two non-zero skirt widths have similar shapes. As a final observation, it should be pointed out that a wind tunnel
It is interesting to note that there is a definite tendency for these is not able to model, in all details, the fluid flow conditions that are
curves to come together at smaller angles as well as at larger angles. encountered in the outdoor environment, e.g., unsteadiness and
The enhancement effects for the wider of the two skirts persist to turbulence characteristics. On the other hand, owing to the variability
angles that are 10-15 deg larger than for the narrower skirt. of the outdoor environment, it is unrealistic to think in terms of a
The characteristics evidenced by the no-skirt case are altogether universal set of fluid flow conditions for solar collectors (the specifi-
different from those for the finite-width skirts. First of all, there is cation of the mean wind speed does not characterize unsteadiness or
no significant enhancement of the heat transfer coefficient for angles turbulence). Thus, model experiments of any kind, whether per-
greater than 15-20 deg. Furthermore, an undershoot is in evidence formed in a wind tunnel or outdoors, can only approximate any par-
at lower angles, whereby the heat transfer coefficient falls slightly ticular solar collector application. While it is well to keep such con-
below the large-angle separation-free value. cerns in mind, it is still necessary to employ the results of model ex-
It is not surprising that there are such different trends when a skirt periments for solar collector design. In this light, the present results
is absent and when a skirt is present. These differences can be at- merit consideration for incorporation into solar collector design cal-
tributed to differences in the respective flow fields. Thus, for example, culations.
the separated region at the foot of the step (between points A and B,
Fig. 1, lower left) is no longer present when there is no skirt. Fur-
Acknowledgment
thermore, a stagnation line which would otherwise lie in the skirt may
shift to the step when the skirt is absent. This research was performed under the auspices of the American
Society of Heating, Refrigerating, and Air-Conditioning Engineers.
In common with the trends of earlier figures, the effect of higher
Reynolds numbers, as evidenced in Figs. 10 and 11, is to increase the
References
degree of enhancement and to extend the range of angles where en-
1. Duffie, J. A., and Beckman, W. A., Solar Energy Thermal Processes,
hancement occurs. Wiley, New York, 1974, p. 83.
2. Jordan, R. C. and Liu, B. Y. H., eds., Application of Solar Energy for
Concluding Remarks Heating and Cooling of Buildings, ASHRAE, New York, 1977, Chapter 8.
The research reported here was aimed at determining heat transfer 3. Krieth, F., and Krieder, J. F., Principles of Solar Engineering,
McGraw-Hill, New York, 1978, p. 139.
coefficients for a thermally active surface elevated above a host sur- 4. Sparrow, E. M, and Tien, K. K., "Forced Convection at an Inclined and
face. Supplementary flow visualization experiments, intended to Yawed Flat Plate—Application to Solar Collectors," ASME JOURNAL OF HEAT
provide insights into the patterns of fluid flow adjacent to the surface, TRANSFER, Vol. 99,1977, pp. 507-512.
were also carried out. 5. Tien, K. K., and Sparrow, E. M., "Local Heat Transfer and Flow Field
Characteristics for Airflow Oblique or Normal to a Flat Plat," International
The visualization studies demonstrated the presence of a region Journal of Heat and Mass Transfer, Vol. 22, 1979, 349-360.
of flow separation which blankets the forward part of the elevated 6. Sparrow, E. M., Ramsey, J. W., and Mass, E. A., "Effect of Finite Plate
surface. Fluid recirculation within the separated region was inferred Width on Heat Transfer and Fluid Flow about an Inclined Rectangular Plate,"
from the streak lines on the surface. The streamwise extent of the ASME JOURNAL OF HEAT TRANSFER, Vol. 101,1979, pp. 199-204.
separated region was found to increase with decreasing angle of attack. 7. Ramsey, J. W., and Charmchi, M., "Variances in Solar Collector Per-
formance Predictions due to Different Methods of Evaluating Wind Heat
On the other hand, at larger angles of attack, separation does not Transfer Coefficients," ASME JOURNAL OF HEAT TRANSFER, Vol. 102,1980,
occur. Separated regions of greater extent were encountered at larger pp. 766-768.
step heights and for greater widths of the skirt which thrusts forward 8. Eckert, E. R. G., "Analogies to Heat Transfer Processes," in Measure-
into the flow. There is also a tendency toward larger separated regions ments in Heat Transfer, E. R. G. Eckert and R. J. Goldstein, eds., Hemisphere,
Washington, 1976.
at higher Reynolds numbers. Since, at the Reynolds numbers inves- 9. Zukauskas, A., "Heat Transfer from Tubes in Crossflow," in Advances
tigated, the separation zone contains vigorously recirculating fluid, in Heat Transfer, Vol. 8, Academic Press, New York, 1972, pp. 93-160.
larger separated regions give rise to higher heat transfer coeffi- 10. Sogin, H. H., "Sublimation from Disks to Air Streams Flowing Normal
cients. to their Surfaces," Trans. ASME, Vol. 80,1958, pp. 61-69.
11. Skelland, A. H. P., Diffusional Mass Transfer, Wiley, New York, 1974,
At a fixed Reynolds number and for fixed geometry (i.e., fixed step p. 51.

Journal of Heat Transfer AUGUST 1981, VOL. 103 / 447

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 Unsteady Stagnation Point Heat
Takao Sano
Department of Mechanical Engineering,
Transfer with Blowing or Suction
University of Osaka Prefecture,
Sakal, Osaka, Japan
The effects of blowing and suction on unsteady heat transfer at a stagnation point due
to a step change in wall temperature are examined. Two asymptotic solutions for the tem-
perature field at large and small Prandtl numbers are presented. It is shown that the as-
ymptotic solution for large Prandtl number gives sufficiently accurate results for the sur-
face heat transfer even for the moderate values of Prandtl number ifEuler transformation
is applied to the series.

1 Introduction fm +ff" + 2- ; (l - / ' 2 ) = 0, (2)

The thermal response behavior of laminar steady flows at a front dB
stagnation point due to a step change in wall temperature or in wall = B" + Pr/0', (3)
dr
heat flux has been investigated by several authors [1-4]. The present
paper extends this problem to the case with a uniform blowing or where prime denotes differentiation with respect to i\, and the initial
suction at the surface. This extension has already been made by Jeng and boundary conditions as
[5]. The method of solution adopted by Jeng is to obtain two asymp-
0 = 0 for T = 0,
totic solutions respectively valid for small and large time, which are
joined together at some moderate value of time. However, the / = /o, / ' = 0, B = H{T) for v = 0 (4)
matching seems unsatisfactory especially for small Prandtl number
and the method for obtaining the large-time solution has a drawback f-* 1, 6—-0 as?;—- ">,
of the lack of uniqueness [4]. Jeng gave the numerical results only for where H(T) is the Heaviside step function. Positive values of f0 cor-
the case of blowing and only for Pr = 0.01 and 0.7. The latter restric-
respond to suction and negative ones to blowing. The momentum
tion on the value of Pr is probably due to the fact that the convergence
equation (2) has already been solved numerically by several authors
of the series for the parameters which are included in the large-time
[6-8]. In the following sections, we shall obtain asymptotic solutions
solution as functions of P r becomes poor when Pr is large.
of (3) for small and large Prandtl numbers.
In the present paper, in order to study the effects of blowing and
suction on unsteady heat transfer process for a wide range of Pr, we 3 Asymptotic Solution for Small Prandtl Number
obtain asymptotic solutions for large and small Prandtl numbers for We first consider the case of small Pr. Assuming that 6 may be ex-
the case of a step change in wall temperature. These solutions are panded as
uniformly valid for all time and it is found that the solution for large
Pr is valid even for moderate values of Pr if its convergence is im- d = 0O(T, i)) + Pr0!(T, ij) + Pr 2 0 2 (r, TJ) + . (5)
proved by the Euler transformation.
and substituting this into (3), we have
2 Governing Equations 30o_ ;
0, (6)
Consider steady, incompressible, laminar flows near a front stag- dr
nation point of a two-dimensional or axisymmetrical body. We assume
that there is a uniform blowing or suction through the surface. Ini- = W, (7)
tially, the surface of the body and the fluid are at the same tempera-
ture T„, whereupon at time t = 0 the surface temperature is suddenly etc. The solutions of these equations can easily be obtained by using
changed to a constant value Tw. By introducing the following vari- the Laplace transformation technique. The final result for the wall
ables temperature gradient is
/2''KU/2 2lKt
-(B')^0=(TTT)-UI + PT
v=y r=
hr iv 2U7 2
2
(1) + (JTT)-1'2 f °° (/ - V - f3)e-i2'rdV + 0(Pr ), (8)

B = (T - T„)I(TW - T „ ) , where

where x and y are the coordinates along and normal to the body, v the
(3 = lim (f - v). (9)
kinematic viscosity of the fluid, L the characteristic length of the body,
K a constant related to the velocity up just outside the boundary layer
It is seen that, as T —* °°, ($')i;=o becomes infinite, that is, (8) and hence
by Up = Kx, i a constant being zero for plane flow and one for axi-
(5) are singular for large T. The reason for the failure of the expansion
symmetrical flow, T the temperature of the fluid and i/' the stream
(5) at large 7 can be explained as follows. At an early stage (when r
function defined by
= 0(1)) in the unsteady heat transfer process, the thermal layer is
_ IL\' cM// _ ILV d ^ confined to the velocity boundary layer (inner region) near the surface,
\xj dy \xj dx where the effect of convection on temperature is small and the
equation (3) and the expansion (5) are adequate for describing the
u and v being velocities in the x and y directions, respectively, the temperature field. As r becomes larger, however, the thermal layer
governing equations for the velocity and the temperature fields can diffuses into the "outer region" outside the momentum layer where,
be written as [2] no matter how small Pr is, both the convection and conduction terms
are of the same order of magnitude. It is the existence of this outer
Contributed by The Heat Transfer Division for publication in the JOURNAL
OF HEAT TRANSFER. Manuscript received by The Heat Transfer Division May region which prevents the expansion (5) from becoming a uniformly
11,1979. valid approximation to the temperature.

448 / VOL. 103, AUGUST 1981 Transactions of the ASME

Copyright © 1981 by ASME
Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 In order to obtain a uniformly valid solution, we shall use the 8* = AoVp(£) + Pr 1/2 (-Ao/3V p _i(£) + AiV p (£)j
method of matched asymptotic expansions. From the argument stated
above, it is reasonable to construct two expansions valid in the + Pr - V P - 2 ( | ) - A 1 / ? V P - 1 ( J ) + pr3/2.
large-time region in addition to (5); one is an "inner expansion" valid b
in the inner region and the other an "outer expansion" valid in the AlF
outer region. These expansions are constructed in such a way that: + ~Y~ Vp_2(f) - A2(3Vp-i (£) + A3VP(£) + 0(Pr 2 ), (20)
(1) The inner expansion satisfies the boundary condition on the sur-
where
face. (2) The outer expansion satisfies the boundary condition at in-
finity. (3) The two expansions match identically in the overlapping U/2
VP«) = iFi 2P/2r|^+l
domain in space where both expansions are valid and also match the 2'2' 2
small-time solution (5) at small values of T*, T* being a time variable
in the large-time region.

as
As the time variable in the large-time region, we introduce T* such
-$i*i
.£ + Ii_q/2<p-w2r
2 2,2, 1)1
iFi(a, b,x) being the confluent hypergeometric function [9], and An's
N (21)

are unknown constants to be determined through the following

T* = P r r s 2'Kt, (10) matching condition:
in terms of which the energy equation can be written as
lim6»* : l i m 9 * or, lim 8* = lim9* (22)
d0*
Pr — = 8*" + Pr/0*', (11)
dr* In view of (20), the inner expansion should be of the form
where 8* = 0O* (T*, i)) + Pr^fli'HT*, ij) + Pr0 2 *(T*, ij)
0 * ( T * , I J ) = 0(T,TJ). (12) + Pr 3 / 2 0 3 *(T*, v) + 0(Pr 2 ). (23)
(11) is an inner equation valid in the inner region. In the outer region, Substituting this into (11) yields the following equations for 0„*:
we introduce following variables
:
0,
£ = Pr 1/2 ?/, F(0 = 7vmf(n), 9 * ( T * , £) = 0 * ( T * , IJ), (13)
;
-f8n-2*' + ^ - ^ (for n > 2). (24)
in terms of which the energy equation may be written as dr*
dO* Solutions of these equations satisfying the boundary condition on the
• = e*+F(£)G*, (14) surface are
dr
where the dot denotes differentiation with respect to £. Since, in this 1 + a0(j*)n,
region, r) is very large, the stream function F is expressed by the fol-
lowing asymptotic form: = <2I(T*)}/, (25)

F~£ + PPi1/2 + (exp-),

where (exp—) denotes exponentially small terms. Neglecting the
(15)
U *i

o
fi \ 1 dao

Vfdt] - T) I fdr) + — - — 7 j 3 + a 2 (r*)7;

Jo / 6dr*
03* = ai(r*) Vfdt] fd
terms, (14) becomes
io \
6 dr*
ae*
- = 9* + (£ + /3Pr1/2)9* (16) Matching condition (22) determines the unknown constants An and
dr a„. Only the final result for the wall temperature gradient calculated
Since the small-time solution decays exponentially as ?j - » °°, the from the inner solution is shown below.
following matching condition holds as an initial condition for 9*: 2 U/2
-0*'),-o •
Mm 6* = 0. (17)
l,r(l - e-2'*)J
T* — 0 2|8PrjE(yi - e~^*) - e- 2 T *K(Vl - e"2T*)]
Therefore, application of the Laplace transformation 7r(l-e-2^*)
1/2 i(
, (-1 .-9^*1-1/9 "^ -' > p r 3/2 + Q(P
e* = C" e*e-PT*dT*, (18) •0 7T

where
to (16) yields
6* + (£ + j S P r ^ G * - p 9 * = 0. (19)
The solution of (19) satisfying the boundary condition at infinity is V / l -"" 2 T *vT : r F{2B(fe) - (2 - fe2)K(fe)j
expressed as an expansion in terms of Pr 1 ' 2 as follows:
/(T*)=
X fe3Vl-e-2T*-fe2
dfe, (27)

-Nomenclature^
f,F = dimensionless stream function T = temperature 0,0*, 9*, 0 = dimensionless temperature
/o = value of/ at the surface of the body u = velocity in x direction v = kinematic viscosity
i = constant being zero for plane flow and one v = velocity in y direction £ = dimensionless coordinate defined in
for axisymmetrical flow x = coordinate along the body measured from (13)
„
K = proportionality constant related to the the front stagnation T,T*,T = dimensionless time
velocity outside the boundary layer y = coordinate normal to the body
\p = stream function
L = characteristic length /? = constant defined in (9)
p = parameter in Laplace transform y = constant defined in (27) Subscripts
Pr = Prandtl number f = dimensionless coordinate defined in s = steady-state
s = parameter in Laplace transform (30) u> = condition at wall surface
t = time T\ = dimensionless coordinate defined in (1) °° = free stream condition

Journal of Heat Transfer AUGUST 1981, VOL 103 / 449

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 K(k) and E(k) being the complete elliptic integrals of the first and Table 1 N u m e r i c a l v a l u e s o f /3 a n d y
the second kind, respectively. It is easy to show that the large-time
solution (26) matches with the small-time solution (8) and gives for h & 7
T* —• co the steady-state solution which has already been obtained by
-3 -4.8585 2.590
the present writer [10]. 0.7234
-1 -1.9450
When r = 0(1), the surface heat transfer may be calculated from -0.5 -1.2810 0.5081
(8), and when T* = 0 ( l ) ( r = CHPr- 1 )) from (26). It is, however, pos- i =0 0 -0.6479 0.3595
sible to construct from (8) and (26) a single composite expansion which 0.5 -0.0423 0.2584
1 0.5407 0.1895
is valid uniformly for all values of time. 1 Employing the so-called 2.7329 0.0678
3
"additive composition method" [11], we get
I 2 U/2 1 h*Yli j = l 0 -0.8046 0.5360
-(«Vo = P r l/2
2
nr(l - e- **)) ~2\ir
The solutions of (33) satisfying the initial and boundary conditions
( 7 r T ) -l/2 C" fe-vVrdv_l
Jo 2 can easily be found by using the Laplace transformation technique.
The final result for 0U is
Pr
ir(l- *•) I j e r f c ( £ ^ ± M ^ + .-^e^zM^J,
2W 2 12 2/3 2 /(T*)
7(1 -e-^*)- ' - (34)
,7r;

. 7 („. T *)-i/2 Pr 3 / 2 + 0(Pr 2 ). (28) which represents the temperature distribution in the limit Pr -* <».
It should be noted that, when /o < 0, do —"• 1 as f —• °°, that is, the so-
For calculating /3, y, and the integral included in (28), we need the lution (34) does not satisfy the boundary condition at infinity when
numerical values of the stream function / . For i = 0, Evans [7] ob- f —• °°. This means that for the case of blowing the expansion (32) is
tained numerical solutions off for several valus of/o- For i ~ 1, on the invalid in the outer region when both f and fare infinitely large, and
other hand, they are obtained only for /o = 0 [6] and hence the cal- that an outer expansion, which is not obtained in the present paper,
culation of the integral and constants mentioned above is performed prevails there. This failure of (32) at large f and large f is due to the
only for this special case. The calculated results for /3 and y are shown fact that, as f increases, the thermal boundary layer, which initially
in Table 1. adheres to the wall, becomes to be blown away from the surface due
to the convective effect of the injected fluid and the approximate
4 Asymptotic Solution for Large Prandtl N u m b e r equation (29) for / becomes to be invalid in the thermal boundary
We shall now proceed to obtain the asymptotic solution of (3) for layer. In the injection layer near the surface, (29) is still a valid ap-
large Prandtl number. When Pr is large, the thickness of thermal layer proximation for / even when f is large and (32) is valid for all values
is much thinner than that of viscous layer, so that the velocity field off.
in the thermal layer may be expressed by the following series form: The solutions of (33) are obtained up to §4, which are not shown here
to save space, and it is found that, for/o < 0,8 n - • 0(ra > 1) as f —• °°.
f t 4-(f°"\ 2 J0'" 3 >«"> 4 . (29) This means that, when f is large, deviation of the value of 0 from unity
I 2/ 6 24 is exponentially small in the injection layer and the temperature drop
where subscript 0 denotes the values of functions at 7)l : 0. We assume from unity to zero occurs only in the thermal boundary layer far from
t h a t / o ^ 0. Introducing the following variables the surface. The result for the wall temperature gradient is

f = Pr?/, f = Pr 2 r, 0(f, f) = 0(T, IJ), (30) -(«Vo = £ + erfc(-T) P r

and substituting (29) and (30) into (3), we have
K + —= e - ^ + e r f c O - T ) Pr-
^-=^ + (/o + ^rPr- + ^f3 P r -3 /o 2 "ifil
2 2
af af r 2 6 T T j, r2 2
id0 3 — + — = + - e - + e r f c ( - T ) Pr""
r 4Pr-4 + ... (3D /o 2 V^ 2
24 'af /O / 4r 6 T4 2T
y2 . 1 e-r2+erfc(-T)
Assuming that 6 may be expressed as U5V^~ 2
foiiv) - 4 T
3
0 = 0„( f, f) + h(r, f)P)•- 1 + 02(f f)Pr- 2 + ... , (32) / 1
72 _
2T_
4
• l | e - T 2 + e r f c ( - T ) ' Pr-
and substituting this into (31), we have "2/0 3V^r" V7T
+ 0(Pr-4), (35)
afl0 d%
0,
df df 2 where
2
dfli d §! d(?!
0, (33) T ••-V~r. (36)
df df 2 "
d0 2 d202 ,2?!° It is seen that, when/o < 0, (0'),,=o ~ 0 ( P r _ ° ) as f ->• <*>, that is, for the
^ f ~ d f 2 ~ / 0 df 2 of, case of blowing, the wall temperature gradient becomes to be expo-
nentially small as Pr -> °= and f ->• <=°.
a 2 0 3 , a0 3 / 2 d0!
+ r
df af2~/oaf = ~6 df 2 df 5 Discussion
Using the asymptotic solutions obtained in the preceding sections,
1 we calculate the wall temperature gradient for Pr ranging from 0.01
According to Van Dyke [11], the construction of a single uniformly valid
composite expansion can be carried out in a large variety of ways and their re- to 100 and for /n from —3 to 3. (For axisymmetrical case, the small
sults may be different because of the non-uniqueness of the composite expan- Prandtl number results are not obtained since the numerical values
sion; but they will all be equivalent to the order of accuracy retained. for /, which are necessary for calculating the temperature field, have

450 / VOL. 103, AUGUST 1981 Transactions of the ASME

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 0.14 100
i=0 =^_ Pr = 100
^ ^
-Present results
10
-Chen & Chao (3) ^=*"*=**=^::^
^^~~^~-^_
"Jeng (5), large-time sol. X=rJ^_t___ | —

a> 0.10
T

Pr = 0.01

0.07 10 ~-9j7
-^?
wr
fo 5

Kt I l
Fig. 1 Comparison of the present results with the previous works at Pr - 0.001 0.01 0.1
0.01
2' Kt
Fig. 2 Effect of fa on heat transfer response

100

i= 0
f.=_QJ
^^ 1

5" loi- 0
n-
's

^s^.
^^^,
^**^^^^-
OD l
0.001 Q01 0.1
2 ' Kt
Fig. 3(6)
Fig. 3 Effect of Prandtl number on heat transfer response

not yet been given.) It is found that for sufficiently large or small Pr due to mass transfer at the surface has small effect on the temperature
the present results for T - • <» agree well with the numerical ones given field. In Fig. 3, on the other hand, the effect of the Prandtl number
for steady state [8] and t h a t the convergence of the asymptotic solu- on the ratio qwlqw,s is shown for some values of /n- It is seen that for
tions is better for smaller values of time. For moderate value of Pr, the case of blowing and for impermeable surface the response time
we apply the Euler transformation to the asymptotic solution for large increases monotonously with increasing Prandtl number, while for
Pr. The Eulerized series shows good convergence even for P r = 0.7 the case of suction the trend is reversed provided that/n is larger than
when r* is smaller than about 1.5. some threshold value fo*. The exact value of fo* can not be determined
In Fig. 1, our small Prandtl number results for plane flow are in the present paper. For 0 < fo < fo*, the response time is not a mo-
compared with those 2 obtained previously at Pr = 0.01 and fo = —1,0. notonous function of Pr; it is maximum at some value of Pr (de-
It is seen that the results of Jeng [5] are not satisfactory. Figure 2 pending on the value of fo) and decreases as P r - * 0, » . This aspect
shows the effect of mass transfer on the variation of the ratio qwllw,s may be considered as the transitional one between the two diamet-
with time for Pr = 0.01,1 and 100, where qw is the instantaneous wall rically opposite trends described above. It is also seen from Fig. 3 that
heat flux and qw,s the steady-state value of qw. The results for /n = blowing has greater effect on qw/qw,s for axisymmetrical than for the
0 are obtained using the solution in [4]. It is seen from Fig. 2 that the two-dimensional case, as is already noted by Jeng, but that the effect
response time of heat transfer increases monotonously as fo decreases of suction is almost the same for both cases.
from positive values (suction) to negative ones (blowing) and the effect
of mass transfer is larger for larger value of Pr. The reason why the References
effect of mass transfer on the response time is small for small Pr is that 1 Cess, R. D., and Sparrow, E. M., "International Developments in Heat
when Pr is small the thickness of momentum layer is much smaller Transfer," Trans. ASME, Part II, Sec. B, 1961, p. 468.
than that of thermal layer and the change in the velocity distribution 2 Chao, B. T., and Jeng, D. R., "Unsteady Stagnation Point Heat Trans-
fer," ASME JOURNAL OF HEAT TRANSFER, Vol. 87,1965, p. 221.
3 Chen, J. L., and Chao, B. T., "Thermal Response Behavior of Laminar
2
The author wishes to thank one of the referees, who kindly made available Boundary Layer in Wedge Flow," International Journal of Heat Mass
to the author the numerical data of Chen and Chao [3]. Transfer, Vol. 13,1970, p. 1101.

Journal of Heat Transfer AUGUST 1981, VOL. 103 / 451

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 4 Jeng, D. R., Lee, M. H., and De Witt, K. J., "The Final Approach to 8 Elzy, E., and Sisson, R. M., "Tables of Similar Solutions to the Equations
Steady State in Nonsteady Stagnation Point Heat Transfer," Journal of En- of Momentum, Heat and Mass Transfer in Laminar Boundary Layer Flow,"
gineering Mathematics, Vol. 10,1976, p. 173. Bulletin No. 40, Engng. Experiment Station, Oregon State Univ., Carvallis,
5 Jeng, D. R., "Unsteady Heat Transfer in Laminar Boundary Layers," 1967.
Ph.D. thesis in mechanical engineering, University of Illinois, 1965. 9 Slater, L. J., Confluent Hypergeometric Functions, Cambridge Uni-
6 Hartree, D. R., "On an Equation Occurring in Falkner and Skan's versity Press, 1960.
Treatment of the Equations of the Boundary Layer," Proceedings of the 10 Sano, T., "Forced-Convection Heat Transfer in Laminar Boundary
Cambridge Philosophical Society, Vol. 33,1937, p. 57. Layer at Low Prandtl Number," Bulletin of Univ. Osaka Prefecture, Series A,
7 Evans, H. L. "Mass Transfer through Laminar Boundary Layer-8. Vol. 21,1972, p. 27.
Further Solutions to the Velocity Equation," International Journal of Heat 11 Van Dyke, M., Perturbation Methods in Fluid Mechanics, Parabolic,
Mass Transfer, Vol. 5,1962, p. 373. 1975.

SECOND NATIONAL SYMPOSIUM ON

NUMERICAL METHODS IN HEAT TRANSFER
Sponsored by the
National Science Foundation and the Office of Naval Research
to be held at
University of Maryland
College Park, Maryland
September 28-30, 1S31
Twelve sessions are arranged:
(I and II) Finite Difference Methods
(III and IV) Finite Element Methods
(V and VI) Comparison of Finite Difference, Finite Element and Other Schemes
(VII and VIII) Novel or Existing Methods
(IX and X) Applications to Heat Transfer Problems
(XI) Numerical Convective Instability
(XII) Five-Minute Presentations
Invited lectures will be delivered by
Professors W. F. Ames, D. K. Edwards, B. A. Finlayson, R. B. Kellogg,
T. M. Shih and D. B. Spalding.
Participants who wish to present five-minute talks should submit their complete papers to the principal organizer
prior to September 1,1981.
Advance registration is strongly recommended. For further information, please contact the principal organizer:
Professor Tien-Mo Shih
Department of Mechanical Engineering
University of Maryland
College Park, MD 20742
The Organizing Committee:
I. Babuska, U. of Maryland at College Park; J. deRis, Factory Mutual Research; D. K. Edwards, UCLA; B. A. Finlayson,
U. of Washington at Seattle; H. B. Keller, Cal. Tech.; W. J. Minkowyez, UICC; S. V. Patankar, U. of Minnesota; S.
F. Shen, Cornell U.; D. B: Spalding, Imperial College; E. M. Sparrow, U. of Minnesota; C. L. Tien, UC Berkeley; K.
T. Yang, U. of Notre Dame.

452 / VOL. 103, AUGUST 1981 Transactions of the ASME

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 K. A. Joudi
Lecturer, Department of Mechanical Engineering,
University of Basrah, Basrah, Iraq
Surface Contamination,.
D. D. James Rejy¥enationf and the
Lecturer, Department of Mechanical Engineering,
UMIST, Manchester, England
Reproducibility of Results in
Nucleate Pool Boiling
Contamination of stainless steel boiling surfaces at nucleation sites by corrosion products
in water, and by oxide deposits in methanol are investigated experimentally. Reproduc-
ibility of boiling data is achieved by nitrogen sparging of test liquids and employing inert
components in the apparatus. Heating surface rejuvenation is effected by boiling in re-
frigerant R-113. The order of testing a surface with various liquids is investigated.

Introduction
The important feature of any reliable scientific investigation is that during experiments, was bubbled through a 6 mm glass tube fitted
the results obtained should be reproducible at different times under flush in a corner of the base plate of the boiling chamber and posi-
similar conditions. A feature which is peculiar to nucleate boiling, tioned 80 mm from the nearest edge of the test specimen.
more so than the other regimes of heat transfer, is the difficulty en- The flow rate of nitrogen was kept to a minimum during tests such
countered in controlling the condition of the heat transfer surface, that it did not influence the natural convection currents and liquid
and also maintaining the chemical purity of both the surface and the circulation to and from the heating surface. Nitrogen bubbles did not
boiling liquid. A consequence of this difficulty is reflected in the in- induce bubble generation from the heating surface and did not appear
consistency and scatter of experimental data caused by the "time- to have an observable effect on the boiling process.
in-use" or so-called aging of a surface on which boiling occurs. A test surface was fitted with a 6 mm thick "pure natural" rubber
Details of how the effect of aging is overcome are not usually given backing piece which provided thermal insulation as well as sealing.
in the majority of the published literature, except for general remarks These items were then placed on an Araldite block which fitted over
to the effect that the data were taken after the surface had aged. two gold plated brass electrodes as shown in Fig. 2. The surface tem-
Generally, surfaces which have been in service for long periods of time perature was determined from the average indication of ten Chro-
require higher temperature differences to maintain the same heat flux. mel-Alumel thermocouples attached to the base of the test specimen,
Rohsenow and Choi [1] attribute this trend to either added resistance and situated so as to measure the temperature distribution across the
of an oxide or deposited layer, or to the decrease in cavity size resulting length and width of the sample. Measured temperature profiles in-
from mild oxidation. Localized deposits of corrosion products at nu- dicated that approximately 96 percent of the boiling area was iso-
cleation sites have been cited by Nestrueva, et al. [2] in experiments thermal.
involving boiling heat transfer with water. They reported that the
formation of oxide film, and deposition of corrosion products was the
most probable reason for the considerable scatter of the available data
on heat transfer in developed boiling.
This work describes an attempt to systematically inhibit and pos- NEEDLE
/ VALVE
sibly eliminate surface contamination under boiling conditions so that
results from boiling experiments over a period of time are reproduced AR aEED
VALVE
and can be relied on. The use of nitrogen sparging as an oxygen-in- ^WCUUM
duced corrosion preventer is an innovation which will be of interest I DVM - AC
~RESERV0R
TRAP
to those concerned with reproducibility of results in boiling heat CHAMBER
transfer. Heat transfer data are reported for stainless steel surfaces
boiling in three liquids; Analar 1 water, methanol and refrigerant R-113
over a pressure range including 0.25 atm., 0.5 atm., and atmospheric
DOUBLE 10 KVA
pressure. VARWC TRANSfORMER

Fig. 1 Schematic diagram of experimental apparatus

Description of the Equipment
The experimental apparatus and instrumentation are shown
6mm RUBBER COPPER STAINLESS STAINLESS-
schematically in Fig. 1. A test liquid was boiled from an electrically PIECE TERMINAL STEEL TEST

V-fo^
STEEL NUT
/
SURFACE
heated horizontal rectangular stainless steel plate housed in a cubical
stainless steel boiling chamber with clear inside dimensions of 210 mm
on all sides. An auxiliary heater made of resistance wire in silica glass
tube was used to bring the test liquid to boiling and maintain it at
THERMOCOUPLE -
WIRES
¥
saturation temperature during low heat flux operation. Hot vapours ^ ^
were conducted to a stainless steel condenser, and the condensate was a&xw^ ffi^sT
returned to the boiling chamber, through the base plate, by glass
tubing. Nitrogen gas which was used for sparging water and methanol
3 mm RUBBER
CASKET
TUFNOL
-STAINLESS
STEEL TUBE H
BUSHWG

1
Analar water is a double de-ionized commercial product obtained from
British Drug Houses.
Contributed by the Heat Transfer Division for publication in the JOURNAL
OF HEAT TRANSFER. Manuscript received by the Heat Transfer Division
December 14,1979. Fig. 2 Test surface assembly

Journal of Heat Transfer AUGUST 1981, VOL. 103 / 453

Copyright © 1981 by ASME
Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 Tests were carried out with flat specimens of varying surface
roughness, made of type EN 58E stainless steel measuring nominally
100 X 50 X 1.6 mm. The variation in surface roughness was produced
by a vaqua blasting technique, which consists of the impingement on
the surface at high pressure of an alumina particle suspension in
water.
Low pressure operation was achieved by evacuating the system
from the cold end of the condenser which reduced loss of vapor to a
negligible level. Pressure inside the chamber was regulated by a
throttling needle valve connected to a vacuum reservoir-trap chamber.
System and vacuum pump pressures were measured with mercury-
in-glass manometers.
Heating of the test surface was accomplished by the passage of a
heavy a-c current through the specimen at low voltage from a trans-
former which was regulated by a double variac.
The apparatus was dismantled completely at the conclusion of each
experiment and cleaned dry of all the residue of the test liquid. Prior
to each experiment, the test surface was cleaned with acetone before
the introduction of the test liquid. A more detailed description of the
Fig. 3 Evidence 01 heating surface contamination
apparatus, instrumentation and measurement is given by Joudi and
James [3].

Surface Contamination and Associated Experimental

Refinements
Contamination as referred to here is not the normal tarnishing of
a surface which causes what is known as aging in boiling research.
Rather, it is the appearance of small circular "doughnut" shaped
spots, of approximately 0.5 mm in diameter, at the base of boiling
bubbles. In fact, the total number of nucleation sites can be assessed
by counting the number of these spots at the end of a test. In this work
it was observed that, generally, these spots became larger and darker
in color with prolonged or more vigorous boiling. Two different forms
of surface contamination were encountered, the first with methanol,
and the other with water, each of which was used as a test liquid
boiling on stainless steel test surfaces. The two types of contamination
are described separately, together with the consequent refinements
of the experimental apparatus which were employed to combat the
contamination.
The photograph in Fig. 3 was obtained using a Scanning Electron
Microscope (SEM), and shows spots which were incurred by methanol
contamination. Small sections of a methanol contaminated surface
were examined, for analysis, at high resolutions in an SEM, and at
even higher resolutions under an Electron Probe Microanalyzer
(EPM). Figure 4 represents a scan with the EPM across the edge of
a crater (spot) from the outside to the center of the crater, showing
Fig. 4 Electron probe mlcroanalyser: scan 01 surface contamination
zinc ion concentration. The peak of the graph, above background
noise, was a result of concentration of zinc deposit on the surface, the
density of which diminished as the scan approached the clean centre
of the crater. With the EPM contamination spots were analyzed for
concentrations of iron, copper, nickel, chrome, magnesium, lead,
manganese and zinc. Of these elements, only zinc was detected, and
upon comparing its content to a standard, approximately 2.5 percent I ~
, '", ~
zinc was found in the annular space. ILl'
~
lL
I
Figure 5 is a qualitative analysis obtained with the SEM of the
constituent elements, immediately under the electron beam of a test I
section containing methanol contamination spots. The very high ~ <L H
peaks correspond to the presence of the parent stainless steel material,
":
LJ
'"
co 'z",
I
i.e., iron, nickel and chrome. The smaller peaks represent other ele- I
ments, and in particular zinc, which is present in all its phases. An a

aluminium peak is evident and is caused by deflections from the -: ",,,

, ' '"
0
Vl(3 N
specimen's aluminium mounting stub, whilst a small sulphur peak I
is attributed to the parent stainless steel material, which more often 1:1 I I
than not, contains small amounts of sulphur. Identification of the I
elements present is accomplished by determining the energies of the
peaks in the spectrum. A marker is positioned on to each peak shown Fig. 5 Scanning electron microscope: analysis of a contamination spot

_____Nomenc1ature __

q/A '" heat flux Tw '" heated surface temperature TRat ", iiquid saturation temperature

454 / VOL. 103, AUGUST 1981 Transactions of the ASME

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 in the spectrum and direct readout is obtained of the energy level of
the peak.
The above analyses indicate clearly the presence of zinc in the spots
of contamination which are produced by the boiling of methanol. The
source of zinc is the dezincification of brass, which comprises a copper
rich phase and a zinc rich phase. Zinc ions dissolve in the boiling
methanol forming a white coloured zinc oxide, which precipitates on
the boiling surface at the base of bubbles, once its concentration ex-
ceeds a certain value. The generation of boiling bubbles accentuates
precipitation, and in particular at the base of bubbles where the
surface is hottest. This type of contamination was totally eliminated
by replacing all the original brass fittings of the boiling chamber with
stainless steel fittings, and by gold plating the brass electrodes and
the copper terminals of each test piece. Another necessary measure
was the distillation of methanol prior to its use in a boiling test, thus
eliminating possible dissolved solids.
The second type of surface contamination was localized corrosion,
which was encountered with water boiling on the stainless steel
heating samples. Microscopically, a vaqua blasted test surface is very
rough and contains innumerable irregularities that could be consid- Fig. 6 Electron micrograph of surface finish
ered as pits, as can be observed from Fig. 6. Stainless steel is a ferrous
alloy, and in a smooth surface the distribution of alloys is uniform. one or more of the boiling test liquids, or were contributing to the
However, on a rough surface the distribution of alloys is nonuniform, contamination of test surfaces.
and a higher concentration of iron is obtained in pits, which conse- Consequently, all the brass and copper electrical terminals were
quently makes them more readily corrosive. From the microscopic gold plated, all fittings were manufactured from glass and stainless
pits in the surface, iron ions dissolve in the boiling water to form fer- steel, silicone rubber was used for gaskets and pure natural rubber
rous oxide with the oxygen dissolved in the water. This corrosion para B was used to seal the underside of the test specimen. It was
product appears on the surface at the mouth of the pit as ferric oxide found necessary to pre-treat the rubber prior to installation. Full
(rust) and an accumulation of this product forms a small brown information covering these experiments is recorded by Joudi [5).
spot.
Corrosion spots with water occurred at the base of bubbles, as did Results and Discussion
those of methanol contamination. The shape of these spots was ex- Conditions of surface contamination appeared in the form of small
actly the same as those of Fig. 3. Pits were not only a source of iron ions spots at nucleation sites when a stainless steel test surface was boiled
but also acted as bubble nucleation sites. In fact, the ebullition cycle in water or methanol. In water the spots were the result of localized
of a bubble acted as a pump for ion transportation from inside the pit corrosion of the surface, whereas in methanol these spots were the
to the cavity mouth, while the successive dryout and rewetting at the result of the deposition of zinc oxide at bubble sites. Reproducibility
base of a bubble stream provided an excellent condition for the supply of boiling data was established for several stainless steel test surfaces
and deposition of the corrosion products. Ross [4) reviewed the effects and experimental conditions. However, for the purpose of this pre~
of heat transfer on the corrosion phenomenon, and stated that special sentation typical results of only one specimen are presented.
features were introduced with boiling from a heated surface. Accel- As a consequence of surface contamination, higher temperature
erated reaction rates between liquid and surface were suggested as differences were continually required for a given value of heat flux.
a consequence of dryout and rewetting of the surface by a boiling This was true not only from test to test, but also at anyone power
liquid. In addition, electric heating of the test surface together with input setting. In a single test, the surface temperature continued to
the use of dissimilar metals under boiling conditions might possibly increase with time as the contamination became more severe with
set up a galvanic action, which would make matters worse. The extent sustained boiling.
of such a reaction could not be assessed, however, and no apparent Refrigerant R-1l3 which is sometimes used as a strong cleaning
effect of a galvanic action was observed. agent, was found to cause a reversal of the pattern of surface aging.
Corrosion spots could not be analyzed by a Scanning Electron A surface which was originally boiled in water and then in R-1l3, and
Microscope, because the presence of iron in a spot would be masked then boiled again in water exhibited a marked increase in the number
by the abundant iron of the parent stainless steel material. P.eaks, of bubbles. In the latter test (with water) this augmentation in bubble
similar to those of Fig. 5, for contaminant material would not be ob- population was accompanied by a substantial reduction in surface
tained. This would be particularly the case when the thickness of the temperature for the same values of heat flux as those originally ob-
corrosion product is less than 10 J.tm, which in fact was the situation tained in the initial test. This phenomenon will be termed "rejuve-
encountered in this work. nation" of a boiling surface and is illustrated in Fig. 7, together with
To reduce the rate of corrosion of the stainless steel, oxygen dis- the effect of surface contamination.
solved in the water must be removed. A general approach for the re- It would facilitate interpretation of Figs. 7 and 8 if the sequence of
moval of dissolved oxygen from water has been to employ prolonged experimental runs is presented at this point. Tests were conducted
nitrogen sparging of the water in a packed column. Nitrogen me- in the chronological order listed in the following table.
chanically removes the dissolved oxygen by mixing with and replacing
Fig. Curve Run Liquid Notes
it.
7 IW 1 Water Fresh surface
In the present work, double de-ionized Analar water was sparged
8 IR 1,2 R-1l3
with nitrogen for several hours before using it in a test. Bubbling with
7 IlW 1 Water Rejuvenation
nitrogen was continued inside the rig for the duration (If a test. This
IlIW 1 Water
also provided an inert nitrogen atmosphere inside the rig, which
IVW 1,2,3 Water Nitrogen sparged water
minimized possible redissolution of oxygen in the condensing vapors.
8 IlR 3,4 R-1l3
The use of double de-ionized water with adequate nitrogen sparging
9 VW 1,2 Water Second rejuvenation
eliminated corrosive contamination of the heat transfe! test sur-
face. The data presented in Fig. 7 were obtained with a flat specimen of
Several components of the apparatus, employed at the preliminary surface roughness 0.67 J.tm in boiling water at atmospheric pressure.
stage of experimentation, were subsequently found incompatible with Curve I W was obtained by boiling the fresh surface in de-ionized..-

Journal of Heat Transfer AUGUST 1981, VOL. 103 / 455

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 three times distilled water. For this test, laboratory tap water was first
I0 J
distilled, then de-ionized in a "Permutit Mark 7" de-ionizer, and fi-
nally twice distilled in an all glass apparatus. The test was carried out
in the rig with its final refinements incorporated for the reduction of m w. iv w
contamination, but the liquid was not sparged with nitrogen. At the
conclusion of this experiment with the fresh specimen, faint brown
spots were observed on the surface as a result of localized corrosion.
According to well established reports (e.g., Rohsenow and Choi [1])
in the published literature concerning the performance of a boiling
surface, the results obtained with this fresh specimen should give
lower temperatures, for the same heat flux, than subsequent experi-
ments.
This surface was next boiled in refrigerant R-113 on two consecutive
nw
days for a minimum of three hours per test, to investigate whether or
not contamination with this liquid would occur. As can be observed
in Pig. 8 curve I R, the results are reproduced remarkably well in the
two tests with R-113, It has been reported by Akagawa, et al. [6],
among others, that refrigerant R-113 as well as other organic refrig-
erants decompose at very high temperatures under conditions of film
© FRESH SURFACE, DEIONIZED-3 DISTILLED WATER
boiling. Fouling on platinum heating wires was observed and mea-
O REJUVENATION, 2 DEJONIZED ANALAR1 WATER
sured to be 0.1-2 /urn in thickness depending on the degree of de- 0 DISTILLED- 2 DEDNISED AKALAR' WATER
composition of the liquid refrigerant and the severity of contamina- O.AJ7 RUN 1,2,3 NITROGEN SfftRGED ANALAR'VWER
tion. However, it was also reported that at temperatures below 100°C,
10
R-113 was very stable, and that no deposit was built up in nucleate 2 4 6 8 10 12 14 16 18 20 22 24
boiling. In the present work almost exact reproducibility was obtained (Tw-Tsot) °C
for tests with refrigerant R-113 and no contamination of any sort was Fig. 7 Surface rejuvenation, contamination, and reproducibility of boiling
encountered with this liquid. In fact, its cleaning action is thought to data with water
dissolve possible contaminants which might have built up on the
surface during tests with other liquids.
The surface was next boiled in double de-ionized Analar water in
order to determine whether this high purity water would alleviate the
corrosion problem. Contamination was still present but to a slightly
lesser extent. However, the remarkable change in the boiling surface
performance was striking as can be seen in curve II W of Fig. 7. A
substantial reduction in (TV ~~ ^sat) has occurred as a result of the
intervening experiment with boiling R-113. The test surface reju-
venated with a noticeable increase in bubble population accompanied
by a decrease in bubble diameter as compared with the results of the
E
fresh surface (curve I W). These differences were visually observed.
The so-called ageing effect is thus reversed. As mentioned above,
surface rejuvenation is attributed to the dissolution in R-113 of con-
taminants which reduce cavity size or choke the cavity mouth. Fur-
ther, refrigerant R-113 cleanses the surface of oxide film thus acti-
vating cavities that would otherwise be inactive. A similar hypothesis
was put forth by Turton [7] concerning the behaviour of a stainless
steel tube in boiling water and refrigerant R - l l under the influence
of increased gravity and pressures higher than atmospheric. However,
Turton explained what might be called a rejuvenation of the surface
in successive tests with water as being the result of contamination
which changed the wetting properties of the surface. Such contami-
nation was said to increase the effective cavity size and reduce the
superheat required for nucleation. Contamination being the cause 6 8 10 12 14 15 B 20
of surface rejuvenation in water seems very unlikely in the light of ITw-WC
what has been put forward, and also in what follows. Fig. 8 Effect of order of testing of liquids on boiling data of R-113

Another attempt was made to eliminate surface corrosion, this time

by using distilled double de-ionized Analar water. The results of the The successive step-wise change in the characteristic performance
experiment are shown in curve III W of Fig. 7. The shift in the curve of a surface, such as shown in curves II W, III W and IV W of Fig. 7,
to the right may be caused by the contamination of the surface in- is what may be attributed by some investigators to so-called aging of
curred in the previous experiment of Curve II W, even though the a surface, particularly concerning boiling experiments with water.
surface was thoroughly cleaned with acetone prior to the test. In spite Fand and Keswani [8] report on the particulars of the ageing problem
of the liquid being distilled and double de-ionized, localized corrosion associated with sub-cooled boiling of water from a stainless steel tube.
still occurred although to a much lesser extent. They stated that a set of five experimental measurements corre-
The result of the first test with nitrogen sparged Analar water is sponding to five increasing values of heat flux were repeated in total
shown in curve IV W of Fig. 7. Again, a small shift of the curve to the again and again until the test cylinder was aged; that is, until two
right is observed due to the contamination produced in the previous consecutive sets of five measurements were negligibly different. It was
test (Curve III W). However, at the end of this particular experiment reported that this situation occurred after six hours of operation, and
no visible traces of corrosion spots whatsoever could be detected on the data presented at one heat flux value for two consecutive sets of
the surface. Corrosion was inhibited due to the deoxygenation of the measurements indicated a difference of approximately 0.7°F between
water by the sparged nitrogen gas, which removed the dissolved the two measurements. Furthermore, after aging the surface was ac-
oxygen as described earlier. tivated only for the minimum time sufficient to gather data. It would

456 / VOL. 103, AUGUST 1981 Transactions of the ASME

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 not be wise to speculate, but in the light of what has been mentioned 10-
RUN
concerning contamination, it would not be unreasonable to assume
RUN
that some degree of surface contamination was taking place, the re-
sults of which were being attributed to the so-called aging effect.
In experiments with copper tubes boiling in refrigerant R-Il, Go-
renflo [9] reported that constant values of heat transfer coefficient
were obtained only after 60-80 hr of continuous boiling, and that the
E
final value was approximately 12 percent less than that obtained with
a fresh surface. No assessment was made of the surface condition; but
for copper such a trend would most probably be due to extensive V W
tarnishing and oxidation of the surface.
A second test (run 2, curve IV W) of three hours duration with ni- IV W ,
trogen sparged Analar water and the same surface was conducted four FIG 7)
days after the first run. Immediately after this second test, the power
input to the specimen was turned off for one hour whilst the liquid
was continually boiled by the auxiliary heater and sparged with ni-
trogen. After this period, another test run (No. 3) was then carried
out, the results of which together with those of run No. 2 are shown
in Fig. 7, curve IV W. It can be clearly observed that leaving the sur-
face clean and dry for four days between runs 1 and 2, and that also 20
the prolonged boiling of test runs 2 and 3 did not affect the repro- _L_
ducibility of results. The variation in (Tw — Tsat) for the same heat 6 0 12 V, 16 18 20 22

flux in runs 1, 2 and 3 with nitrogen sparged Analar water was less
than 0.5°C, which is well within the experimental accuracy of ± Fig. 9 Second rejuvenation of surface In water

0.4°C.
It is well known that corrosion cannot be completely eliminated. O RUN 1 , 21 / 5 / 7 6 1
However, localized corrosion of the test surfaces employed in this work CD RUN 2 , 24 / 5 / 7 6 1 ATM.
was inhibited to a large degree by using nitrogen sparged double de- e RUN 3 , 1 3 / 7 / 7 6 .
ionized Analar water, such that experimental data obtained within A RUN 1, 21 / 5 / 7 6
05 ATM
V RUN 2 , 2 4 / 5 / 7 6 .
several hours or indeed several days of operation can be totally relied
, 0 RUN 1 , 2 - 2 5 / 5 / 7 6 , 0 2 5 ATM,
/
upon and reproduced. In the light of this, the concept of a so-called
aging effect should be critically scrutinized, particularly for boiling
experiments in water.
Due to the rejuvenation of the surface for experiments with water
(curves I W and IIW, Fig. 7), it was considered important to investi-
gate the influence of the order of boiling of the liquids. Two test runs
(3 and 4 curve I I R , Fig. 8) were made with refrigerant R-113 on two
consecutive days, one run per day, after all the tests with water
mentioned above had been completed. As can be observed from Fig.
8, curve II R, the data are remarkably reproducible for these two
particular tests but are not the same as those previously obtained
(curve I R). The difference between the two curves is the result of a
change in surface condition due to the intervening tests with
water.
The next logical step was to determine whether the surface would
again rejuvenate in water after the last tests with R-113. Two further
tests (runs 1,2 curve V W) were carried out with N2 sparged Analar
water the results of which are shown in Fig. 9. A second rejuvenation 10
12
did occur with a visually observed increase in bubble population and
( T w - T ^ ) °C
a reduced bubble diameter, as was the case for the first rejuvenation
test (curves I W and II W, Fig. 7). However, the change in surface Fig. 10 Reproducibility of data in boiling with methanol
performance in this case was much less pronounced than that of Fig.
7. The data of run 2 Fig 9 are closer to the values of the reproducible
experiments with water (curve IV W, Fig. 7) than those of run 1. It was between 0.25 and 0.5°C. This small variation was systematic and
thus appeared as if the surface was aging, without corrosion. present in all experiments and was attributed solely to the effect of
Tests with boiling pure methanol were then conducted after the the heat capacity of the heat surface assembly. On this basis, data were
liquid had been distilled in an all glass apparatus with considerable recorded only for decreasing heat flux, so that consistent values of
care being taken to minimize absorption of water moisture from the surface temperature would be obtained.
atmosphere, for which methanol has a very high affinity. Distilled pure Due to the changes in surface performance associated with the order
methanol was sparged with nitrogen, only inside the pool boiler for of boiling of the liquids, it is recommended that tests with surface
the duration of an experiment. Reproducibility of data with boiling liquid combinations described in this work be carried out firstly with
methanol is illustrated in Fig. 10 for all operating pressures. The time refrigerant R-113 then with methanol, and lastly with water.
intervals between the successive tests, as indicated in the figure, are
of the order of days and weeks. Conclusions
Changes in surface performance obtained by increasing and de- Contamination of a heating surface should not be associated with
creasing the power input in established boiling were encountered with the so-called ageing effect of a boiling surface. Localized corrosion of
all the test liquids. For the range of heat flux in fully developed boiling, a stainless steel heating surface in boiling water can be inhibited by
the surface temperature with increasing power input was slightly nitrogen sparging of double de-ionized water.
lower than that obtained with decreasing power input for the same Dezincification of brass components and contamination of a test
value of heat flux. The difference in temperature at any one setting surface in methanol can be totally eliminated by gold plating the brass,

Journal of Heat Transfer AUGUST 1981, VOL 103 / 457

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 distilling the methanol, and employing rig components which are of Heat Exchange During Boiling," Teploenergetika, Vol. 23, No. 6, (1976), p.
nonreactive with methanol. 12.
3 Joudi, K. A., and James, D. D., "Incipient Boiling Characteristics at At-
Rejuvenation of a stainless steel heating surface can be effected by mospheric and Sub-Atmospheric Pressures," ASME JOURNAL OF HEAT
boiling the surface in a strong cleaning agent such as refrigerant R-113 TRANSFER, Vol. 99,1977, p. 398.
which dissolves surface contaminants. 4 Ross, T. K., "Corrosion and Heat Transfer—a review, "British Corrosion
The order of boiling of liquids on a given test surface can result in Journal," Vol. 2,1967, p. 131.
5 Joudi, K. A., "Surface Geometry Variations in Nucleate Pool Boiling,"
variation in boiling characteristics with each liquid due to rejuvenation Ph.D. Thesis, UMIST, Manchester, UK, 1977.
or oxide film deposit on the surface. 6 Akagawa, K., Sakaguche, T., and Fujii, T., "Influence of Fouling on Boiling
Heat Transfer to Organic Coolants," 5th Int. Heat Transfer Conference, Tokyo.
Acknowledgments JSME, Vol. IV, Paper No. B1.6,1974, p. 25.
7 Turton, J. S., "The Effects of Pressure and Acceleration on Pool Boiling
Thanks are due to Dr. J. A. Richardson of the Corrosion and Pro- of Water and Arcton 11," International Journal Heat Mass Transfer, Vol. 11,
tection Centre, UMIST, for much useful discussion. 1968, p. 1295.
8 Fand, R. M. and Keswain, K., "The Influence of Sub-Cooling on Pool
Boiling Heat Transfer from a Horizontal Cylinder to Water," Proceedings 5th
References International Heat Transfer Conference Tokyo, JSME, Vol. IV, Paper No.
1 Rohsenow, W. M. and Choi, H. Y., Heat, Mass and Momentum Transfer, B1.3,1974, p. 11.
Prentice Hall, International, London, 1961. 9 Gorenflo, D., "Zum WSrmeubergang bei der Blasenverdampfung an
2 Nestrueva, E. I, Romanovskii, I. M, and Kashinskii, V. I., "Some Aspects Rippenrohren," Diss T. H. Karlsruhe, W. Germany, 1966.

S59 Laminar Mixed Convection in a Shrouded Fin Array

S. Acharya and S. V. Patankar
566 Combined Free and Forced Convection in Vertical Tubes with Radial Internal Fins
C. Prakash and S. V. Patankar
S73 Some Fundamental Relationships for Tubular Heat Exchanger Thermal Performance
K. P. Singh
579 A General Theory of Wet Surface Heat Exchangers and Its Application to Regenerative
Evaporative Cooling
I. L. Maclaine-cross and P. J. Banks
S86 A Parametric Study of a Particulate Director Contact Heat Exchanger (78-WA/MT 16)
R. Letan
591 Conduction of Heat across Rectangular Cellular Enclosures
J. Eftekhar, G. Darkazalll, and A. Haji-Sheikh

TECHNICAL NOTES
596 A Closed-Form Analytical Solution for Freezing Adjacent to a Plane Wall Cooled by Forced
Convection
C. F. Hsu and E. M. Sparrow
598 Phase Change around a Circular Cylinder
V. J. Lunardini
600 Heat Conduction through "Yin-Yang" Bodies
J. H. Lienhard
601 Transient Thermal Expansion of Solids During Inert Heating, Phase Change, and Surface
Gasification
L, W. Hunter
602 Extension of the Adams-Welty Fluid Bed Heat Transfer Model to the Packed Bed Case
R. L. Adams
604 Thermal Storage Regenerator Parameters for Almost Constant Gas Discharge Temperature
F. E. Romie
606 Effect of Knudson Number on Dropwise Condensation
H.Tanaka
527 Erratum on a previously published paper by C. E. Hickox

458 / VOL. 10.3, AUGUST 1981 Transactions of the ASME

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 Destabilization of Film Boiling
Due to Arriwai of a Pressure
A. Inoue 1
S. G. Bankoff Shock: Part I Experimental2
Chemical Engineering Department, Transient heat transfer from an electrically-heated 3 mm o.d. horizontal tube, initially
Northwestern University, in subcooled film boiling, was measured immediately after passage of a shock wave of 1-5
Evanston II. 60201
X 10b N/m2 over-pressure. The fluids tested were Freon-113 and 95 percent ethanol-5
percent water at initially 0.5-2 X 105 N/m2 at 22-24°C. Transient heat transfer rates, av-
eraged over 0.5-1 ms after vapor film collapse, ranged up to 20 times the steady-state
value. The maximum transient flux occurred at supercritical contact temperatures, with
frequently a minimum in the range of contact temperatures between the homogeneous
nucleation and the critical temperature. Photography at 5000 frames/s showed apparent-
ly complete vapor film collapse within one or two frames, followed by re-establishment
of film boiling in ~ i ms, and eventually nucleate boiling in ~100 ms. The surface temper-
ature which gave the highest peak transient flux shifted appreciably with increasing
shock pressure, which indicates some compressibility even after "contact" was made. Im-
plications for vapor explosions are discussed.

1 Introduction 2 Equipment and Procedure

Violent, or even explosive, boiling may result when a very hot liquid, A low-pressure vertical shock tube was constructed for measuring
such as molten UO2, steel or aluminum, is mixed rapidly with a colder, destabilization of film boiling on a horizontal nickel heater by plane
vaporizable liquid, such as sodium or water. Such processes, desig- shocks of up to 2 MPa overpressure (Fig. 1). It consisted of a test
nated generically as fuel-coolant interactions (FCI), are of concern section constructed of thin-walled 50 mm o.d. stainless steel tubing
in the nuclear reactor field, among others. It is believed that an im- separated from a high-pressure driver section by a thin mylar mem-
portant step in the triggering of small-scale explosions and in the brane. The pressure in the driver section was controlled by a regu-
triggering and propagation of large-scale explosions consists of the lating valve from 0-2 MPa. The membrane was broken by means of
destabilization of the film boiling resulting from coarse premixing of a piston actuated by a solenoid valve with a time delay adjusted by
the two liquids. Destabilization is thought to be favored by a decrease means of a needle valve. This time delay was necessary to prevent the
in the fuel or coolant temperature (or both), increase in relative ve- electrical noise produced by opening the valve from contaminating
locity of the two fluids, or by a sudden increase in pressure. However, the pressure transducer signals.
the time scales for the pressure rise in explosive boiling are 0 (10 - 4 ) Prior to breaking the membrane the liquid was circulated at a low
s, so that it is very difficult to obtain quantitative information in velocity by a pump through an external heat exchanger to maintain
liquid-liquid systems. It seemed likely, therefore, that useful infor- a constant temperature. When the shock was generated it traveled
mation concerning the mechanics of triggering and propagation might about 0.3 m down the test section through nitrogen and 0.11 m
be obtained by studying the transient heat transfer behavior of an through the liquid before reaching the heated tube, and then an ad-
horizontal electrically-heated nickel tube, initially in film boiling ditional meter through the liquid column to reach the bottom of the
with respect to the surrounding subcooled liquid, upon arrival of a shock tube. Some metallic sponge was placed at the bottom of the test
moderate pressure shock wave. In this part the experimental results section to reduce the reflection of the shock.
with Freon-113 and ethanol over a range of initial pressure, final A sidearm with a second diaphragm similarly controlled (MV2) was
pressures, shock rise times, and surface temperatures are presented. available for adjusting the shape of the pressure pulse, but was not
The second part deals with a theoretical treatment of the vapor film used in this series of experiments.
collapse process. The test section could be maintained initially at sub-atmospheric
There seem to be no previous studies with which the present results pressures by means of a vacuum pump protected by a cold trap. Just
can be directly compared. In the general area of liquid-solid film before starting each experimental run, the main test section was iso-
boiling destabilization, Bradfield [1] and Stevens and Witte [2] lated from the circulation loop by the magnetic valves, after which
plunged a hot graphite cylinder and a hot silver sphere, respectively, the shock was generated. The steady-state pressures in the driver
into cold water, and observed that a very rapid destabilization of film section and in the test section were measured by high precision
boiling (termed a "transplosion") occurred when the wall temperature pressure gauges, which were also used for calibration of the pressure
cooled to a threshold value, usually above the critical temperature. transducer for transient measurements of the shock pressure. The
In a different vein, a pseudofilm boiling phenomenon has been ob- thermocouple for measurement of the liquid temperature was placed
served above the critical pressure of water, due to the rapid variation below the glass section. The initial test section pressure could be
of specific heat in the neighborhood of the pseudocritical temperature controlled from the vapor pressure of the liquid to 0.5 MPa.
[3]. A series of orifices was used to control the pressure rise time. As-
suming a first-order response of the form

P~(t) = Pi + (PF - Pi) (1 • • e x p ( - t / T p ) ) (1)

1
Associate Professor, Research Laboratory for Nuclear Reactors, Tokyo the time constant could be varied from T. = 8 X 10~5 s with no orifice
Institute of Technology, O-okayama, Meguro-ku, Tokyo, Japan.
2
This work was supported by the U.S. Department of Energy. The test heater on which film boiling was established consisted of
Contributed by the Heat Transfer Division and presented at the 19th a centerless-ground nickel tube 3.0 mm o.d., 20-25 mm in length and
AIChE/ASME National Heat Transfer Conference, July 27 to 30,1980, Orlando,
Fla. Revised Manuscript received by the Heat Transfer Division November 0.06-0.08 mm in thickness. The tube was silver-soldered at its ends
7,1979. to 4 mm dia copper busbars which entered the test section through

Journal of Heat Transfer AUGUST 1981, VOL. 103 / 459

Copyright © 1981 by ASME
Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 teflon seals. The nickel tube was heated by passing direct current from
six 200 amp-hr 6 V truck batteries connected in parallel. The current
was controlled by a water-cooled stainless steel trombone rheostat
with copper busbar slides driven by individual lead screws. In view
of the significant axial heat losses to the busbars, 0.1 mm nickel wire
leads were spot welded across the center 10 mm span for voltage
measurements of the isothermal portion of the tube.
The time-dependent test heater power and average temperature
were determined by measuring the current and voltage drop across
the center section of the heater tube by means of a double bridge
circuit. It was necessary to provide a bypass current sufficiently large
to avoid burnout at the transition from nucleate to film boiling (Fig.
2). Details of the circuit design are given in Appendix I. The transient
pressure was measured 70 mm above the test heater by a piezoelectric
transducer with 2 jus rise time; 250 Khz resonance frequency; 700
N/m 2 resolution. The travel time of the shock from the transducer
to the test heater was about 0.05 ms for a shock speed of 1500 m/s. The
shunt voltage drop, the bridge unbalance voltage and the pressure
transducer output were recorded on an FM tape recorder with a
maximum frequency response of 5 KHz. The temperature of the test
section, the pressure, the heat transfer coefficient and the heat flux
were then calculated by playing the tape recorder into an analog
computer at 32 speed. The computer output was recorded on an os-
cillograph. Additional experimental details are given elsewhere [4].
The instantaneous heat flux, qm, was calculated from a heat bal-
ance:

qm = qss ~ Mt (1)
dt
where qss is the steady power to the test heater per unit surface area;
Mt is the mass of the test heater per unit surface area; and Tm is the Vacuum
average temperature of the center portion of the test heater. The heat
transfer coefficient, h, was then computed from qm and ATm - Tm
— TV The peak transient heat flux, qp, was estimated by fitting a
—6 ns Pump

tangent to the Tm versus t trace immediately after passage of the

shock. Usually this trace exhibited a linear decrease over the first Fig. 1 Sketch of shock-tube experimental facility: T = temperature ther-
0.5-1.5 ms. The thermal boundary layer thickness in a nickel slab is mocouple; P = pressure gauge
~0.3 mm after 1 ms, so that the assumption of uniform metal tem-
perature in the heater tube is acceptable. A more detailed analysis is
given in Appendix II.

3 Experimental Results RUN 771

Figures 3 and 4 illustrate the effect of the initial metal temperature
on the pressure transducer and the heater temperature signals as a
function of time after the passage of a sharp shock (rise time 80 ms).
The vapor film disappears completely in 1-2 frames at 5000 frames/s.
With an initial metal temperature of 195°C the film begins to be re- R = 1.0«105 N/m2abs
established after several milliseconds, but is collapsed again by the R=6.03 •
reflected shock from the bottom. This repeats several times. At no iP-5.03 •
time is nucleate boiling visible, as evidenced by the absence of a milky V 80/s
TL= 22 t
appearance due to scattered light. With an initial metal temperature
of 331°C, the liquid-solid "contact" time is shorter, and the vapor film
which then forms is more stable. Nucleate boiling commences in times
of order 10 2 -10 3 ms.
Additional pressure and temperature traces are given in reference
[4]. In general the temperature behavior just after the shock appears
to fall into three classifications depending upon the initial wall tem-
0 . 5 10 t 15 m s
perature, Tmo, the shock pressure rise AP = Pp — Pj, the pressure rise
time, TP and the initial test section pressure Pi: Fig. 2 Boiling curves under steady-state and transient conditions

-Nomenclature-
Mt = mass of heated tube per unit external Ti = calculated interfacial contact tempera- Subscripts
surface area ture crit = critical
Pi = initial pressure of test section Tm = average temperature of center portion Homo = Homogeneous nucleation (same as
Pp — final pressure of test section of heated tube spontaneous nucleation for well-wetting
q = surface heat flux of heated tube J mo = heated tube initial average tempera- surface)
qp = peak surface heat flux after shock pas- ture of center portion of heated tube £ = liquid
^ sn = spontaneous nucleation temperature max = maximum
T = temperature TP = pressure rise time

460 / VOL. 103, AUGUST 1981 Transactions of the ASME

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 1 1 '
MW RUN 2 5 0 - 276
Pi =1.0X10* N/m'
F-113
P F = 3.99 x10 s »
- A P =2.99X105 »
r.=80/*s
o
T, =22 - C

o
- o
0
oo o o
0

1
o o
0
o 0 o

if
r o
I
II
o
o o

I 1
I Steady state heat flux.
1
100 , 200 _. 300 400
1

Tmo

Fig. 3 Time histories of pressure 7 cm above heated tube and the metal ig.5 Peak heat flux versus initial metal temperature at shock pressure AP
5 2
temperature for sharp shock in freon 113 at the Initial temperature 19S°C 2.99 X 10 N/m , pressure rise time T P = 80 (is for freon-113

RUN 753 MW RUN200-241

m' Pj=L0X105 N/m'
F-113
PF-2.27X10S a
3
AP = 1.27X10*1"
AIJAAYV"^ r = 80#sec
T,=22°C
o
R=1.0 «105N/m2 1p
f?=6.03 « 2
*P=5D3 "
o
TL=22 °C
O
o
o
1 -
o ° i
H o o
-3
0
0 °
"» ° n °
F o " ^
»• %: Steady state heat «ux° ° ° ° °0aP o
1

0 5 10 15 t 20 25 ms 30
1

100
1
, 200
, 300
i

500
: lmo

Fig. 4 Same as Fig. 3, except the initial metal temperature 331°C Fig. 6 Same as Fig. 5, except AP = 1.27 X 105 N/m2

1 No significant temperature change, corresponding to the ratio occurs at a much higher theoretical contact temperature than the
of the peak heat flux to the steady-state film boiling heat flux qp/qSs critical temperature of the liquid. However, the actual surface tem-
< 1.5, found under conditions where T P > 3.3 » 300°C perature may have been subcritical, due to the presence of a thick
01" 1 mo surface oxide layer. Similar results have been reported by Ando and
« 300°C Caldarola [11]. In the present tests the oxide layer, if any, was of
2 Slow temperature decrease but no rapid decrease in the initial negligible thickness on the polished nickel surface, and there seems
1 ms: <j p /<? ss ~5 to be no doubt that the initial contact temperature was supercritical.
3 A sharp temperature drop in less than 1 ms after the shock with The maximum of qp around Tmo ~ 300° C is thus independent of
qPlqSs > 5 nucleation and bubble growth phenomena, which control the behavior
There were no sharp transitions between each classification. The last below Tcrjt. The heat flux maximum can be explained by the opposing
category is favored by the following conditions: T m o = 250-350°C, tendencies represented by the increased temperature driving force
corresponding to reduced temperatures Tr = 1.07-1.28; APp/APj > for heat transfer versus the increased vapor film stability as the metal
3; and rp < 0.15s. The last two parameters seem to be more important temperature increases. This is consistent with measurements of liq-
than the other two. More detailed information concerning these uid-solid contact in steady-state film boiling of water from a gold-
parametric effects can be obtained by investigating the peak heat plated surface [6]. An attempt to make similar measurements here
flux just after the shock. with an ethanol-water mixture was unsuccessful due to vibration of
The data for a large number of runs are summarized in the form of the fine electrical-resistivity probe after passage of the shock.
peak heat flux, qp, as a function of the initial metal temperature, Tmo.
Figure 5 shows the results for a sharp shock with Freon-113 initially It is also interesting that qp may exhibit an apparent minimum in
at atmospheric pressure. There is a considerable scatter of qp, partly the range Tsn < T,- < TCT[t, where Tsn is the spontaneous (in this case
due to errors in fitting a tangent to the curve of Tm(t) at t = 0, and homogeneous) nucleation temperature, T"Homo. This is the case with
partly due to random variations in the initial vapor film thickness at Freon 113, but not with 95 percent ethanol-5 percent water. The ap-
the time of shock passage. Nevertheless, it is clear that the peak heat pearance of this minimum is the opposite of the behavior observed
flux increases sharply with initial metal temperature above the critical with liquid-liquid contacts, where this range is explosive, providing
temperature up to about 300° C and then begins to decrease. Since coarse premixing followed by a trigger (vapor film collapse) is
the interfacial contact temperature, T;, is within 10° C of Tmo, this achieved. There is, however, a simple and consistent explanation.
implies that very high heat fluxes can be obtained in the absence of Above Tsn film boiling is re-established very rapidly. With a solid
true liquid-solid contact. It may be noted that Zyszkowski [5] also (constant-area) surface this may correspond to decreased heat transfer
observed that a violent interaction with a molten copper drop in water below T cr it, depending on the thermal properties of the solid and

Journal of Heat Transfer AUGUST 1981, VOL. 103 / 461

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 liquid. However, a liquid surface can exhibit very rapid increase in Table 1
contact area due to pressure-driven local mixing, and hence increased
AP, MPa T T
energy transfer.
0.13 270
Figures 5 to 7 show the effect of the shock pressure, PF, with an 0.30 300
initial pressure Pi = 0.1 mPa, and all other conditions remaining 0.50 330
constant. One finds that T m a x , the temperature at which the maxi-
mum <jp appears, increases with AP, the shock pressure jump, as
shown in Table 1.
The effect of reducing the initial test section pressure, Pi, by a MW RUN 700-729
m' F-113 o P i = 1 . 0 X 1 0 s N/m'
factor of two can be seen by comparing Fig. 8 with Fig. 5. It is seen that P F =6.03X10 5 H
3
the heat transfer coefficient is increased by roughly a factor of two, AP =5.03X10 5 n
indicating again closer approach of the liquid to the wall, with the peak T =80^sec
heat transfer coefficient being roughly proportional to the initial mass Ti=20~25°C
O

of vapor per unit area of heater surface. This suggests that the relevant 1p o°
parameter may be PF/PI, rather than AP. This is in agreement with 2
- o
-
o o
the experimental results of Sharon and Bankoff [9] for molten
metal-water interactions, and also with the theoretical model, to be °

described in Part II. Figure 9 shows the effect of the pressure rise time, o °
which was varied by employing orifices of different diameters. The
peak flux becomes quite high for small T P , as shown for T P = 80 ^ts and
1
rf> ° » o -
125 fis. For TP > 3.3 ms, qp is less than twice the steady-state value, H
and is not strongly affected by Tmo or T P . One concludes that evapo-
q „ : Steady slat hent flux
ration from the approaching liquid surface becomes important with
slow rise times, but is negligible for fast rise times until the intervening (1 , , ,
200 „,
I
300
1
•c 500
vapor film becomes very thin. 1 mo

The collapse of the vapor film was photographed at 5000 frames Fig. 7 Same as Fig. 5, except A P = 5.03 X 105 N/m 2
per second. The collapse of the vapor film appears to be completed
at 0.1-0.3 ms after the shock. This is the same time interval during
which the very sharp temperature drop and peak heat flux are ob-
MW
served. It is not possible to determine from these films whether the RUN 3 0 0 - 3 3 3
F-113
vapor film is completely or partially collapsed, but no liquid-vapor P, =0.567X10 5 N/m'
interface is visible. This condition continues until about 0.7 ms after PF =4.16X10* •'
the shock when the formation of a vapor layer becomes apparent. No P = 3.41x10 s a
small bubbles are observed during this time, as would be indicated r =80*:s
by light scattering, and it would appear that nucleate boiling does not T, = 20~24"C
occur, as might be expected when Tw > Thn-
The peak flux just after the shock for 95 percent ethanol-5 percent
water at Tmo of 341°C is only slightly higher than for F-113 at 295°C,
as shown by comparing Figs. 5 and 10, despite the fact that the ther-
mal conductivity of the ethanol is three to five times greater than for
Freon-113 in the liquid and vapor states. The minimum in qp for Tm
< Tcrjt is also not apparent. The reasons for these differences in be-
havior are not known, but may be related to wetting properties, as well
as thermal properties, of the fluid. q „ : Steady slate heat flux

200_ „ 300 400

4 Discussion and Conclusions Tmo
The completeness of initial collapse of the vapor film will be dis-
Fig. 8 Same as Fig. 5, except initial test section pressure P, = 0.567 X 105
cussed in more detail in Part II. We note here, however, that the ob- N/m2abs, A P = 3.41 X 10s N/m 2
served maximum average heat fluxes can be accounted for heat con-
duction through a high-pressure vapor layer of 10~ 7 to 1 0 - 8 m
thickness. Since the surface roughnesses are 0(10~7 m), at least partial .MW
liquid-solid "contact" is expected. We define contact here as the ab- m' RUN 400-496
P,=1.0X10 S N/m
F-113
sence of an intervening vapor phase. Thus, at any particular location 1.6 P P =3.99X10 B a
the local pressure may be above the critical pressure, which corre- o T, =20~25"C
sponds to local contact, or below the critical pressure, in which case 1.4 O ° r,=80#s
a vapor layer can exist. True liquid-solid contact can exist only when 0 * r,=125//s
Ti < Tcrit. For Ti > TCrit the fluid compressibility is significant in the 1.2 * r p =3.3ms
q
°
P n T„ = 19.4ms
neighborhood of the solid surface. Hence one would expect that the 1
space- and time-average contact efficiency would increase with AP, • Tp = 0.11sec
0.8
O 0 v r p =0.5sec
as shown in Table 1.
One can draw from these results the important conclusion that the 0.6-
energy transfer in triggering and early propagation of vapor explosions
occurs by rapid fragmentation and formation of new contact area 0.4
during the elementary propagation cycle.
Another conclusion relates to the upper bound of metal vs. coolant 0.2 »,?•" V
Q u : Steady State Heat Flux
temperature interaction zone (TIZ) reported by Reynolds, et al. [8]
0
for small quantities of tin poured into water. The results of Table 1 0 100 200 _ 300
suggest that this upper bound, corresponding to the establishment Imo

of effective contact, can be shifted upwards by a shock wave, corre- Fig. 9 Same as Fig. 5, except pressure rise time constant r p varied from
sponding to an energetic trigger. This is confirmed by recent mea- 80 fis to 0.5 s

462 / VOL. 103, AUGUST 1981 Transactions of the ASME

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 MW 1 1 ' 10 Ochiai, M., and Bankoff, S. G., "A Local Propagation Theory for Vapor
m'
RUN 500-541. Explosions," Paper SNI 6/3, Proceedings Third Special Meeting on Sodium/
Fuel Interactions in Fast Reactors, Tokyo, Japan 1976.
P , = 1.0X10S N/m' 95%Ethanot . 11 Ando, M., and Caldarola, L., paper presented at OECD CSNI Joint In-
P F = 3.99X10 5 11 terpretation Exercise on Selected Fuel-Coolant Interaction Experiments,
T =80/is Karlsruhe, Germany, Dec. 1980.
T,=68'C 12 Amblard, M., and Jacobs, H., "Fuel-Coolant Interactions: The CORECT
0 II V02-Na Experiment," Proceedings ANS/ENS International Meeting on
qP 0 0 Fast Reactor Safety, Seattle, Wash., Vol. 3, 1979, pp. 1512-1519.
- % 0 0
0
So 0
- 0
APPENDIX I
0
0
00
0
0
Heating and Temperature Measurement Circuit
0
o
*~~ H Q% °° The power to be dissipated increased with heater temperature in
i? o°o constant-current heating, since the resistance increases with tem-
3? 1 perature, possibly resulting in departure from nucleate boiling (DNB).
'. steady state heat flux
1 1
Therefore, insofar as possible a constant-voltage power circuit is de-
100 200 300 400 500 sired. Another consideration is that the ends of the heated nickel tube
Tmo "C
will be cooler than the center portion, due to heat losses to the copper
busbars to which the tube is attached.
Fig. 10 Same as Fig. 5, except 95 percent ethanol-S percent water at the
liquid temperature 68°C To measure the average temperature of the uniform central region,
a double bridge method was used. The bridge elements consisted of
the test heater resistance Rt (M).7 — 1.0 X 10~ 3 fi); a precision shunt
surements by Sharon and Bankoff [9] in a shock tube of liquid resistance Rs (5 X 10~5 Q); a variable bypass resistance, Rb = Rt, and
metal/water interactions with initially established film boiling. In fact, a slide resistance Rv (0.003 - 0.186 12), made from 3 ft lengths of
the plots of heat flux immediately after shock passage versus metal water-cooled stainless-steel tubing; a standard resistance Rc (5 0 ) ;
temperature shown in Figs. 5-10 are remarkably similar to plots of and a decade box reading to four digits (0 — 999.9 Q).
peak pressure versus metal temperature in molten tin-lead alloy/water We assume that the resistance of the copper cables and the contact
interactions following shock wave passage (Sharon and Bankoff resistance of each connection can be neglected.
[9])- The resistance of the nickel heater is
These results further imply that vapor explosions can propagate
in the supercritical contact temperature, but subcritical pressure, Rt = Rto (1 + a A T m ) , (I-D
range, with only a weak pressure shock as the initial triggering event. where a is the temperature coefficient of resistivity, Rto is the test
In fact, no large-scale explosions have been reported when T; < Tcrn, heater resistance at the DNB temperature and ATm is the tempera-
and nearly all peak pressures obtained to date have been below the ture rise of the heater above the DNB temperature. In this case it can
coolant critical pressure (Amblard and Jacobs [12]). be shown that
One can carry this reasoning one step further, and suggest that
vapor explosions are nearly always propagated by boiling, or boil-
ing-like, mechanisms. This means that at subcritical pressures behind
the shock front mixing and fragmentation result from bubble growth
(1 + aATm) J^LATm d-2)
[Qo] 1+
and collapse. At supercritical pressures regions of large density vari- i + K2
ations can also exist, giving rise to similar behavior. However, the fact
that "contact" is not uniform means that large pressure gradients Rt,
where #i =
tangential to the interface can exist. This may be the principal Rs + Rto
mechanism for rapid mixing. A simplified model for local mixing due
RbRu
to local variations in contact pressure is the "splash" theory of Ochiai K2 = -
and Bankoff [10]. (Rb + Rv) (.Rs + Rto)
and [Qo] is the heater power at DNB, while [Q] is the power at a
References temperature other than DNB.
1 Bradfield, W. S., "On Liquid-Solid Contact in Stable Film Boiling," Rpt. In the present case Kt = 0.95 (when Rt0 = 10~ 3 Q at DNB), Rb +
No. 36, State Univ. of New York, College of Engineering, Stony Brook, 1965. Rto and R„ » Rto
2 Stevens, J. W., and Witte, L. C , "Destabilization of Vapor Film Boiling Therefore Kx = 1. It then follows that
around Spheres," International Journal of Heat Mass Transfer, Vol. 16,1973,
pp. 669-678.
3 Hebel, W., and Decreton, M., "Mechanisms of Critical Heat Flux in a (1 + aATm) d-3)
Stagnant Water Annulus," 2nd Multiphase Flow and Heat Transfer Sympo- [Qo] U + 0.475 ATm
sium—Workshop, Miami Beach, Fla., Apr. 1979.
The curve of [Q]/[Q0] is shown in Fig. 10. From the intersection of the
4 Inoue, A., and Bankoff, S. G., "Destabilization of Film Boiling due to
Arrival of a Pressure Shock," Rpt. C00-2512-13, Chem. Eng. Dept., North- power curve with the film boiling curve it is seen that the temperature
western Univ., Evanston, 111., 1978; also in Topics in Two-Phase Heat Transfer at DNB is limited to 520° C. However, in actuality the resistances of
and Flow, S. G. Bankoff, ed., ASME, N.Y., 1978, pp. 77-88. the copper cable and contacts are of the same order as that of the
5 Zyszkowski, W., "Experimental Investigation of Fuel-Coolant Interac- heater and the actual circuits, so that in the present equipment about
tion," Nuclear Technology, Vol. 33,1977, pp. 40-59.
6 Yao, S. C , and Henry, R. E., "An Investigation of the Minimum Film a 20 percent decrease in power is measured at DNB.
Boiling Temperature on Horizontal Surfaces," ASME JOURNAL OP HEAT
T R A N S F E R , Vol. 100, May 1978, pp. 260-267.
7 Hsu, Y. Y., and Graham, R. W., Transport Processes in Boiling and APPENDIX II
Two-Phase Systems, McGraw-Hill, New York, 1976.
8 Reynolds, J. A., Dullforce, T. A., Peckover, R. S„ and Vaughan, G. J., Test Heater Surface Temperatures
"Fuel-Coolant Interactions—Some Basic Studies at the UKAEA Culham
Laboratory," Paper SNI6/2, Proc. Third Spec. Meeting on Sodium/Fuel In- As noted above, the average wall temperature of the nickel tube is
teraction in Fast Reactors, Tokyo, 1976. determined by measuring the electrical resistance. The nonuniformity
9 Sharon, A., and Bankoff, S. G., "Fuel-Coolant Interaction in a Shock in the temperature in the axial direction can be avoided by measuring
Tube with Initially-Established Film Boiling," Proceedings ANS/ENS In-
ternational Meeting on Fast Reactor Safety, Seattle, Wash., Vol. 4,1979, pp.
the resistance only of the central portion of the tube, but the radial
1796-1810. differences may be significant under highly transient conditions.

Journal of Heat Transfer AUGUST 1981, VOL. 103 / 463

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 a Temperature Difference at Steady-State. The temperature where
difference at steady-state between the mean metal temperature and
Oimt
the wall surface temperature is given by the steady-state heat con- t*
duction equation. For a steady-state surface heat flux qss = 100 w/cm 2 ,
2
Tm-Tw=lQ8m /km = 0.40°C, where the tube wall thickness 5 = If this function is approximated by e t*/T*) the time constant T* is
0.07 mm, Tm is the average metal temperature and Tw is the wall
2
surface temperature. This difference is therefore negligible for heat ft
^0.33 or T = 0.33- (II-7)
fluxes in the film boiling regime in this experiment (14-19 w/cm 2 for
Tm = 450°C).
For the heated tube r = 0.33 (0.007)2/0.1496 = 0.11 ms, so that the
b Temperature Difference under Transient Conditions.
time lag between the average and surface temperature may be ne-
an> = a a 2 r m t Q 01-1)
glected.
dt dx2 pmCp (ii) Temperature Difference during Ramp Transient. To get
actual surface temperature Tw(t) from the average metal temperature
I.C. Tm(x, 0) = TW(Q)•+• If-
. A ( 5 m 2 - x2) Tm(t), by rearranging equation (II-3):
Km

where: km = pmCpmam [Ta(p)] - — - ~ 5 m 2 - = 8mqm (1 + e- 2 "" 5 -) £ e-2^"

dTm p okm p n=0
B.C. a t * = 0:- 01-2) TmW]
dx X [Tm(p)] - (H-8)
at x = S:Tm(8,t) = Tw(t)
We truncate the series after the first term for a short transient:
The space-averaged Laplace transform solution [Tm(p)] of equation
(II-l) with conditions (II-2) is 2
[TW(P)] - ^ ^ - - S - ^ - = V ? - (1 -e-2"^)
P 3km p
[Tm(p)]-[Tw(p)}=\^bm2-
3xm p Tm(0)}
oQm°m -QmSm
[Tm(p)] - 01-9)
[Tw(p)}
+ ( Qmim+
e e-«m«m)8m<[„
By inverse transformation of equation (II-9)
TW(Q) f e-"^" p<im&m

(eqmSm + e-qmt>m) ftmqm

Ta(t)-Tw(0)-]-^-5m2
10 1 a Km
2 ( 1 — e - 2 l 7m5m) £ (—\)n e-2nqmim
: -—ft - + Tm(0)\
3km O &mQm n=0 = £.- [8mqm (1 - e~2i^) \[Tm(p)}- (11-10)
T»(0)
- [T„(p)] (H-3)
If Tm(t) is approximated as a ramp function, Tm(t) = Tm(0) — ATmt,
where in the initial stage of the transient, one obtains, by substituting the
Laplace transform of Tm(t) into equation (11-10),
;
qm
Tw(t) - Tu(0)-~ 6m*
(J) Step Response of Tm(t). 3 km
ATm Sm2/am 2p
Let Tw(t) = T J O ) - ATwH(t) (II-4)
2
V t * ( l - e - ' - + 2erfcl. (11-11)
where H(t) = 0 , t = 0 yf-K
H(t) = 1, t > 0
Substituting Laplace transform of equation (II-4) into equation (II-
3): ™~Zm(l);iQSmmm - »• --1= V? (i - e+) (n-12,
[Tm(P)]~\^-bm2- In our case
3 km p
AT„/ 1 _ fi-29m Tw(t)-Tm(t)--^-8m2
- £ (-l^e-^m^l (H-5)
+ P I 5m(Jn n=0
2km
= 3.3 X 10r 4 Atm It* - —= yjt* (1 - c"'1')] (H-13)
If we take n - 0 and 1, since we are only interested in small time be-
havior, and perform the inverse transformation, equation (II-5)
gives
< 3.3 X 10- 4 Afm X 0.3
Tm(t) - Tw(t) 4
= l--=y/t*(l-2 e-W + e- /'* ) In the present experiments, the maximum of ATm immediately after
ATW
passage of the shock was found to be 10240°C/s at $8m = 19 w/cm 2 .
+ 4|erf—L=-erf-jL) (II-6) In this case, the maximum difference of temperature is Tw(t) — Tm(t)
\/t* Vt*l < 1.06°C, and hence is negligible.

464 / VOL. 103, AUGUST 1981 Transactions of the ASME

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 Destabilization of Film Boiling Due
A. Inoue to Arriwal of a Pressure Shock2
Associate Professor,
Research Laboratory tor Nuclear Reactors,
Tokyo Institute of Technology, Part II: Analytical
O-okayama, Meguro-ku,
Japan
Vapor explosions are believed to be triggered by the rapid collapse of film boiling of cool-
ant in contact with molten fuel, probably due to local pressure waves from an initially
A. Gangufli1 small interaction. In Part I of this work the heat transfer during the first two ms after pas-
S. G. Bankoff sage of a shock past a hot nickel tube surrounded by subcooled Freon-113 or ethanol was
Chemical Engineering Department,
studied. The following important results were obtained: (1) The peak heat flux exhibits
Northwestern University, a maximum of a heater surface temperature of 280-350° C, depending upon the strength
Evanston, III. 60201 of the shock. This is well above the critical temperature, so that nucleation considerations
are irrelevant. (2) The maximum of the peak heat flux envelope depends upon the shock
AP, indicating that only partial contact is made upon collapse of the vapor film. (3) The
collapse is rapid (1-2 frame at 5000 f/s), and is produced by relatively weak shocks (AP
= 2-3 atm.). In the present work, the vapor film collapse is studied analytically, in order
to obtain additional insight into the mechanism. A Lagrangian transformation due to
Hamill and Bankoff is introduced to immobilize the moving boundary, and a polynomial
temperature distribution in the transformed mass variable is assumed in the vapor region,
as well as in the liquid region. This leads to a set of coupled nonlinear ordinary differen-
tial equations in the three regions, which are solved numerically. Two models were devel-
oped: (1) A detailed model, taking into account the Knudsen layers at the vapor-liquid
and vapor-solid interfaces, and (2) a simplified model, in which these layers were neglect-
ed, and a linear temperature profile in the Lagrangian vapor phase variable was assumed.
It is found that the initial vapor mass is a key variable determining whether collapse is
achieved. In practice, this is a stochastic variable due to bubble departure, which explains
the observed heat flux data scatter. The analytical results are in general agreement with
the experimental data.

1 Introduction 2 Model
Violent, or even explosive, boiling may result when a very hot liquid, Complete and simplified models are developed for the approach
such as molten UO2, steel or aluminum, is mixed rapidly with a colder, of the vapor-liquid interface toward the solid surface after the shock
vaporizable liquid, such as sodium or water. Such processes, desig- has passed. The aim of this analysis is to study the effect of several
nated generically as fuel-coolant interactions (FCI), are of concern parameters on the peak heat flux and to estimate the possibility of
in the nuclear reactor field, among others. It is believed that an im- direct contact.
portant step in the triggering of small-scale explosions and in the Even the complete model represents a first-order approach to the
triggering and propagation of large-scale explosions consists of the actual phenomenon, which is considerably more complex. The surface
destabilization of the film boiling resulting from coarse premixing of roughnesses are of order 2ju, so that the calculation loses physical
the two liquids. Destabilization is thought to be favored by a decrease meaning for vapor layer thicknesses <1 /i. In addition, the unstable
in the fuel or coolant temperature (or both), increase in relative ve- vapor-liquid interface generates random waves, which may be quite
locity of the two fluids, or by a sudden increase in pressure. However, large in amplitude compared to the small vapor layer thickness. Also,
the time scales for the pressure rise in explosive boiling are 0 (10 - 4 ) an unstable vapor dome forms at the top of the heated horizontal tube,
s, so that it is very difficult to obtain quantitative information in releasing vapor bubbles nearly periodically, while the vapor layer at
liquid-liquid systems. On the other hand, studies on the collapse of the bottom stagnation point is quite thin prior to the shock arrival.
Leidenfrost boiling of single drops of ordinary fluids, such as ethanol Hence the actual film collapse is expected to be asymmetric, which
or Freon, on molten Woods metal or solid metal surfaces due to a rapid is not taken into account in these calculations. Finally, physical
increase in pressure indicate little difference in behavior [1]. Pre- properties of the fluid (Freon 113) are either extrapolated from a low
sumably this is due to high interfacial tension of the liquid metal pressure, low temperature range (for vapor) or assumed constant (for
relative to either the vapor or liquid, which tends to inhibit liquid- liquid). Since the rise in gas pressure and temperature is considerable
liquid mixing in this configuration. It seemed likely, therefore, that (often to supercritical values), this might cause some error in the re-
useful information concerning the mechanics of triggering and sults.
propagation might be obtained by studying the transient heat transfer Since the characteristic lengths of the phenomenon are much
behavior of an horizontal electrically-heated nickel tube, initially in smaller than the heater diameter, curvature effects can be neglected
film boiling with respect to the surrounding subcooled liquid, upon and the problem is treated in rectangular coordinates. The principal
arrival of a moderate pressure shock wave. difficulty in solution arises from the fact that the moving boundary
Inoue and Bankoff [1] have reported experimental results with two is induced by external forces and not by heat transfer. This sets it
liquids over a range of initial pressures, final pressures, shock rise apart from other moving boundary heat conduction problems with
times and surface temperatures. This paper presents two theoretical change of phase and eliminates similarity solutions. Since it is prob-
treatments of the vapor film collapse process, which are compared ably impossible to get an analytical solution, numerical methods were
with the experimental data. used after simplifying or eliminating some of the equations. The
complete model is shown in Fig. 1 and the following assumptions were
1
Present address: Heat Transfer Research, Inc., Alhambra, Calif. made: (1) Metal temperature is uniform. (2) Molecular transport
2
This work was supported by the U.S. Department of Energy. layers exist at both ends of the vapor layer. (3) No heat storage in the
Contributed by the Heat Transfer Division for publication in the JOURNAL
OF HEAT TRANSFER. Manuscript received by the Heat Transfer Division molecular layers. (4) Vapor follows the ideal gas law. (5) Pressure
November 7,1980. inside the vapor film is uniform. (6) Volumetric heat release rate in

Journal of Heat Transfer AUGUST 1981, VOL. 103 7 46S

Copyright © 1981 by ASME
Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 and
-8»_i'
T„,(0) ' mo = given (3)
Considering the metal side molecular transport layer, the Knudsen
treatment shows that the heat flux, qm (t), at the wall interface region
is given by [2, 3, 4]:

2 7 - 1 \2 w Mgj y/Tei (£)

According to assumption (3) this also represents the heat flux at the
P« (t)
left-hand edge of the vapor conduction zone. The thickness of this
layer can be assumed to be one half of the mean free path of gas
molecules [5], so that

MTgtM.Pgit))- (5)
2y/2 sn
Fig. 1 Sketch of the model
The collision cross section, s (=7r<i2), does not change in the present
range of temperatures and pressures. The number density, n, however,
is a function of the gas temperature and pressure. Tg\(t) is taken to
metal is constant. (7) All viscous dissipation and gravity effects are be the average temperature in this layer:
negligible. (8) Constant liquid properties. (9) Velocity of vapor purely
Tgi(t)=h(Tm(t) + Tsl(t)) (6)
radial during collapse.
Assumption (1) has been justified by the analysis reported in Part The mass of this molecular transport layer, is given by:
I. Assumption (2) is required because (a) the influence of the Knudsen
M'
number becomes dominant when the vapor film gets thin, and (6) the M\i - M'gn \\ = -^-j~ = Mo = constant (7)
heat flux, which goes toward infinity when the vapor thickness ap- 2y/2s
proaches zero, can be limited by considering the molecular transport since the collision cross section remains constant during the transient
mechanism. The third assumption is justifiable in view of the large period.
heat flow and the negligibly small heat capacity involved. Assumption Using the ideal gas law in the definition of the mean free path, one
(4) is expected to hold since the gas is highly superheated. The fifth, finds
seventh and eighth assumptions are customary, and assumption (6)
reflects our experimental condition. Assumption (9) takes into ac- x m_Tgl(t)pgy»
A
l W ~ 77i ,„> „ , v A0 (8)
count the short collapse times. Later, we will impose additional re- Tgl(0)Pg(t)
strictions, as required to facilitate the solution. Thus, an energy bal- where Xo is half the mean free path at Pg(0), Tgi(0) and can be de-
ance on the metal tube gives: termined easily. X0 is about 300 to 400 A at N T P .
dTm(t) The mass and thickness of the liquid side molecular layer can be
(1) found similarly. Inside the vapor conduction region, the mass balance
dt and energy balance equations can be written, in view of assumption
where the steady-state heat flux is (7), as [5]:
PRt _d_
constant (2) + — (Pgug) (9)
irDLt dt dx

-Nomenclature-
A; = profile coefficient, equation (31) for i = Pp = final test section pressure 7 = Cp/C„
0,1,2 Pi = initial test section pressure T - latent heat
Bi = profile coefficient, equation (62) for i = AP = PF- Pi ^sat.sub = initial vapor film thickness for sat-
0,1,2 q = transient heat flux urated or subcooled boiling
C p = specific heat at constant pressure qp = peak heat flux, maximum of qm bm = thickness of heated tube
Co = specific heat at constant volume qss = steady state heat flux X = half of mean free path
D = diameter of heated tube
R = molar gas constant /i = viscosity
H = enthalpy
Rt = resistance of heated tube p = density
/ = current
s = collision cross section of gas molecules a = condensation or evaporation coeffi-
k = thermal conductivity t = time cient
£ = thickness of transient boundary layer in ta,c - time for first bounce, collapse ac,e = condensation, evaporation coeffi-
liquid Teo = steady-state liquid temperature pro- cient
L ( = length of heated tube file \p = Landau transformation variable, equa-
L°° = length of liquid column Te<° — initial liquid temperature far from the tion (23)
m = Lagrangian coordinate in vapor layer heater (or outer edge of boundary layer) TP = pressure rise time constant
rhc = condensation mass flux, molecular T'g = perturbation in Te, equation (53) X = X + Xi + X2, total vapor film thickness
me = evaporation mass flux, molecular
T m o = initial metal temperature
rht = rhc - me Subscripts
u = velocity
M = total mass of gas in conduction layer x = coordinate direction, Fig. 1 avg = average
M'g = molecular mass of Preon-113 X = vapor conduction layer thickness m = metal
Mg = molar mass of Preon-113 y — coordinate direction, Fig. 1 g = vapor
Afj.1,2 = mass of the Knudsen layer at left, 81, §2 — vapor properties at points shown in
a = thermal accommodation coefficient
right a = thermal diffusivity Fig.l
n = number density of molecules in gas £ = liquid
P = pressure 0 = initial value
P„ = pressure far from the heater ft s = saturation

466 / VOL. 103, AUGUST 1981 Transactions of the ASME

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 dTg In view of the number of boundary conditions available, a quadratic
PSCP,
dt
- + ue (10)
profile in \p is chosen for Tg [7]:
dx dx dx i
where Tg (iP, t) = A0(t) + AAt) f + A2(t) V (31)
pg = pg(x,t); ug = ug(x,t) (ID To satisfy the boundary conditions in Tg, this becomes:

Tg = Tg (x, t) (12) Tg(iP, t) = Tgl(t) + (Tg2(t) - Tgl(t) - A2{t)) \P + A2(t) iP* (32)
Using the Lagrangian coordinate system employed by Hamill and Substitution of the above profile into the integral energy equation and
Bankoff [6] a little simplification results in

m(x, t) = i pgdx (13) dTgl dTg2 ldA2. 1 dMI 4A2fe

Jo
• X(t)
dt dt 3 dt M dt \ -••41= M2
(33)

M{t)>
o
J"-A(tJ
PedX
(14) The boundary conditions on the conduction heat flux at the left and
right hand edges of the conduction zone become:
the continuity equation then becomes

dm qgi(t) = -^f(Ts2-Tg1-A2) (34)

— + pgUg = 0 (15) M

and the convective term can be eliminated from the energy equation, Qg2 (t) = (Tg2 - Tm + A2) (35)
M
which reads:
For the molecular transport layer on the right-hand side, the absolute
dTg *2Tg condensation mass flux, mc, and the absolute evaporation mass flux,
(16)
dt dm2 m e , are obtained from kinetic theory [8]:
where Me
mc = ocPg(t) (36)
2TCR Tg2(t)
8 g =^^ g (Pg) (17)
Me
Tg = Tg(m, t) (18) <rePs(t) (37)
2irRTs(t)
It is found that pgkg is nearly constant over a wide temperature range, In general, <rc = <re - a and the net condensation mass flux is
for a particular pressure, so that /3g is not a function of temperature.
The boundary conditions are: ^ / Pg Ps
rht = mc (38)
r
— Ts
Tg (0, t) = Tgl (t) (19)
and, for M(t) > 0,

1g\\t) = -Pgkg -— (20) d_M_

dm m=o mt = — (39)
dt
Tg (M, t) = Tg2(t) (21)
The net energy flux at the vapor-liquid interface, <?;(£), is
dT,
qg2 (t) = -Pgkg (22)
dm qi(t) = mtHg(Ts(t),Ps(t))+ mc(Hg(Tg2(t),Pg(t))
-Hg(Ts(t),Ps(t))) (40)
To eliminate the moving boundary condition at m = M(t) we make
use of the Landau transformation [6]: Then the heat flux to the bulk liquid, from an interfacial energy bal-
m (x, t) ance on the deposition layer, is
(23)
M(t) qe(t) + mtH£ ((T s (t),P s U)) = Qi(t) (41)
The energy equation then becomes:
where Hg(T, P) and H((T, P) are the vapor and liquid enthalpies,
dTg_±dMdTg= 1 &Tg and
(24)
g
dt M dt dyp M 2 d\p2 r = Hg(Ts,Ps)-HATs,Ps) (42)
where
since Pg (t) is not very different from Ps (t), so that Hg (T s , Ps) is nearly
(25) equal to Hg(Ts, Pg). Therefore,

The boundary conditions are: qt(t) = m ( r + mcCpg (Tg2(t) - Ts(t)) (43)

Te (0, t) = Tgl (t) (26) where T is the latent heat of evaporation at the saturation tempera-
ture. After substitution of the appropriate quantities, one obtains,
PgkgdTg
„ <t\- ^8"«"Jg (27)
Qgi (t) TTT7
M d\p if=0 Pg(t) P.(t)
e(t)
Tg (1, t) = Tg2(t) (28) [WTgWJ VTAtj,
PgkgdTg . Pg(t)
Qgi(t) ' (29) Cpe (Tg2(t) - T„(t)) (44)
M dip *=i y/TgW)
Upon integrating the energy equation from \p - 0 to ip = 1, one ob- On the other side of this molecular transport region, the energy bal
tains: ance becomes, including energy transfer by mass transfer:

mtHg(Tg2, Pg) + Qg2(t) - qdt)

e
(Jo *) M dt \ Jo ^ dip
.h.[*£l.
2 (30)
M \M *=i dip if=0; dt

Journal of Heat Transfer AUGUST 1981, VOL. 103 / 467

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 In the liquid the momentum equation is, assuming cylindrical com- 2k p
pression of liquid
qAt) = qSs + ~(Ts(t)-T,(0)). (66)
•C\t)

d2X«)x (dX(t)Y P,(0-P-(i) The Clausius-Clapeyron equation at the interface is

(46)
dt2 \ dt I p(b dTAt) RTsHt) dPAt)
where P » ( t ) = Pi + AP(1 - e x p ( - t/rp)) (67)
dt MgTPAt) dt
X(t) = X(t) + Xi(t) + X2(t) (47a) From the ideal gas law
C •x(0
1-ill-^, fe = - ^ n Y ; Y = LJD (476,c) Pg(t)dx=— j pg(x,t)Ts(x,t)dx (68)
D o Mg Jo
The energy equation, using a coordinate y which is fixed to the moving After integration and simplification, one finds
interface, becomes for short times, 1 R
dT£ d2T£ Ps(t)x(t) Mx (Tm(t) + Tgl(t) + TgAt)
(48) 2 Ma
ae 2
dt dy
+ TAt)) +M(t)(Te2(t) + Tsl(t)~-A2(t))\ (69)
with the initial and boundary conditions

TAy, 0) = Tea (y) (49)

Te(0,t) = TM (50) 3 Calculation

Two computer codes were constructed, based on the above model,
TAW), t) = T£0(£(t)) (51) the second being a simplified version. Both of them indicated the same
dT£ _ _, dTjo trends and showed the same order values in the peak heat flux and
-k. = Qss (52) peak gas pressure when a parameter survey was run to check the effect
dy y=e(t) dy y-nt)
of Tm0, Pi, AP and r.
where £(t) is the transient thermal boundary layer thickness in the The following equation was used from the analysis of Baumeister
liquid. Next, we put and Hamill [9] to determine an initial film thickness:
TAy, t) = Teo(y) + T'Ay, t) (53) ksixg (Tm0 - Ts0)U/4
X(0)-f exp 4.35 (70)
with T'D On as j

d2T£0(y) where,
'0 (54)
dy 2 V' = T 1 +
1 " ^pg W m Ts0)
(71)
and B.C. 20
This value of the initial film thickness, however, always came out to
T£0 (0) = T,(0) (55)
be larger than that calculated by the steady-state pure conduction
T£0(£o) = Te- (56) equation. The difference is attributed to turbulence and surface waves
which increase the effective gas thermal conductivity. A superficial
Substituting equation (53) into equation (48), one obtains gas thermal conductivity kg* was thus defined to satisfy the following
dT'Ay, t) &T'Ay, t) boundary conditions
ae (57)
dt dy 2 , , , Qss x(0)
at X £ X (0) (72)
with the initial condition: (Tmo ~ Tso)
kg* = kg a t X ^ Xcond (73)
T'Ay, o) = o (58)
and to produce a smooth fit, as a function of x between x(0) and Xcond-
and boundary conditions:
Here, Xcond represents the smaller film thickness calculated from the
TA0, t) = Ts(t) - T£o(0) = TAt) - TA0) (59) pure conduction equation:
T'A£(t), t) = 0 (60)
Xcond -
(Tmo ~ Tso) (74)
ZT'Ay, t) qss
-k£- =0 (61) and kg is the actual gas thermal conductivity which is a function of
y-m dy
.the average gas temperature and pressure. Thus, for Xcond < x ( ' ) <
Assuming a polynomial profile in the liquid [7] of the form
x(0) one obtains:
TAy, t) = B0lt) + Bx(t) f-f-) + B2(t) (-f-f (62) kg*(Tg,Pg) = kg(Tg,Ps)
X(t) ~ XcondlL£
and satisfying the boundary conditions, one obtains + (ke0-ke(Tg,Pg)) (75)
-
X(0) XcondJ
T'Ay, t) = (TAt)-TAO))\ (i-^f
£(t)l
(63) A value of 1.5 was chosen for the exponent after finding that signifi-
cant changes in qp did not occur if the exponent was taken as 1.0 or
Applying the integral method to equation (57), the energy equation 2.0. In our experimental condition, a and 6 in equation (49 b,c) are
becomes 500 and 0.003, respectively.
(a) Simplified Model. If one neglects heat storage in the gas film
- - H C ^ t ) - TA0)) £(t)\ = ae-£-(TAt) - TS(0)) (64) by setting f$g (= pgkg/Cpg) -* <», one obtains A2(t) = 0, which means
A at £(t) that the temperature profile in the gas is linear in i/'. This simplifies
and the equation set considerably.
Secondly, the molecular transport layer on the gas-liquid interface
jdTAy, t)\ \dT'e(y, t)\
qAt) -ke\ =qSs-ke\ (65) was suppressed and a simple heat balance was used to determine the
)y-o \ <>y )y=o rate of condensation or evaporation.
\ dy
or using equation (63) Two different methods were used to determine the heat flux in the

468 / VOL. 103, AUGUST 1981 Transactions of the ASME

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 liquid, in addition to the one reported above. One follows the Du-
hamel's integral approach, while the other is based on a general profile
method, as shown in reference [10]. All three gave nearly the same
results, and the last one was chosen for calculations, being the
quickest. Thus, the final set of equations is:

' Pm^pmOn

d2XA [dx^ = 500, b = 0.003 (78)

—2r + a — a
dt \dt
dTs {2TMg*Pgx\
ft!
, dX
0 (79)
dt R2TS dt R J dt ' ' 0.10
Tgi is found from equation (766), which is an intermediate balance «-q/!'/| 2 Q.W

of equations (76), the step-by-step heat balance for the whole system.
'!1 !i i!>! i
0,30
4 0.33
Equation (77) represents heat transfer on the liquid side, while ii ! h
t 1 I!
equation (78) governs the motion of the liquid under the pressure ! ii j
gradient. Equation (79) is the combined form of the ideal gas law and 1! i
the Clausius-Clapeyron equation, after differentiation. This system -.„.l._.U!..i..L........
of five nonlinear first-order ordinary differential equations was solved o.i • I . •
numerically for Tm, Ts, M, qe and x as a function of time explic- i O.I 0.3
itly. T. ms
Fig. 2 Simplified model results. Time history of q,P and x 'Of changing AP.
(b) Complete Model. Since the Knudsen layer equations are al- Tm0 = 300°C, P, = 0.1 MPa, T p f = 80 jts, T/ = 22°C. <jM = 0.16
gebraic in form they were converted to differential equations by ap- Mw/m 2
propriate differentiation, where necessary. Finally, a system of ten
non-linear ordinary differential equations are obtained, which can cient, a, on the peak heat flux, qp, the minimum film thickness, XP<
be solved numerically for Tm, Tgi, Tg2, T„, A2, Pg, PS,M,£ and x as and the peak gas pressure, Pp. The results are quite independent of
a function of time. The Runge-Kutta-Gill method with double pre- a for AP = 0.2 MPa but there is a significant change if AP = 0.3
cision was used in the numerical calculation. The initial time step was MPa.
10~ 5 s but the step size decreased automatically, depending on the Measured values of a for Freon-113 have not been reported.
magnitude of the time derivative x and x itself in the region of rapid However, some early experiments have shown that u becomes very
change. Final forms are rather lengthy and are given in reference low for antisymmetric molecules. Heideger and Boudart [10] found
[12]. a to be 0.05 for glycerol while Delaney, et al. [11] reported a value of
~0.03 for water. Assuming Freon molecules to be antisymmetric one
4 Results might therefore expect <x to be around 0.05. If these molecules were
(a) Simplified Model. A parametric survey was run to check the symmetric, <x would be higher. However, the noncondensible gases
effect of TmQ, Pi, AP and r on the behavior of the gas film after pas- present in the system would lower its value considerably, by adding
sage of the shock. The time history of the heat flux, film thickness and to the interfacial resistance. Very minute quantities of noncondens-
gas pressure were calculated for each parameter. The results are quite ibles are required for a large reduction in a, as they get concentrated
sensitive to the initial film thickness, which is actually a stochastic while the vapor film is condensing. Therefore, a = 0.05 was chosen
variable, due to surface waviness and bubble breakoff. For these for further computations to check the effects of AP, Pi, T and Tmo-
calculations, the stopping criterion was arbitrarily set at x ^ 0.1 n, Figure 4 shows the time history of the variables for changing AP, and
corresponding to the estimated surface roughness, which might play can be compared directly with Fig. 2, since identical conditions were
an important role during an actual vapor-liquid contact on micro- used for both. The trends are quite similar, except that the gas film
scale. However, the film always bounced back before reaching this apparently collapsed for AP = 0.35 MPa, and no bouncing was de-
cut-off value and computations were continued until the second re- tected for AP = 0.1 MPa, within about 0.25 ms. Figure 5 is a plot of
bound occurred. In these cases, the left hand-side molecular transport these peak values against AP. The dashed and the solid lines show
layer had practically no effect and Tgi was always within 1°C of the results of Figs. 2 and 4, respectively. The experimental ranges of
qp are shown by vertical lines, and are found to be in the same range.
Tm.
For shocks with AP > 4 atm., very high heat fluxes with supercritical
Figure 2 shows the effect of shock pressure. Bounce times varied
gas pressures and partial contacts are predicted, according to Fig.
from 0.1 to 0.2 ms for AP values of 0.1-0.35 MPa. The heat fluxes
5.
increased by an order of magnitude or more, but stayed high over a
rather short period only. The gas pressure always remained below Figure 6 shows the effect of P/. These calculations indicate that the
critical, the value of which is 3.37 MPa for Freon 113. The trends gas film would collapse for P / < 0.1 MPa, producing rather high heat
shown are as expected. fluxes and gas pressures exceeding the critical. This is also in agree-
(b) Complete Model. In this case the parametric survey in- ment with Fig. 5, since a high value of AP = 0.34 MPa was used. The
volved two additional parameters, i.e. a and a. It was found that bounce or collapse time of ~0.13 ms was independent of P/. The ex-
changes in a from 1.0 to 0.05 were of negligible effect, so that a was perimental qp values agree with theory for Pj = 0.1 MPa. They are
chosen to be 1.0 always. The calculations are continued up to x ^ 0.1 much lower for Pj = 0.057 MPa, which is not unexpected. A sharp
ix, but unlike the previous case, complete collapse was predicted for peak whose half width value is less than 10 /us cannot be detected by
a few cases. the experimental measurement circuit and the measured peak is
Figure 3 shows the effect of the condensation-evaporation coeffi- therefore greatly attenuated.

Journal of Meat Transfer AUGUST 1981, VOL. 103 / 469

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 MPa MPa
MW
10 r — T 1 r— 1 I03
10 10'2 ^

Pp-

Mil
•
10 I
^r^-
I 10
"PT5—
5"

• %
. - ~ ^ ^ ^ - "- X r
—

I 0.1 "~\/^T -10

0.1 I
S

0.1 0.01 -

-
— P "0.299 Mpa
— P"0.20 Mpa

0.1
0.01. 0.1
03 0.36
Fig. 3 Complete model results. Effect of condensation-evaporation coeffi- MPa
cient. qp, P and %p versus a. Tma = 300°C, P, = 0.1 MPa, T p = 80 pis, 7> Fig. 5 Comparison of results of Figs. 2 and 4. qp, Pp and Xp versus A P for
= 22°C, qss = 0.16 Mw/m 2 and A P = 0.299 and 0.22 MPa both models. Same condition as Figs. 2 and 4

M P o ^
IO io* io*
MPa MW

10 Iff I0 3

r
p
i io

I IO
^p

O.I I

0.I I

0.I

0.I

V UP'O.I MPa
.0.2
=0.3
C.S.'COMPUTATIONS _ | »0.35
0.05
STOP AT 8.
NEXT STEP O 0.I
0 Q08 Ql 0.I5 f 0.2 0.2S Fig. 6 Comparison of qp, Pp and XP versus P, for simplified and complete
ms models. Same conditions as Fig. 2 except for changing P(. Tm0 = 300°C, AP
Fig. 4 Same as Fig. 2, except for complete model with o~ = 0.05 = 0.341 MPa, Tp ± 80 pis, Jt = 22°C and 0^ = 0.16 Mw/m 2

470 / VOL. 103, AUGUST 1981 Transactions of the ASME

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 MPo - M *
0 * /

/ o EXPERIMENT

o o o o

<° >'
" X { 0 ) BY BAUMEISTER ,-•*» O &<
CORRELATION

Fig. 8 c/p versus Tmg for the complete model and two values of the Initial film
thickness. Also compared with experimental data points. P, = 0.1 MPa, A P
= 0.299 MPa and 0.32 MPa, T p = 80 fis, 7> = 22°C and a = 0.05

the simplified model predicts a "bounce" (minimum vapor film

thickness), and where experimentally it is known that collapse takes
place.
2 The experimental data scatter is most probably due to random
variations in film thickness at the instant of shock arrival, as seen from
an analytical sensitivity study. There is also an experimental averaging
of the peak heat flux data due to instrument limitations.
Fig. 7 Comparison of q„ and Pp and %p versus r p for both models. Same 3 Effective direct "contact" between solid and liquid, with only
conditions as Fig. 2 except for changing TP, Tm0 = 300°C, P( = 0.1 MPa, A P a thin intervening layer of supercritical fluid, is attainable by shocks
= 0.299 MPa, r> = 22°C and fe = 0.16 Mw/m 2
with rise times of order 0.1 ms-1 ms and pressure ratios less than 10.
Collapse times of the order of 0.1 ms can be expected, which is ade-
Figure 7 shows the effect of the pressure rise time constant, r. It quate for efficient fuel-coolant interactions.
shows good agreement with experimental qp values at r = 80 and 120 4 When the thickness of the vapor layer becomes less than lfi, the
vapor pressure increases to more than 10 MPa and reaches supercri-
The solid lines in Fig. 8 show two plots of qp versus the initial metal tical pressures under some conditions. The transient local pressure
temperature, Tmo, in the sensitive range for the film collapse or in the thin vapor layer might cause fragmentation and mixing of fluids
bouncing of AP = 2.99 and 0.32 MPa, respectively. However, when near the interface.
the initial vapor film thickness is chosen to be about twice that cal-
culated by the Baumeister and Hamill correlation [9]. (Since the References
correlation gives an initial film thickness which is about half the av- 1 Inoue, A., and Bankoff, S. G., "Destabilization of Film Boiling Due to
erage thickness calculated from the initial heat flux and metal tem- Arrival of a Pressure Shock. Part I: Experimental," Report No. COO-2512-13,
perature, the computational results, shown by the dotted lines are Chemical Engineering Dept., Northwestern University, Evanston, 111, 1978.
quite different. This illustrates the powerful effect of the initial film 2 Loeb, L. B., The Kinetic Theory of Gases, McGraw Hill, New York,
1927.
thickness on the tendency for the vapor film thickness to collapse 3 Patterson, G. N., Molecular Flow of Gases, John Wiley and Sons, New
below about 0.08 ft, corresponding to the Knudsen layer thickness (cf. York, 1956.
Fig. 4).) 4 Alofs, D. J., Flagan, R. C , and Springer, G. S., "Density Distribution
Measurements in Rarefied Gases Contained between Parallel Plates at High
The experimental points are generally bracketed by the two sets
Temperature Differences," Physics of Fluids, Vol. 14,1971, pp. 529-533.
of curves. It appears therefore that by a reasonable adjustment of the 5 Bird, R. B., Stewart, W. E., and Lightfoot, E. N., Transport Phenomena,
assumed initial vapor film thickness the calculated qp versus T m o John Wiley and Sons, New York, 1960.
curves can be made to follow the data trends, but that the behavior 6 Hamill, T. D. and Bankoff, S. G., "Growth of a Vapour Film at a Rapidly
is quite sensitive. Further, it is recognized that the initial vapor film Heated Plane Surface," Chemical Engineering Science, Vol. 18, 1963, pp.
355-364.
thickness, and hence collapse, is far from uniform from top to 7 Goodman, T, R., "The Heat Balance Integral and its Application to
bottom. Problems Involving a Change of Phase," Trans. ASME, Vol. 80, 1958, pp.
335-342.
6 Conclusions 8 Collier, J. G., Convective Boiling and Condensation, McGraw Hill, New
York, 1972.
1 Two models for destabilization of film boiling around a heated 9 Baumeister, K. J., and Hamill, T. D., "Laminar Flow Analysis of Film
cylinder by a sudden increase in ambient pressure are presented. The Boiling from a Horizontal Wire," NASA T N D-4035, July 1967, p. 44.
10 Heideger, W. J., and Boudart, M., "Interfacial Resistance to Evapora-
more complete model takes into account heat transfer in all three
tion", Chemical Engineering Science, Vol. 17,1962, pp. 1-10.
phases—solid, vapor and liquid—as well as the Knudsen layers on 11 Delaney, L. J., Houston, R. W. and Eagleton, L. C , "The Rate of Va-
the liquid and solid surfaces. The simplified model neglects heat porization of Water and Ice," Chemical Engineering Science, Vol. 19, 1964,
storage in the vapor film and the Knudsen layer at the liquid surface. pp. 105-114.
Both models predict peak fluxes in general agreement with those 12 Inoue, A., Ganguli, A., and Bankoff, S. G., "Destabilization of Film
Boiling Due to Arrival of a Pressure Shock Part II: Analytical," Rpt. No.
found experimentally, but the complete model is considered more COO-1512-15, Chemical Engineering Dept., Northwestern University, Evanston,
reliable, since it predicts vapor film collapse under conditions where III, 1978.

Journal of Heat Transfer AUGUST 1981, VOL. 103 / 471

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 An Experimental Inwestigation of
N. Seki
Professor.
Boiling Heat Transfer of
S. Fukusako Fluorocarbon R-11 Refrigerant for
Associate Professor.

K. Koguchi
Concentric-Tube Thermosyphon
Graduate Student.
The characteristics of the boiling heat transfer for a concentric-tube open thermosyphon
are examined in detail. Fluorocarbon R-ll refrigerant as a testing fluid is utilized. Out
Department of Mechanical Engineering,
Hokkaido University, of a number of possible controlling variables, the effects of the heat flux, the void ratio,
Sapporo 060, Japan and the diametric ratio of the concentric-tube on the heat-transfer performance are deter-
mined. From the present experimental results it is demonstrated that total heat flux can
be appropriately predicted by a superposition of the heat flux due to the available corre-
lations for free convection and due to the correlations experimentally determined for boil-
ing heat flux.

Introduction
Recent recognition of the finite nature of fossil fuels such as natural vestigated analytically the general problem of condensing thermo-
gas and oil has led inevitably to an urgent call for the development syphon in a variable acceleration field, and concluded that the effects
of alternative energy resources. of temperature difference and the Prandtl number are similar to those
Prom a practical point of view, the potential for utilization of geo- in other condensing problems. The possible application of heat
thermal energy has extensively been described in several references, transfer by evaporative cooling of a liquid metal under high centrifugal
for instance, Hayakawa [1]. The major use of the geothermal energy acceleration to turbine blade cooling was demonstrated by Genot and
up to now has been power produced from steam-dominated or from LeGrives [6]. A study of a two-phase thermosyphon with large filling
liquid-dominated reservoirs. The great potential of geothermal en- quantity was done by Larkin [7]. In his study it was pointed out that
ergy, which is probably to be found in locations of normal geothermal it was not possible to develop a precise relationship for boiling and
gradient, would be expected to be devoted, for instance, to melt snow condensing heat-transfer coefficients, but an appreciable knowledge
on roads in the northern cold districts. of the process was gained. A recent study of the two-phase thermo-
A number of important problems, however, remain to be resolved syphon was carried out by Lee and Mital [8]. They performed an ex-
before utilization of such energy in very low density becomes feasible. perimental study using water and fluorocarbon R - l l refrigerant as
Especially, the means of energy transport to the surface is of prime testing fluids, and evaluated the effects of many possible controlling
importance in economical and technical evaluation of the heat-re- variables on the heat-transfer characteristics of a two-phase ther-
moval from such a very low-density geothermal energy. mosyphon. McDonald, et al. [23] investigated the characteristic op-
A two-phase open thermosyphon, which utilizes buoyancy force erating behavior of a two-phase thermosyphon loop which they are
and phase change of liquid contained in a vertically long tube, may proposing as an efficient waste heat recovery system and found that
be considered to be one of the convenient devices to achieve this, the optimum performance is mainly concerned with the orientations
because a power supply and a pump to circulate the working fluid are of the evaporator and condenser tubes. A simulation of the behavior
not required, and, furthermore, the low cost and simplicity of con- of the thermosyphon loop was made by Ali and McDonald [24], who
struction of the device are quite advantageous. It is well known that indicated that the agreement between the simulated performance and
the high performance of the two-phase heat-transfer system is at- experimental data is very good. McDonald and Ali [25] also presented
tributed to the fact that it utilizes mainly the heat-transfer charac- the results of a number of loop simulations which were carried out to
teristics of phase change which are more efficient means of heat illustrate typical behavior of thermosyphon loops as various param-
transfer than those of single-phase. eters are changed.
In addition to the aforementioned expectation, it was recently Experiments of the heat-transfer performance and the critical heat
shown by Seki, et al. [2] that the markedly decreased heat-transfer flux for a single-tube open thermosyphon using water, ethyl-alcohol,
performance under impeded or similar flow conditions can signifi- normalhexane, and carbontetrachloride as testing fluids were con-
cantly be improved by installing an additional inner tube which could ducted by Kusuda and Imura [9, 10]. They claimed that the heat
be occupied by the downward return-flow in this concentric-tube open transfer for their thermosyphon is greater than that of pool-boiling,
thermosyphon. while the critical heat flux has a tendency to become small, and that
There have been few investigations on the two-phase thermosyphon the heat-transfer coefficient is nearly independent of the tube length
other than those reported in the open literature [3-10] and thus it and tube diameter, and also of the flow-pattern characteristic.
appears that the fundamental design information and operating From the foregoing literature survey, it is apparent that very little
technique of the system may be not satisfactory. Cohen and Bayley is known about the boiling heat-transfer characteristics of the con-
[3] conducted a series of experiments using hollow thermosyphon centric-tube open thermosyphon of the present system, and especially
tubes in a rotating jig which were filled with water in quantities about the fundamental informations which are essential to apply the
ranging from 0 to 100 percent of the tube volume. They reported that system in extraction of thermal energy from deep in the earth and in
over a wide range the heat transferred from the heated section to the bringing it to the surface thermal plant.
cooled section of the thermosyphon is independent of the quantity The objective of the present paper is, therefore, to make an ex-
of coolant enclosed. Long [4] tried to use the closed two-phase ther- perimental study into the heat-transfer performance of a stationary
mosyphon as a thermal rectifier to keep permafrost. Chato [5] in- open thermosyphon having a concentric annulus, so-called the present
concentric-tube open thermosyphon. In particular, the effects of the
Contributed by the Heat Transfer Division for publication in the JOURNAL
OP HEAT TRANSFER. Manuscript received by the Heat Transfer Division July heat flux, the void ratio, and the diametric ratio of concentric tube
3,1980. were extensively investigated.

472 / VOL. 103, AUGUST 1981 Transactions of the ASME

Copyright © 1981 by ASME
Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 Experimental Apparatus and Procedures
A schematic diagram of the experimental set up is depicted in Fig.
1. The main heat-transfer loop of the concentric-tube open thermo-
syphon consists of a vertical concentric cyclinder as test section where
both the top and bottom of inner tube are opened, a main electrical
resistance heater assembly, a guard heater, a reservoir, and associated © TEST SECTION
© RESERVOIR
instrumentations. ® MAIN HEATER
The copper outer tube of 100.0 mm i.d. and 2 mm thickness with © GUARD HEAER
© WATER GUAGE
2000 mm length was utilized. Values of the annular-space width © BRINE INLET
ranging from 5.7 to 34.0 mm were employed by using vinyl chloride © BRINE OUTLET
© AIR PURGE
tubes with diameters of 32.0, 48.0,60.0, 66.6, 76.0, and 88.6 mm with VALVE
2000 mm length as the inner tube. Between the copper outer tube and © 0RAIN
® THERMOCOUPLE
the cooling reservoir a lucite tube 100.0 mm in length was installed PROBE
for visual observation of boiling behavior in the annulus. Closure of © VISUALIZATION
SECTION
the outer tube was performed by screwing a threaded cap into the
bottom end, which also carried a brass tube to a drain valve.
The surface temperature of the outer tube was uniformly main-
Fig. 1 Schematic diagram of experimental set up
tained at a constant temperature by using nine independent and
controllable main heaters, which were respectively surrounded by
corresponding guard heaters to eliminate the heat loss from the main desired wall-temperature. Continuous thermocouple readings were
heaters to the environment. These heaters were made of electric re- made on the temperatures in the total system.
sistance rod (chromel-iron wire 1.6 mm in diameter) wound around In a series of experiments, a fresh fluorocarbon R - l l refrigerant
the outer tube surface. All heaters were supplied with electric power packed in 20 kg vessel, which has been preserved in the tempera-
through separate circuits from a variable transformer rated 10 KVA, ture-controlled room at 15° C, was always charged to the thermosy-
100 V, respectively. The heat from the test section was removed at the phon system.
reservoir through a heat exchanger consisting of coiled copper tubing
and a temperature-controlled coolant, Results and Discussions
Measurements of wall-surface temperature were obtained from In all, 104 experimental runs were conducted for the various tem-
thermocouples carefully plugged into the tube wall at stations cor- peratures of heated outer tube. Measurements were made as the
responding to the center plane of each heating coil. In addition, a long temperatures of heated outer tube varied from 15.6 to 52.5 °C. All
slender shielded probe with thermocouple was axially traversed along physical properties were based on the arithmetic mean temperature
the axis of the tube to measure the core temperature of the working of the heated outer tube and that of the entrance-fluid into inner tube.
fluid in the inner tube. Outputs of these thermocouples were con- A more complete documentation of the data is included in [26].
tinuously recorded on YOKOGAWA Multiplex Recorders. Power to Visual Observation. Qualitative visual observations were con-
the main heaters subdivided were measured by means of ammeters ducted to examine the effect of the presence of inner tube on the
and voltmeters provided for each heating coil. boiling phenomena which take place near the open end of the sin-
The pressure inside the thermosyphon was measured at the res- gle-tube open thermosyphon or the concentric-tube one.
ervoir by a pressure gauge. Measurements of the mean void ratio, Flow visualization for the boiling phenomena of the single-tube
which is defined as the ratio of the vapor space to the total volume of open thermosyphon and the vertical narrow annular channel has been
the annulus, were undertaken by using a Hg-manometer measuring studied extensively by previous investigators [9-11]. The present
the pressure difference between the bottom and open end of the observations were mainly focused on the flow-exit phenomena with
concentric-tube thermosyphon. In the procedure mentioned above, boiling near the open end of the concentric-tube open thermosy-
it was assumed that the head of the fluid friction in the rising two- phon.
phase flow could be neglected comparing with the gravity head of the For a single-tube open thermosyphon using fluorocarbon R - l l re-
two-phase flow (maximum ratio of the fluid-friction head to the frigerant as a working fluid, it was assessed that the present results
gravity one was approximately evaluated to be less than 1.0 per- observed may be approximately similar to those of Kusuda and Imura
cent). [9] using pure water as working fluid. It was observed that bubbles
Before starting actual experiments, the thermosyphon system was generated along the heated tube rise without coalescence and carried
thoroughly cleaned by repeated washing with acetone followed by back some of the cool fluid (flow downward), as schematically shown
water. After the system had been perfectly dried, the charge valve was in Fig. 2(a). Kusuda and Imura [9], furthermore, claimed that onset
opened, and the flow of the working fluid was initiated. Special care of bubble coalescence and accompanying pulsation throughout the
was taken to insure that there were no air bubbles within the con- whole channel occurs as the heat flux is increased. Such a phenome-
centric tube. Power then was supplied to both the main and guard non, however, was not observed within the range of present experi-
heaters. The power to main heaters was increased in steps up to a mental parameters examined. This is mainly attributed to both the

-Nomenclature.
a = inner radius of heated tube, D/2 9cal = calculated heat flux based on equations Tw = temperature of heated surface
d, = inner diameter of inner tube (2) or (3), (4) + equations (5), (6), respec- AT = temperature difference between
d0 = outer diameter of inner tube tively heated wall and entrance-fluid, Tw — Te
D = inner diameter of heated tube r = equivalent heat-transfer radius, (D2 — ATS = temperature difference between
g = gravitational acceleration do2)/(2D) heated wall and saturated fluid at open end
L = tube length of thermosyphon ta - dimensionless temperature based on of thermosyphon, Tw — Ts
p = dimensionless pressure normalized by inner radius, gflATa3/(vK) a;, = boiling heat-transfer coefficient, qb/(Tw
atmospheric pressure tr = dimensionless temperature based -T.)
q = total heat flux from heated tube on equivalent heat-transfer radius, j8 = coefficient of volumetric expansion
qb = heat flux due to boiling heat transfer g/3ATrs/(i>K) K = thermal diffusivity
qex = measured heat flux Te = temperature of entrance-fluid at open X = thermal conductivity
qc = heat flux due to free convective heat end of thermosyphon H = viscosity
transfer Ts - temperature of saturated fluid v = kinematic viscosity

Journal of Heat Transfer AUGUST 1981, VOL. 103 / 473

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 OUTER TUBE INNER TUBE i 1 ' " "1 "
TW=4B.8°C
CONCENTRIC TUBE
D = 100.0mm O ANNULAR
45 d0= 60.0 mm s INNER

40 ----
CONCENTRIC TUBE
~~~~-
(b)
Fig. 2 Schematic drawing of flow pattern near open end of thermosyphon o ~ o - _. INLET
o o- - - . K21m
35 o
differences of the working fluid and the dimension of the heated tube 8 9 9 0 0 e
(Kusuda and Imura [9] used a markedly small channel of the rec- 1 1 1 I i
tangular section of 300 mm in length and 50 mm in width). 2.0
On the other hand, as will be appreciated by an inspection of Fig.
1, for the concentric-tube open thermosyphon the primary effect of Fig. 3 Distributions of core-fluid temperature for concentric-tube thermo-
heating the outer tube should be to cause flow upward along the wall syphon
of the tube due to the buoyancy effect and an associated return flow " r- - r " " r "' ! •
downward in the inner tube via continuity. It was, thus, observed that
the cool fluid without bubbles in the inner tube comes down to the Tm=/.1.5°C
bottom while the hot fluid with a number of bubbles due to evapo- /
ration rises up the annulus like the pure free convection in the con- SINGLE TUBE -
centric-tube open thermosyphon. It was, furthermore, found that with D = 100.0mm
small do (large annular space) some of the cool fluid (flow downward)
tends to enter into the annulus, hence a serious interaction between --P
~-~.^0
the ascending flow with bubbles and the descending cool fluid occurs
~~ *~o^ /T 5
near the open end of the annulus, as will be shown in Fig. 2(6). It was
also observed that for large do (small annular space) a three-dimen-
sional flow with bubbles occurs near the place where the tube meets ~"0~---
"" < X --._ 1 INLET
the reservoir. - - ^ r 2.1m —
O IT-
The effects of annular-space width on the boiling heat transfer in i
a vertical annular channel was extensively studied by Ishibashi and 1 1 1 1 i
Nishikawa [11]. They showed that the boiling phenomena without 1.5
any coalescence of bubbles when the annular space is greater than 5
mm in width are closely similar to those of pool boiling along the Fig. 4 Distribution of core-fluid temperature for single-tube thermosy-
phon
vertical surface, while for annular space smaller than 5 mm in width
the bubbles tend to coalesce and continue to grow further to large size the inner tube into the annulus, thus indicating that the heat trans-
and are accompanied by pulsation at low frequency as the annular ferred between upward and downward flows is unexpectedly quite
space becomes smaller than 2.7 mm in width. small. It should be, thus, noted that for the concentric-tube open
Therefore, the fact that the coalesced bubbles could not be obtained thermosyphon there might be two different flow regimes with and
in the range of present experimental parameters examined (annular without boiling in working fluid even if the working fluid at the en-
space a 5.7 mm in width) appears to be well consistent with results trance section of the inner tube is completely saturated.
obtained by Ishibashi and Nishikawa [11]. Figure 4 shows a typical temperature distribution of the working
Temperature Distributions. Typical temperature distributions fluid along the core of a single-tube open thermosyphon without an
in the inner tube (flow downward) and in the annulus (flow upward) inner tube. The difference between temperature distributions of Figs.
for the concentric-tube open thermosyphon are shown in Fig. 3, where 3 and 4 is quite evident. The temperature distribution mentioned
do denotes the outer diameter of the inner tube. The ordinate is the above is closely in accordance with that of the saturation pressure-
fluid temperature of the core in the inner tube or in the annulus, while temperature relationship, which reveals that the boiling within the
the abscissa is the actual distance from the bottom of the thermosy- working fluid may take place throughout the thermosyphon tube.
phon. The dotted line in the figure indicates the saturation-temper- Such a trend as shown in Fig. 4 was also reported by Kusuda and
ature distribution of the working fluid which is evaluated by taking Imura [9] who conducted an experimental study on a single-tube open
the local static pressure into account. thermosyphon under the condition of uniform heat flux.
It is quite natural that the heating of the outer tube will cause a flow Boiling Heat-Transfer Curve. As pointed out earlier, it appears
upward in the annulus due to buoyancy and an associated return flow that the process of the boiling with free convection in a vertical an-
downward in the inner tube via continuity, as pointed out earlier. The nulus heated from outside may be still more complicated than that
temperature of the fluid in the annulus is, thus, expected to increase of common boiling in a pool, because the vapor-bubbles cannot rise
as the fluid flows up along the outer tube. It is, however, clear that the freely from a submerged heater, but are constrained to rise in a rela-
temperature of the fluid in the annulus increases initially with an tively narrow space, becoming more and more crowded and displacing
increase in axial distance from the bottom, goes through a maximum more liquid. It is firstly, however, natural to evaluate the character-
at a certain distance from the bottom, and then decreases with further istics of boiling heat transfer for the present thermosyphon in the form
increase in L. This phenomenon may be physically understood by of conventional boiling curve. In the present study, to more clearly
considering that after the temperature of the fluid in the annulus investigate the transition regime from the free convection to the de-
becomes identical to the saturation temperature at a certain distance veloped boiling, AT may be firstly defined as the temperature dif-
from the bottom, boiling within the working fluid becomes possible ference between the heated surface temperature and the inlet fluid
in the annulus, depending upon the gradient of the saturation pres- temperature at the open end of the thermosyphon.
sure-temperature relationship of the fluid. Figure 5 shows the heat transfer rate q as a function of the tem-
On the other hand, the temperature of the fluid in the inner tube perature difference AT. As will be seen in Fig. 5, the q increases at a
appears to be quite uniform until the cool fluid turns upward from moderate rate with AT, while at approximately AT = 13°C and AT

474 / VOL. 103, AUGUST 1981 Transactions of the ASME

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 1
-
C0NCENT1C TUBE
CONCENTRIC TUBE
// Eq.(5) "
d 0 mm D mm
d0mm Dmm
o 32.0-]
o 32.0-1 i° r
0 48-0
o 66.6 MOO.O
o 600
• 88.6-1 -100.0
c 2 - o 666
SINGLE TUBE off o 760
o 100-0

a*
1?
E
103^
1,03
-
• 88.6
SINGLE
o
J

TUBE
1000
if "
.«- -
Eq.(6)' ?7»»
13* -
e»
o
o /

K
AT °C sT s °C
Fig. 5 Boiling curves showing departure from free convection heat Fig. 6 <fc versus ATS
transfer

= 18°C for the concentric-tube and the single-tube open thermosy- qc = 0.0158- t a °- 4 4 3 AT (8 X 106 £ ta) (4)
phons, respectively, the rate of the increase becomes to a great extent a
larger. It was reported by Seki, et al. [12] that for the concentric-tube where r denotes the equivalent heat-transfer radius, tr the dimen-
thermosyphon of closed system, bubbles take place only on a few se- sionless parameter based on r, a the inner radius of the heated tube,
lected spots of the heating surface in the range of smaller AT than and ta the dimensionless parameter based on o, respectively, as will
about 10°C. It is, thus, expected that the smaller number of bubbles be defined in Nomenclature. The combination of equations (1) and
may have a minor contribution on heat transfer and that the heat- (2-4) results in qt,, representing that portion of the heat-transfer rate
transfer rate may be mainly governed by free convection in the an- which is only due to bubble-motion effects.
nulus. However, the number of spots on which bubbles start becomes The resulting graph of qt, versus AT S is shown in Fig. 6. In the
larger with an increase in AT, and at a temperature difference beyond present study, ATS may be defined as the temperature difference
about AT = 13°C heating surface is so densely populated with bubbles between the heated surface temperature Tw and saturation temper-
that their separation may cause an appreciable stirring action in the ature T s at the open end of the thermosyphon under the prescribed
ascending fluid which gives rise to an increase of the heat transfer. pressure of the reservoir. As will be shown in Fig. 6, it is elucidated
This might be considered to be the reason for steeper increase of the that the points representing <j(,-versus-ATs could be correlated ap-
q in the region AT > 13°C. proximately by a straight line on a log-log plot for both the single-tube
Of interest is the fact that for the single-tube open thermosyphon and the concentric-tube open thermosyphons, respectively. Fur-
after the so-called inception of developed boiling, AT decreases thermore, it can be seen that the heat-transfer rate for the single-tube
rapidly with an increase in q, yielding a marked neck of boiling curve, open thermosyphon is greater than that for the concentric-tube open
as will be clearly seen in Fig. 5. Such a great overshooting of AT at near one in the whole range of parameter studied, yielding a clear contrast
the inception of developed boiling appears to be caused by the rapid to the tendency in the free-convection region (see equations (2-4)).
transition from the subcooled boiling to saturated pool boiling due This characteristic may be understood by the fact that for the
to the more serious stirring-action of the bubbles which may take place single-tube open thermosyphon the boiling within the working fluid
more successively with an increase in heat flux. It should be noted that would be taking place axially throughout the whole tube because the
in the range of the developed boiling the temperature of entrance-fluid temperature of fluid is closely in accordance with that of the satura-
at the open end of the thermosyphon-tube was nearly maintained at tion pressure-temperature relationship, while for the concentric-tube
the saturated one under the prescribed pressure in the reservoir. For open thermosyphon there exist two kinds of flow regimes with and
the concentric-tube open thermosyphon, on the other hand, this trend without boiling even if the working fluid at the open end of the ther-
was not observed because subcooled fluid turns upward from the inner mosyphon is completely saturated, namely, developed boiling might
tube into the annulus at the bottom of the concentric-tube even if the be taking place only on an upper portion of the heating surface, as
working fluid at the entrance section of the inner tube is completely pointed out earlier in Fig. 3.
saturated, as will be suggested from the saturation-temperature re- It is elucidated from the Fig. 6 that the present experimental results
lationship in Fig. 3. using fluorocarbon R - l l refrigerant as a working fluid can be corre-
Correlation of Boiling H e a t T r a n s f e r . The present experi- lated by following equations in the range of parameters investi-
mental results were firstly examined to be correlated using a similar gated.
method to that adopted by previous investigators [13-16] for corre- For the concentric-tube open thermosyphon
lating forced-convection boiling heat-transfer data. It was assumed
that in the case of boiling with free convection the convection effects qb = 3.85 X 10" 3 (AT S ) 6 (5)
can be superimposed on the bubble motion effects, and the total heat For the single-tube open thermosyphon
flux thus is
qb = 1.21 X lO-HAT,) 6 (6)
q = qb + qc (1)
As elucidated above, it is thus evident that AT S is also an important
The free covective heat-transfer rate was calculated from the exper-
variable controlling the nature and the rate of the heat-transfer pro-
imentally determined values by Seki, et al. [2] as follows, for the
cess in the regime with boiling for the present system as well as for the
concentric-tube open thermosyphon
common boiling in pool. The exponent in equations (5) and (6) stands
in good accordance with values reported in [17].
qc = 0.980 - t r ° - 2 6 A T (1 X 106 =s tr) (2)
r Boiling Heat-Transfer Coefficients. It may be of great interest
to compare the present experimental results with those for the pool
qc = 0.197 - t r °- 388 AT (t r < 1 X 105) (3) boiling reported by Nishikawa, et al. [18,19], who extensively inves-
r tigated the effects of the saturation pressure on the pool-boiling heat
for the single-tube open thermosyphon transfer using fluorocarbon R - l l and R-113 refrigerants as working

Journal of Heat Transfer AUGUST 1981, VOL 103 / 475

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 q b W/m !
10'
6 8 103 2
+
CONCENTRIC TUBE ,-
d0mm Dmm
-
_ © 32.0 -i Ret [18] '
a 48.0

u -
o.
o
o
60.0
66.6 -100.0
76.0
• 8a6-! /' e$ '^ x > S
,'<s*Utr
-2
-E
^
E
IlO 3
»5£"«
CONCENTRIC TUBE
d0mm 0 mm
o 32.0 -
- SINGLE TUBE / « • < $ 2 a a 48.0
. 0 _ 100.0/ / « r ~ o 60.0
o 66.6
/ ° »* 0 Eq.(7) o 76.0
»•*<?
_ • 88.6 -

j?wx%
SINGLE TUBE
o 1000
L
/ s
10' • 1 1 1 1 1 1 I I
1
10J 6 8 10"
qb kcal/m'h Fig. 8 qb versus £

Fig. 7 Correlations of boiling heat transfer data

W/m 2
fluids. It appears that this comparison may give an evaluation for the 4 5
1 1 1 1 1 1 1 !• 1 >
boiling heat-transfer performances of the present thermosyphons.
CONCENTRIC TUBE / / ~
The heat-transfer coefficient ab as a function of heat flux qb is d0mm D mm / / ~
shown in Fig. 7, where p denotes the dimensionless pressure nor- 7- o 32.0 i «25V. / /
malized by the atmospheric pressure, thus taking into account the 6-- 0 48,0
o 60.0
saturation pressure. In the present program p is varied from 1.0 to 5-
e 66.6
1.65. For the calculation of ab in term of qb/(Tw — Ts) the saturation r.
76.0
temperature Ts at the open end of the thermosyphon was utilized. \0 ® 88.6 J /i,1**? -25V.
u «!
From an inspection of this figure, the slope of the present experi-
mental results for both the concentric-tube and the single-tube open
' 3-
/°v<sv'
thermosyphons seem to be in closely good accordance with that of
Nishikawa, et al. [18,19], though the operating conditions are quite - / /jro /
/ Xi* '
different. It is, however, clear that the heat-transfer coefficients for / »,iT • "/ SINGLE TUBE -
the present thermosyphon are rather small as compared to that of / fi8°a* /' d0mm Dmm
Nishikawa, et al. [18,19], within the whole range examined. This may ' ' «?'* / ° '00.0
be attributed to the fact that in the case of the present program the / /
evaluation of the saturation temperature at the open end of the 3 4 5 6 7 8 9
10'
qca, kcal/m 2 h
thermosyphon is utilized, namely, that the smallest value of Ts
varying along the heated surface is adopted. It would be furthermore Fig. 9 Comparison of <7ex with q cal
found that the present results for the single-tube open thermosyphon
might be in good accordance with that of Nishikawa, et al. [18,19] if
the average of Ts which varies axially corresponding to the saturation in width the heat-transfer coefficient has a tendency to markedly
pressure-temperature relationship would be utilized. increase with a decrease in width due to occurrence of the coalescence
of bubbles, while for the annular spaces greater than 5.1 mm in width
Hence, it appears to be concluded that the boiling heat-transfer
the rate of increase of the coefficient is relatively small. Hence, it
performance for the single-tube open thermosyphon using the fluo-
appears to be reasonable that the boiling heat transfer for the con-
rocarbon R - l l as a testing fluid at least may qualitatively be quite
centric-tube open thermosyphon may be nearly independent of the
similar to that for the common pool boiling of fluorocarbon R - l l re-
diameter of inner tube (i.e., the width of annular space) because in
frigerant. Simple correlations which may satisfy the boiling heat-
the present program width of the annular space ranged from 34.0 to
transfer performances of fluorocarbon refrigerant for both the con-
5.1 mm.
centric-tube and the single-tube open thermosyphons are given as
follows, respectively. Effect of Void Ratio on Boiling Heat Transfer. An attempt
For the concentric-tube open thermosyphon was made to consider the relationship between the boiling heat flux
and the mean void ratio for the present thermosyphon. In Fig. 8 is
ab = 0.311(qb)°-8(p)0-23 (7) plotted the boiling heat flux qb as a function of the mean void ratio
e, which is the ratio of the vapor space to the total volume of the an-
For the single-tube open thermosyphon nulus. It can be seen in this figure that the boiling heat flux increases
ab = 0.398(<7i,)0-8(p)0-23 (8) steeply at first with an increase in mean void ratio; thereafter, it in-
creases with a gentle slope of dependency on the mean void ratio.
Effect of Diametric Ratio of Concentric-Tube on Boiling Heat In the range of qb greater than 103 kcal/m 2 h, it may appear that qb 2
Transfer. Now, in addition to the matters mentioned above, it ap- a e. This relationship is similar to that obtained by Nishikawa [20]
pears to be of importance to investigate the effect of the diameter of for pool boiling experiments using a heated plate facing upwards,
the inner tube on the boiling heat-transfer performance for the con- namely, qb2 * n, where n denotes the number of vapor columns.
centric-tube open thermosyphon. As will be seen from Figs. 5-7, it Gaertner, et al. [21] claimed that for the fully developed boiling regime
seems that the boiling heat fluxes may show no significant dependency the relationship of d0-n = Cx may be obtained, where d0 is defined as
on the diameter of inner tube (i.e., the width of annular space) for the diameter of a bubble just leaving the heated surface, and Ci as a
concentric-tube thermosyphon investigated in this study. uniform constant. On the other hand, it was pointed out by Nishikawa,
Ishibashi and Nishikawa [11] studied the effect of the width of et al. [20] and Jakob [22] that for the various heated surfaces, boiling
vertical annular channel heated from inside unlike the present system heat fluxes, and working fluids, d0-f is a uniform constant, where /
on the boiling heat-transfer characteristics by using pure water as a denotes the frequency of bubble formation. It may be quite natural
testing fluid, and found that for the annular spaces less than 2.7 mm to suppose here that the mean void ratio e is in proportion to do-n-f.

476 / VOL. 103, AUGUST 1981 Transactions of the ASME

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 According to the matters mentioned above, thus, it may be proven Turbine," Fourth International Heat Transfer Conference, Paris, HE 2.2,
that e « n. Therefore, it is concluded that the relationship of qb2 <* 1970.
7 Larkin, B. S., "An Experimental Study of a Two-Phase Thermosyphon
e for the present thermosyphon shown in Fig. 8 might be quite similar Tube," 70-CSME-6,84th Annual Meeting, Engng, Institute of Canada, Ottawa,
to that of Nishikawa [20] in the range of parameters examined. 1970.
Finally, for confirmation, all the data which encompass the boiling 8 Lee, Y., and Mital, U., "A Two-Phase Closed Thermosyphon," Inter-
heat-transfer process within the working fluid were reproduced in Fig. national Journal of Heat Mass Transfer, 15,1972, pp. 1673-1707.
9 Kusuda, H., and Imura, H., "Boiling Heat Transfer in an Open Ther-
9 using the abscissa as qcai (equations (2) or (3,4) + equations (5,6), mosyphon (1st Report)," Bulletin ofJSME, 16,1973, pp. 1723-1733.
respectively). It may be seen that for both the concentric-tube and 10 Kusuda, H., and Imura, H., "Boiling Heat Transfer in an Open Ther-
the single-tube open thermosyphons using fluorocarbon R - l l re- mosyphon (2nd Report)," Bulletin ofJSME, 16,1973, pp. 1734-1740.
frigerant as a testing fluid, calculated and experimental values agree 11 Ishibashi, E., and Nishikawa, K., "Boiling Heat Transfer in Narrow
within an average deviation of approximately ±25 percent. Space (1st Report)," Trans. JSME, 33,1967, pp. 121-129.
12 Seki, N., Fukusako, S., Cheng, K. C, and Kawabe, H., "Heat Transfer
Characteristics of Double Tube Thermosyphon," Proceedings of Brasilian
Conclusions Energy Congress, Rio de Janeiro, A-5,1978, pp. 43-53.
13 Rohsenow, W. M., "Heat Transfer with Evaporation," Lectures at
Boiling heat-transfer characteristics were experimentally investi- Symposium on Heat Transfer, University of Michigan, 1953, pp. 101-1491.
gated for fluorocarbon R - l l refrigerant in the concentric-tube open 14 Clark, J. A., and Rohsenow, W. M., "Local Boiling Heat Transfer to
thermosyphon. Varying the different parameters while maintaining Water at Low Reynolds Numbers and High Pressures," ASME JOURNAL OF
uniform temperature along the heated outer tube, boiling data have HEAT TRANSFER, Vol. 76,1954, pp. 553-562.
15 Bergles, A. E., and Rohsenow, W. M., "The Determination of Forced-
been obtained. It is concluded from the present experimental results Convection Surface-Boiling Heat Transfer," ASME JOURNAL OF HEAT
that total heat flux can be appropriately predicted by a superposition TRANSFER, Vol. 86,1964, pp. 365-372.
of the convective heat flux given by available correlations (equations 16 Blall, T. A., and Adt. Jr, R. R., "An Experimental Investigation of Boiling
(2) or (3, 4)) and pure boiling heat flux given by available correlations Heat Transfer and Pressure-Drop Characteristics of Freon 11 and Freon 113
Refrigerants," AIChE Journal, Vol. 10,1964, pp. 369-373.
(equations (5, 6)). For design purposes for the concentric-tube and 17 McAdams, W. H., Heat Transmission, 3rd edition, McGraw-Hill, New
the single-tube open thermosyphons these correlations may be uti- York, 1963.
lized. 18 Nishikawa, K., Fujita, K., and Hidaka, S., "Nucleate Boiling Heat
Transfer of Fluorocarbon Refrigerants (1st Report)," Preprint of National Heat
Transfer Symposium of Japan, 1977, pp. 130-132.
Acknowledgment 19 Nishikawa, K., Fujita, K., Ota, H., and Hidaka, S., "Nucleate Boiling
Heat Transfer of Fluorocarbon Refrigerants (2nd Report)," Preprint of National
The authors are indebted to Mr. Y. Usami for his help in the ex- Heat Transfer Symposium of Japan, 1978, pp. 148-150.
periments. 20 Nishikawa, K., "Effect of Surface Roughness on Boiling Heat Transfer,"
Trans. JSME, Vol. 20,1954, pp. 808-815.
21 Gaertner, R. N., and Westwater, J. W., "Nobel Method for Determining
References Nucleate Boiling Sites," Chemical Engineering Progress Symposium., Ser.,
1 Hayakawa, M., "Geothermal Energy," NHK Books, 114, Nippon Hoso Vol. 55,1959, pp. 58-61.
Kyokai, 1977. 22 Jakob, M., Heat Transfer, Vol. 1,1949, John Wiley, New York.
2 Seki, N., Fukusako, S., and Koguchi, K., "Single-Phase Heat Transfer 23 McDonald, T. W., Hwang, K. S., and Diciccio, R., "Thermosiphon Loop
Characteristics of Concentric-Tube Thermosyphon," Warme- und Stoffil- Performance Characteristics: Part 1. Experimental Study," ASHRAE Trans-
bertragung, 14,1980, pp. 189-199. action, Vol. 83,1977, pp. 250-259.
3 Cohen, H., and Bayley, F. J., "Heat Transfer Problems of Liquid Cooled 24 Ali, A. F. M., and McDonald, T. W., "Thermosiphon Loop Performance
Gas Turbine Bladed," Proceedings Institute Mechanical Engineers, 169,1955, Characteristics: Part 2. Simulation Program," ASHRAE Transactions, Vol.
pp. 1063-1080. 83,1977, pp. 260-278.
4 Long, E. L., "The Long Thermophile", Proceedings Permafrost Inter- 25 McDonald, T. W., and Ali, A. F. M., "Thermosiphon Loop Performance
national Conference, 1963, pp. 487-491. Characteristics: Part 3. Simulated Results," ASHRAE Transactions, Vol. 83,
5 Chato, J. C, "Condensation in a Variable Acceleration Field and the 1977, pp. 279-287.
Condensing Thermosyphon," ASME Journal of Engineering Power, Vol. 87, 26 Koguchi, K., "Heat-Transfer Characteristics of Concentric-Tube
1965, pp. 355-360. . Thermosyphon," M. S. Thesis, Department of Mechanical Engineering, Hok-
6 Genot, J., and LeGrives, E,, "Transfert de Chaleur des Ailettes de kaido University, Sapporo, Japan, 1979.

Journal of Heat Transfer AUGUST 1981, VOL. 103 / 477

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 M. Z. Hasan 1
Research Assistant,
School of Engineering and Applied Science,
Boiling Burnout During Crossflow
ower Cylinders, beyond the Influence
University of California,
Los Angeles, Calif.
Assoc. Mem. ASME

M. M. Hasan of Gratity 2
Research Assistant,
Assoc. Mem. ASME
New observations of burnout during the upflow and downflow of isopropanol and metha-
R. Eichhorn nol over horizontal cylinders are added to previous observations. A criterion is developed
to determine whether or not gravity influences burnout in any such flow. A prediction that
Dean of Engineering,
Fellow ASME
depends on one empirical constant is then developed to predict burnout when gravity is
not influential. Those of the existing data that are uninfluenced by gravity are represent-
ed within ±20 percent by the prediction. It is also shown that a low-speed downflow can
J. H. Lienhard cause the hydrodynamic burnout to be replaced with a very inefficient buoyancy burnout
Professor, mechanism.
Mechanical Engineering Department,
University of Houston,
Houston, Tex. 77004
Fellow ASME
large overhead
^bubble breaking
away
Boiling and Phase Change Laboratory,
Mechanical Engineering Department,
University of Kentucky,
Lexington, KY 40506 sheet-like jet
eaving cylinder

Introduction
Most industrial heat removal by flow boiling has been done within
tubes, or in the axial channels separating tubes within closely packed
cylindrical heater
tube bundles. However, there is increasing interest in the high heat
fluxes that can be achieved in the crossflow of liquids over single tubes
and tube bundles. Our concern is that of predicting the peak heat flux,
or "burnout" in such flows.
Previous Work. Many previous experimenters have provided free stream approaching
observations of the burnout heat flux for crossflows over single cyl- at velocity, uco
inders. Those data that appear to be free of such nuisance variables
as end-mounting influences, short aspect ratio effects, etc., include Fig. 1 Schematic diagram of vapor removal process during boiling on a
cylinder in a liquid crossflow
Vliet and Leppert's water data [1, 2], Cochran and Andracchio's Freon
113 data [3], Min's water and methanol data [4], Eichhorn's isopro-
traction factor, a. The existing data seemed to suggest that, for low
panol data [5], and Yilmaz and Westwater's Freon-113 data [6]. If
velocity flows, a was close to unity, and for high velocity flows it
there are other data, we are either unaware of them or do not have
enough documentation to deal with them here.
The work of Lienhard and Eichhorn [5] provided a strategy for a =* constant (pf/pg)3/i
predicting the peak heat flux on single cylinders with the help of a
mechanical energy stability criterion. They observed that burnout The Scaling of Gravity and Heater Diameter. The flow in all
occurs when more kinetic energy relative to the flow, PgUg(ug2/2) ceD, previous experiments was upward over the cylinder so that gravity
enters the wake than surface energy 2<ru„, is consumed by it. (See Fig. aided vapor removal. Normally, the Froude number, which might be
1. See also [5] or the discussion in [7]. This analysis gave defined as

u„ I 4 U/3 (1) Fr = y/pfu~2/gD{pf - pg) <* uJVJB (4)

hH determines whether or not gravity actually exerts an influence on the
«» \aWegl flow. Typically, a value of Fr >10 suggests that gravity ceases to be
where 3 We s = pgu„2DI<r. A simple energy balance gives a major influence on the flow. Only the data of Yilmaz and Westwater,
Pgh/g (u„ + Ug) • A-tjet
ic (2) and some isolated points of Min's, and Vliet and Leppert's satisfy this
criterion.
Lienhard and Eichhorn's prediction did not take account of the
influence of gravity and unfortunately the data in [1-5] all fell in the
7T<7n Ua (2a)2'a
1+ -1 •• ct + (3) range: 0.76 < Fr < 16.4. Their correlation might well predict results
h
Ps fs Ua We.1'3 that would be different if the gravity field were altered.
The one problem that remained was the evaluation of the jet con- There is a second influence that is connected to gravity, and that
is local surface tension—surface tension forces that are related to
1
Most of this study was done while J. H. Lienhard and M. Z. Hasan were at bubble behavior at the heater surface. Such bubbles might have a role
the University of Kentucky. in shaping the liquid-vapor interface (and in determining a) at low
2
This work was supported by the National Science Foundation under Grant velocity, but not at high velocity. Gravity will also have a role in
ENG77-25029. shaping these bubbles. Thus there should also be an influence of the
3
Symbols not defined in context or in Fig. 1 are conventional ones. They are liquid Weber number, We/
defined in the Nomenclature.
Contributed by the Heat Transfer Division for publication in the JOURNAL PfUa2D liquid inertia
OP HEAT TRANSFER. Manuscript received by the Heat Transfer Divison July We/ = —- ~ ; (5)
24,1980. a capillary force

478 / VOL. 103, AUGUST 1981 Transactions of the ASME

Copyright © 1981 by ASME
Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 if We/ is small—say less than 10.
We have carefully studied the dependence of (j> on both Fr and We/.
For upflow, the existing </> data correlate reasonably well with We/,
even when We/ is as low as 10. For downflow, there is a diameter effect
of Fr and <f> does not correlate with either Fr or We/ alone unless both
parameters are large. From our observations, however, the demar-
cation between gravity dependent burnout and gravity independent Ugl-UoQ
burnout becomes independent of diameter when | u „ | is large enough.
If this is to be the case, the point of demarcation must be governed
by a parameter, made up of Fr and We/, that does not depend on D.
We designate this parameter (similar to a Kutataladze number) as
G:

G = ^We/Fr2 Ml
•£ g(Pf - Pg)a Vgv/pf
T h e Region of Gravity-Influenced Behavior. We assume that
(6)
* i n left ~p\ UajL
^outright =/>f UootL-aDJ+pgtUg+Uoo) aD
equation (3) is correct for predicting the burnout heat flux, whether '"out sides = * in left - *oul right^f UooaD-^gfUg+UooJaD
or not gravity influences are present. It should only fail when such
influences become so extreme as to replace the hydrodynamic stability
mechanism with some other kind of burnout mechanism. Therefore,
= V"° D H(^> +1 )]
the influence of gravity will merely be that of changing the magnitude Fig. 2 Control volume around heater, and mass balance
of the contraction factor, a, in equation (3).
Objectives of the Present Work. The previous considerations We can only complete the derivation when we can predict Co-
point to the need for more experiments, and a more comprehensive However, consider one preliminary result of this balance. The me-
theory, than are now available. We therefore aim to chanical energy stability criterion, equation (3), makes it clear that
1 Provide observations of g ma x in both upflow and downflow over for high liquid velocities (Weg ~* large) the dimensionless heat flux,
cylinders, with Fr being varied from «—10 to » + 1 0 , and We/ from 0, must approach a. Under those circumstances Co must also ap-
near zero to » 1 0 . proach zero in accordance with equation (8). Such a result matches
2 Use these data to identify a critical value of \G\ — Gc, above our expectation because, at high heat flux, vapor moves at a speed
which gravity exerts no influence. close to u» and blankets almost all of the cylinder, the no slip condi-
3 Determine a for flows for which \G\ > Gc. tions disappears, and the flow approaches a perfect potential flow of
4 Provide a general prediction or correlation of 0 for | G | > Gc. liquid around the cylinder.
Notice, too, that the drag is composed of three contributions:
2
T h e P r e d i c t i o n of g m a x i n t h e H i g h - S p e e d L i m i t CD = + -•[•-J (9)
The Prediction of a. Figure 2 shows a control volume located We/
around the cylindrical heater. We can neglect viscous influences be- liquid momentum cont.
cause the liquid flow around the cylinder is almost a pure slip-flow. surface tension (always positive).
The vapor "lubricates" the liquid flow around the cylinder so the only contribution vapor thrust contribution
significant drag force, F, exerted on it is pressure drag. Then a mo- (always positive) (always negative).
mentum balance gives
When we substitute actual numbers in equation (9), in the next sec-
S(forces and momentum fluxes to right) = F(N/m) tion, we shall see that the vapor jet exerts a thrust, driving the heater
into the flow throughout most, if not all, of the velocity range. Thus
CD is generally negative.
PfU„2L — pfU„2(L — aD) — Pg(ug + u^)2aD The evaluation of Co- The drag coefficient consists of two parts.
The first one is explicitly 2/We/ and the other becomes, with the help
— pfU«,2aD 1 - ^ + 1 + 2<r = F (7) of equation (3),
pf\u
^ .... ri/3
Simplifying this equation and introducing the notations: r = pflpg, +1
Weg = pguJD/o, We/ = pfujDh, CD = Flpfu„2D, and <pla = 1 + r a \a •)-K
) \ rj We/ 1 / 3 We/ 1/3
Ug/uoo, we obtain a relation between a and F and Co: This can be rearranged into

<t> rCD r\CD (8) (10)

We, We/ rl/3 [r J (We/) 1 ' 3 (we 2/

,-Nonienclature-
=
Aj et ; Aheater cross-sectional area of vapor jet ing upon whether u„ is directed upward or = the burnout, or peak, heat flux on the
leaving heater; area of the heater itself downward) cylinder
C = empirical constant defined in equation r = pf/pg
Gc = critical value of | G \ below which gravity
(ID We/, Weg = Weber numbers, p/ua 2D/a, and
influences the flow
Co *= drag coefficient, Flpfiia2D psu„2D/ff, respectively
D = diameter of heater g = acceleration of gravity a = jet contraction factor (width of vapor
F = drag force exerted on the cylinder in the hfg = latent heat of vaporization of boiled jet/D)
direction of flow (may be negative) fluid Pf, pg = saturated liquid and vapor densities
Fr = Froude number, Vp/"» 2 /gD(/o/ - Pg) a-.; ug = velocity of liquid relative to the of boiled fluid, respectively
a uJ\fgD cylinder (positive in the upward direction); a = surface tension of the boiling fluid
G = gravity influence parameter, VWe/Fr 2 =^ velocity of vapor in the jet relative to the (j> = dimensionless peak heat flux,
uj-s/galpj (positive or negative depend- liquid moving at u«, irqmax/pahfg\u„\

Journal of Heat Transfer AUGUST 1981, VOL 103 / 479

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 -up!low
II I
_downflow
5
The desired flow 5
direction is obtained by 5
blocking the flow either ~4. 4
01 A or 01 8. (Dashed .,.,. ,:~ J9
line shows another pipe,
7 7
between 12 and 8 behind 9)
8 c.'.-.:_-_::'-::>
2
y

- 2

3
== 3
3
-
Fig. 3 A schematic diagram oj the experimental facility (1) 200 gpm cen-
Irlfugal pump (2) by-pass (3) orifice-meter (4) heater (5) drain drum (6) filter
pump (7) IIIter (8) test chamber (9) diffuser (10) glass window (11) nozzle Flg.4 Test section (to approximate scale only) (1) test chamber (2) nozzle
(12) overhead tank (13) condenser coil (14) overflow pipe (15) bubble re- (10.16 em X 2.54 em) (3) diffuser (7.66 em circular to 10.16 em square
moval pipe for downflow, liquid return pipe tor uptlow section) (4) brass seat tor holding device (5) glass windows (6) bubble re-
moval pipe for downflow case. (7) power leads (8) thermometer (9) pressure
measurement tap
Finally we recall from [5] that a was guessed to vary as r 3/4 • That may
or may not have been accurate, but we do expect the exponent of the
r dependence of a to be on the order of unity. In any event, (alr)1/3
will exhibit very weak dependence on a or r, if it exhibits any.
Therefore, we seek a relation of the form:
r/J (r/J ) (2a)2/3 C (11)
r1/3 ~ -1 - (We r)1/3 = We/ 13

at high velocity, where C is an empirical constant.

The High-Velocity Prediction of q,. Substituting equation (3)
in equation (11) and solving the resulting quadratic formula for r/J, we
obtain
a ll3 = C(rI16)113 (12)
Substituting equation (12) in (3), we get
C3 r C2 ( r 2 )1/3
r/J=-+-- (13)
16 4 Weg
It now remains only to determine C from experimental data.
There are now three orders of business to be resolved:
We must measure q max over a wide variety of conditions in both
upflow and downflow to identify Ge .
We must use the data for IG I > Ge to establish the value of C
using equations (3) and (11). em
We must then determine whether or not equation (13) predicts
observed values of qmax accurately, based on the empirical value of
C.
Fig. 5 The test-healer holding-device, givIng about 8 em unobstructed active
healer length
Experiment
The hot flow loop described in (4) and [5] and redesigned by M. Z. a calibrated ASME orifice meter. All data presented here were ob-
Hasan and M. M. Hasan has been used for the present tests. Complete tained in the same nozzle configuration.
descriptions of the apparatus and experimentaJ procedure are given The cylindrical heaters were of 14, 20, and 24 gage Nichrome wire.
in (8). Figure 3 is a schematic diagram of the system. The test section Each one was 13 em in length with the two ends electroplated with
(elements 8, 9,10 and 11) is shown in its right-side-up, upflow position a tapered copper sheath to provide good electrical connections and
on the left and in its up-side-down, downflow position on the right: to minimize end effects. This defined an active unplated length that
for upflow: Flange A remains open and flow returns down pipe vari!Jd from wire to wire but stayed in the neighborhood of 8 em.
15. Flange B is blanked off. The test section is replaced with a neutral Figure 5 shows a heater wire.
run of pipe on the right. Alternating current was applied to the wires in small steps. The
for downflow: Flange A is blanked off and B is open. The test burnout was noted both visually and by a sudden drop of applied
section is replaced with a neutral run of pipe on the left. current that accompanied it. (The resistance of the wire was carefully
The system is designed to accommodate a variety of organic liquids obtained prior to the tests.) The probable error of qmax was about ±2.2
or water. The 2.54 cm wide, two-dimensional nozzle, shown in two percent based on the component measurements and the reproduc-
views in Fig. 4, delivers liquid at velocities (u~) up to 2.6 mls over a ibility of observations at any condition was found to be within ±2.5
10.16 cm span. (The cylinder is not shown in place between holders, percent.
4). The nozzle was designed according to the method of Cheers (10) At each condition, a plot of qmax against liquid subcooling was made
and the velocity profile above it is within 2 percent of being uniform, and extrapolated over the last 1°C to zero subcooling. Thus all data
with the flow virtually free of turbulence. The velocity is inferred from are reported here for saturated liquid.

480 / VOL. 103, AUGUST 1981 Transactions of the ASME

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 image of
upflow data
image of .
i \ A data for isopropanol inflow data—.. / o data for isopropanol
>\ flowing over a 0.051 cm flowing over a 0.0813
\\ diam. cylinder diam. cylinder

upflow
data
upflow -
data

-20 -10 -20 -10 20

4/
/
image of
upflow data
a data for isopropanol
flowing over a 0.16 cm
diam cylinder

upflow
data

-20 -10 0 K) 20
Gravity influence parameter, Gauay/^/ger/JD,

Fig. 6 Influence of gravity on burnout

Results and Discussion Notice that the data diverge only modestly in the range 10 > \G\
The Identification of the Regime of Gravity Influence. The > 3. But, for — 3 < G < 0, <t> plunges abruptly and drastically. We be-
collected data for the present experiments are tabled in Appendix A. lieve that in the range 10 > \G\ > 3 burnout continues to obey the
The table includes the raw data (<jmax, u«,, and D) and it includes their mechanical energy stability criterion, but that a is altered by gravity
dimensionless reductions in the form of <j>, Fr, G, We/, and W e e - 1 ' 3 . in ways that are not reflected in equation (11).
It also includes the contraction factor, a, calculated by trial and error When G is between - 3 and zero in a downflow, a dramatic change
from known values of <p and We^ _ 1 / 3 using equation (3). The reader is evident in the vapor wake. We observed that it became very
should be warned that neither 0 pre d nor a are offered as valid for | G \ thick—almost triangular with the apex at the wire. This large,
<GC. oddly-shaped bubble is suspended in the flow around the wire. As u»
The dimensionless burnout heat flux is plotted against G for several is decreased, Qmax
cases in Fig. 6. Downflow is interpreted as the case for which u „ is drops far below the value predicted (see e.g., [9])
negative, so downflow data are plotted on left side, in each case. Notice for pool boiling. Our data include q m a x values as small as 18 percent
that for \G\ > 10 the curves are almost perfect mirror images. This is of the pool boiling limit.
demonstrated by fairing curves through the data on each side, and These low values of g m a x reflect a burnout model quite different
then dashing the right-hand curve in on the left side, and vice versa. from the hydrodynamic stability limit. Burnout occurs, in this case,
Consequently we identify the critical gravity influence parameter as the result of a balance between the rate of buoyant rise of the vapor
as slug around the cylinder and the downward velocity of the liquid. One
Gc = 10. would have to be able to predict the equilibrium diameter of such a
slug, and the heat transfer through it as well, to determine the mini-
Journal of Heat Transfer AUGUST 1981, VOL 103 / 481

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 Table 1 Combined r a w and reduced data

"oo °.max Imax

m MW Weg"' Wef Fr G a MW 0expt. We„ Wef Fr 0pred.
0expt 0pred.
s m2

Isopropanol, r = 334.76; diameter = 0.150 cm Isopropanol, r = 3 2 1 . 5 1 ; diameter = 0.158 cm (continued)

0.279 0.585 4.490 4.000 5.2 2.30 2.29 0.498 -0.794 0.624 1.631 1.934 44.4 -6.38 -6.52 0.289
0.350 0.615 3.763 3.442 8.2 2.89 2.88 0.468 -0.915 0.635 1.440 1.760 59.0 -7.35 -7.51 0.271
0.515 0.650 2.703 2.661 17.8 4.25 4.23 0.402 -2.570 0.985 0.795 0.884 465.5 -20.64 -21.10 0.245 0.918
0.623 0.685 2.355 2.344 26.0 5.14 5.12 0.385
0.808 0.730 1.935 1.971 43.7 6.66 7.79 0.339 Isopropanol, r = 3 2 1 . 5 1 ; diameter = 0.152 cm
1.082 0.770 1.524 1.622 78.4 8.92 8.89 0.320
-0.197 0.320 3.371 4.962 2.6 1.61 1.62 0.250
1.233 0.800 1.389 1.487 101.9 10.16 10.13 0.309 1.383
-0.267 0.170 1.321 4.052 4.8 -2.19 2.19 0.084
1.333 0.808 1.298 1.411 119.1 10.99 10.95 0.298 1.328
-0.332 0.210 1.313 3.504 7.5 -2.72 2.73 0.102
1.476 0.865 1.255 1.319 146.0 12.17 12.12 0.305 1.261
-0.418 0.455 2.259 3.005 11.8 -3.42 3.43 0.269
1.682 0.885 1.127 1.209 189.6 13.87 13.82 0.289 1.182
-0.502 0.465 1.922 2.660 17.1 -4.11 4.12 0.249
1.942 0.915 1.009 1.098 252.7 16.01 15.95 0.274 1.102
-0.577 0.530 1.906 2.424 22.6 -4.73 4.74 0.276
2.202 0.972 0.945 1.010 324.9 18.15 18.09 0.272 1.038
-0.618 0.550 1.847 2.316 25.9 -5.06 -5.08 0.279
2.407 0.990 0.881 0.952 388.2 19.84 19.77 0.262 0.996
-0.663 0.610 1.909 2.210 29.8 -5.43 -5.44 0.308

Isopropanol, r = 334.76; diameter = 0.081 cm

Isopropanol, r = 3 2 1 . 5 1 ; diameter = 0.081 cm
0.290 0.770 5.686 4.792 3.0 3.25 2.38 0.554
0.209 0.250 2.482 5.884 1.6 -2.35 -1.72 0.127
0.350 0.775 4.742 4.227 4.4 3.93 2.88 0.502
-0.231 0.215 1.931 5.504 1.9 -2.59 -1.90 0.096
0.437 0.785 3.847 3.646 6.9 4.90 3.59 0.450
-0.273 0.095 0.722 4.925 2.7 -3.06 -2.24 0.027
0.610 0.865 3.037 2.919 13.5 6.84 5.01 0.424
-0.284 0.170 1.242 4.796 2.9 -3.19 -2.33 0.061
0.796 0.900 2.421 2.444 22.9 8.93 6.54 0.381
-0.290 0.185 1.324 4.730 3.0 -3.25 -2.38 0.068
0.991 0.970 2.096 2.112 35.5 11.12 8.14 0.370
-0.321 0.230 1.487 4.420 3.7 -3.60 -2.64 0.089
1.065 0.955 1.920 2.013 41.0 11.95 8.75 0.346
-0.327 0.150 0.952 4.366 3.9 -3.67 -2.69 0.047
1.230 0.990 1.724 1.829 54.7 13.80 10.10 0.332 1.630
fl.332 0.266 1.625 4.322 4.0 -3.72 -2.73 0.104
1.467 1.060 1.547 1.626 77.9 16.46 12.05 0.326 1.483
0.362 0.380 2.178 4.080 4.7 -4.06 -2.97 0.172
1.760 1.100 1.338 1.440 112.1 19.74 14.46 0.304 1.349
•0.394 0.550 2.897 3.856 5.6 -4.42 -3.24 0.280
1.942 1.130 1.246 1.349 136.5 21.79 15.95 0.296 1.283
-0.433 0.652 3.124 3.621 6.8 •4.86 -3.56 0.338
2.233 1.225 1.175 1.229 180.4 25.05 18.34 0.300 1.196
-0.440 0.645 3.042 3.582 7.0 -4.94 -3.61 0.330
2435 1.270 1.117 1.160 214.5 27.32 20.00 0.297 1.146
-0.482 0.660 2.840 3.371 8.4 -5.41 -3.96 0.323
-0.509 0.720 2.935 3.251 9.4 -5.71 -4.18 0.354
Isopropanol, r = 334.76; diameter = 0.050 cm
-0.545 0.675 2.570 3.106 10.7 -6.11 -4.48 0.310
0.261 0.925 7.589 6.037 1.5 3.73 2.14 0.620 -0.563 0.665 2.451 3.039 11.5 -6.32 -4.62 0.298
0.273 0.900 7.060 5.860 1.7 3.90 2.24 0.581 -0.591 0.760 2.668 2.943 12.6 -6.63 -4.85 0.350
0.375 0.830 4.740 4.741 3.1 5.35 3.08 0.433 -0.604 0.720 2.474 2.900 13.2 -6.78 -4.96 0.320
0.425 0.865 4.358 4.362 4.0 6.06 3.49 0.428 -0.612 0.725 2.458 2.875 13.5 -6.86 -5.03 0.321
0.462 0.928 4.301 4.126 4.8 6.60 3.80 0.451 -0.682 0.670 2.038 2.675 16.8 -7.65 -5.60 0.269
0.465 1.000 4.605 4.108 4.1! 6.64 3.82 0.500 -0.696 0.746 2.224 2.639 17.5 -7.81 -5.72 0.309
0.607 0.955 3.369 3.439 8.2 8.67 4.99 0.401 -0.706 0.760 2.234 2.614 18.0 •7.92 -5.80 0.315
0.610 1.070 3.756 3.428 8.3 8.71 5.01 0.469 -0.775 0.900 2.410 2.456 21.7 , -8.69 -6.36 0.377
0.812 1.000 2.637 2.833 14.7 11.59 6.67 0.360 -1.050 1.055 2.085 2.006 39.8 -11.78 -8.62 0.388
0.985 1.095 2.381 2.491 21.7 14.06 8.09 0.364 -1.108 0.975 1.826 1.935 44.4 -12.43 -9.10 0.337
1.145 1.130 2.113 2.253 29.3 16.35 9.41 0.347 -1.108 0.900 1.685 1.935 44.4 -12.43 -9.10 0.302
1.455 1.355 1.994 1.920 47.3 20.78 11.95 0.384 1.696 -1.372 1.165 1.762 1.678 68.0 -15.39 -11.27 0.376 1.477
1.458 1.250 1.836 1.918 47.5 20.82 11.98 0.343 1.695 -1.447 1.085 1.556 1.620 75.7 -16.23 -11.88 0.329 1.436
1.722 1.287 1.600 1.716 66.2 24.59 14.15 0.322 1.549 -1.480 1.145 1.605 1.596 79.1 -16.60 -12.15 0.349 1.419
2.010 1.385 1.476 1.548 90.2 28.70 16.51 0.322 1.427 -1.702 1.184 1.443 1.454 104.7 -19.09 -13.98 0.333 1.319
2.248 1.462 1.393 1.437 112.9 32.10 18.47 0.322 1.347 -1.722 1.280 1.542 1.442 107.1 -19.32 -14.14 0.368 1.311
-1.850 1.300 1.458 1.375 123.7 -20.75 -15.19 0.358 1.264
Isopropanol, r - 3 2 1 . 5 1 ; diameter = 0.163 cm -2.202 1.320 1.244 1.224 175.2 -24.70 -18.08 0.325 1.158
-2.407 1.445 1.246 1.154 209.3 -27.00 -19.77 0.345 1.108
-0.527 0.600 2.362 2.516 20.2 -4.17 -4.33 0.356
-2.407 1.300 1.121 1.154 209.3 -27.00 -19.77 0.300 1.108
-0.615 0.610 2.058 2.270 27.5 -4.86 -5.05 0.332
-0.678 0.650 1.989 2.127 33.4 -5.36 -5.57 0.341
-0.784 0.690 1.826 1.930 44.7 -6.20 -6.44 0.338 Isopropanol, r = 3 2 1 . 5 1 ; diameter 0.051 cm
-0.808 0.540 1.387 1.892 47.5 -6.39 -6.63 0.237 -0.267 0.265 2.059 5.831 1.6 -3.77 -2.19 0.098
-0.836 0.700 1.737 1.850 50.8 -6.61 -6.86 0.331 -0.322 0.335 2.159 5.147 2.4 -4.55 -2.64 0.124
-1.298 0.830 1.327 1.379 122.5 10.26 10.66 0.315 1.267 •0.386 0.500 2.688 4.561 3.4 -5.46 -3.17 0.201
-1.474 0.855 1.204 1.267 158.0 11.66 12.10 0.301 1.188 -0.398 0.700 3.650 4.469 3.6 -5.63 -3.27 0.321
-1.853 0.960 1.075 1.088 249.6 14.65 15.22 0.300 1.062 -0.458 0.545 2.469 4.069 4.8 -6.48 -3.76 0.207
-2.044 0.980 0.995 1.019 303.8 16.16 16.78 0.289 1.013 -0.501 0.830 3.438 3.833 5.7 •7.08 -4.11 0.360
-2.156 1.015 0.977 0.984 338.0 17.05 17.70 0.291 0.989 -0.643 0.820 2.646 3.245 9.4 -9.09 -5.28 0.306
-2.421 1.035 0.887 0.910 426.1 19.15 19.88 0.276 0.936 -0.703 0.818 2.414 3.058 11.2 -9.94 -5.77 0.290
-2.543 1.075 0.877 0.881 470.2 20.11 20.88 0.279 0.916 -0.763 0.840 2.284 2.896 13.2 1079 -6.27 0.286

Isopropanol, r = 321.51; diameter 0.158 cm Isopropanol, r = 3 2 1 . 5 1 ; diameter 0.050 cm

-0.348 0.300 1.789 3.352 8.5 2.79 -2.86 0.168 -0.220 0.180 1.698 6.678 1.1 •3.14 -1.81 0.061
-0.476 0.455 1.983 2.721 16.0 3.82 -3.91 0.253 -0.273 0.200 1.520 5.783 1.7 -3.90 -2.24 0.063
-0.542 0.518 1.983 2.495 20.7 -4.35 -4.45 0.282 -0.279 0.200 1.490 5.707 1.7 -3.98 -2.29 0.063
-0.692 0.560 1.679 2.120 33.8 -5.56 -5.68 0.271 -0.279 0.195 1.450 5.700 1.7 -3.98 -2.29 0.060

482 / VOL 103, AUGUST 1981 Transactions of the ASME

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 Table 1 (Continued)

"oo flmax "oo Imax

m MM) 4>expt Weg"J Wef Fr Q or.
0pred. m MW <t>expt. Weg"> Wef Fr G a
0pred.
i m2 s "m7
0.355 0.340 1.987 4.854 2.8 -5.07 -2.92 0.119 1.658 1.470 2.059 1.536 169.8 13.62 13.32 0.508 2.232
-0.418 0.725 3.599 4.353 3.9 -5.97 -3.43 0.326 1.672 1.590 2.208 1.527 172.9 13.74 13.44 0.560 2.223
-0.440 0.727 3.428 4.207 4.3 -6.28 -3.61 0.318 1.820 1.600 2.042 1.443 204.8 14.95 14.63 0.534 2.131
-0.502 0.830 3.431 3.853 5.6 -7.17 -4.12 0.356
-0.596 0.910 3.168 3.437 7.9 -8.51 4.89 0.368 Methanol, r = 6 1 5 . 1 1 ; diameter = 0.081 cm
-0.618 0.860 2.888 3.354 8.5 -8.82 -5.07 0.332 1.214 3.073 2.327 48.9
1.606 13.62 9.76 0.561
0.826 0.950 2.387 2.765 15.2 -11.79 -6.78 0.323 1.369 3.139 2.147 62.1 11.00
1.850 15.36 0.631 2.888
-1.039 1.075 2.147 2.034 38.2 -18.69 -10.75 0.334 1.728 1.583 2,971 1.949 83.1 12.72 0.650
2.025 17.76 2.680
-1.235 1.200 2.016 2.114 34.0 17.63 10.14 0.350 1.784 1.744 2.863 1.827 100.8 14.02 0.661
2.150 19.56 2.548
-1.519 1.360 1.858 1.842 51.5 -21.69 -12.47 0.365 1.593 1.855 2.767 1.754 114.1 14.91 0.659
2.210 20.81 2.469
•1.835 1.460 1.651 1.624 75.1 -26.20 -15.07 0.356 1.439
Methanol, r = 6 1 5 . 1 1 ; diameter = 0.051 cm
Methanol, r = 6 1 5 . 1 1 ; diameter = D.151 cm 1.242 1.670 3.123 2.673 32.2 17.56 9.98 0.489
1.171 1.370 2.717 1.936 84.9 9.62 9.41 0.580 1.369 1.980 3.359 2.505 39.1 19.35 11.00 0.583 3.284
1.327 1.482 2.593 1.781 108.9 10.90 10.66 0.594 2.498 1.638 2.240 3.176 2.223 56.0 23.16 13.16 0.617 2.980
.'i i i h i
mum value of g m a x in this case. We make no attempt here to predict
such burnout, nor do we claim to identify the transition to this u> W

mechanism using G = —3. A correct general criterion would doubtless

- -
be more complicated. The data for — 3 < G < 0 do not correlate on
either Fr or We/ alone and they are insufficient to justify their being
represented as a function of G. - -
The Evaluation of C. The left-hand side of equation (11) is
C = 0 245 " —jf
plotted against W e ^ - 1 ' 3 in Fig. 7, based on only those data from the
present study (and those from [1] and [5])4 for which \G\ > 10. They - J_. .y*---^ 0 245 W f ! a -2'
3

give a remarkably straight line on logarithmic coordinates and fix the

constant C—the slope—as 0.245. ; 2 ;
The Prediction of $. Using C = 0.245 we obtain the final high
speed formula for the dimensionless peak heat flux, from equation - ' Data !

o present results
-
(isopropanol, uptlow)
(13). / © present resutts
- / (isopropanol, downflow) —
16.3 7 • present results
<t> = 0.0009191 +-We/ 1/3 (14) / {methanol, upflow)

— j/ V Vliet 8 Upper!
(water) ' —
: • YilmozaWeshwtef
Predicted values of <j> are compared with all observed values for | G \ / (F-II3)
-
>10 in Fig. 8. The data are represented well within ±20 percent by
equation (14). Two Westwater data are high, but they were obtained - -
under about 6°C subcooling.
Figure 9 shows dimensional values of (/max as a function of u „ for
0.16 cm dia. cylinders in isopropanol. The prediction of qmBX from • 1 1 1 • 1
equation (14) is included as a solid line on the curve, in the range | G \ 1.0 Z.0 4.0

> 10. The low velocity behavior (\G\ < 10) is sketched in as a dash-dot
line. Notice the discontinuity of the dash-dot line in low-velocity Fig. 7 Correlation of the liquid and vapor momentum contributions to the drag
on a cylinder during boiling
downflow, and notice how far below (<7max) pool boiling the burnout
io i 1 1—i—i i i i j i 1 r
heat flux falls.
Both Fig. 6 and Fig. 9 suggest that the present correlation will be
valid for G's substantially below 10 in upflows. We opt for caution and _ o present results, isopropanol (upflow)
make no such claim here, since we lack a rationale for doing so. <D present results, isopropanol (downflow)
The Magnitude of a. Equation (12) gives a present results, methanol
v Vliet a Leppert, water
a = 0.000919 r (15) # Yilmaz 8 Westwater, F-II3
For high values of r this formula gives values of a that exceed unity.
Values of a significantly greater than unity are probably not really
attained. It is more likely that a simply reaches an upper limit of about
unity and stays there, in accordance with the assumption [5] that a
=* 1 for slow flows. We should also note that a is velocity-independent
when | G | is large.

Conclusions
1 The influence of gravity on the peak heat flux during boiling
from a single horizontal cylinder in a crossflow becomes inconse-
quential when \G\ > 10.
2 The peak heat flux in this case is given within ±10 percent by
equation (14).
O.I 0.2 0.4 0.6 1.0 2 4 6 10
Predicted dimensionless peak heat flux, <£pred

Fig. 8 Comparison of high-speed burnout prediction with experimental

4
Data from the other sources fall in the range \G\ <10. data

Journal of Heat Transfer AUGUST 1981, VOL. 103 / 483

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 1 1 1 r Acknowledgment
W e a r e grateful t o G e n e U n g a r a n d E . B . Y a t e s for t h e i r h e l p w i t h
the present experiment.

References
1 Vliet, G. C , and Leppert, G., "Critical Heat Flux for Nearly Saturated
Water Flowing Normal to a Cylinder," ASME JOURNAL OF HEAT TRANSFER,
Vol. 86, No. 1,1964, pp. 59-67.
2 Leppert, G., and Pitts, C. C , "Boiling," in Advances in Heat Transfer
T. F. Irvine, Jr., and J. P . Hartnett, eds., Vol. 1, Academic Press, New York,
1964.
3 Cochran, T. H., and AndraCchio, C. R., "Forced Convection Peak Heat
Flux on Cylindrical Heaters and Refrigerant 113," NASA D-7553, Feb.,
-1-0 -0.5 0 05 1.0 1.5 1974.
flow velocity, u<n(m/s),positive upward 4 Min, T. K., "Boiling on Cylinders in Cross Flow—Low Velocity Peak
Heat Flux Measurements for Water and Methanol," Master's Thesis, University
Fig. 9 Typical behavior of data in dimensional coordinates of Kentucky, Mechanical Engineering Department, 1975.
5 Lienhard, J. H., and Eichhorn, R., "Peak Boiling Heat Flux on Cylinders
in a Cross Flow," International Journal of Heat Mass Transfer, Vol. 19,1976,
pp. 1135-1142.
3 The drag coefficient in this case is 6 Yilmaz, S., and Westwater, J. W., "Effect of Velocity on Heat Transfer
to Boiling Freon-113," JOURNAL OF HEAT TRANSFER, Vol. 102, No. 1,1980,
2 0.245 0.0150 pp. 26-31.
CD (9a)
We/ Wey 2/3 Weg 1 ' 3 7 Lienhard, J. H., A Heat Transfer Textbook, Prentice-Hall, Englewood
Cliffs, N.J., 1981.
8 Hasan, M. Z., "Peak Boiling Heat Flux in Upflow and Downflow across
4 When | G | is reduced sufficiently far below 10 the hydrodynamic Single Cylindrical Heaters," Master's Thesis, University of Kentucky, Me-
stability criterion for burnout is replaced by a buoyancy criterion. chanical Engineering Department, 1980.
When this occurs, qmax can fall far below the pool boiling limit, instead 9 Sun, K. H., and Lienhard, J. H., "The Peak Pool Boiling Heat Flux on
of exceeding it, in the downflow case. Horizontal Cylinders," International Journal of Heat Mass Transfer, Vol. 13,
1970, pp. 1425-1439.
5 a is velocity-independent and equal to 0.000919r, when \G\> 10 Cheers, F., "Note on Wind Tunnel Contractions," Aeronautical Research
10. Council, Reports and Memoranda No. 2137, Mar 1945.

484 / VOL. 103, AUGUST 1981 Transactions of the ASME

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 Density-Wawe Oscillations in Sodium
Heated Once-Through Steam
H. € . Unal
Central Technical Institute TNO,
P.O. Box 342,
7300 AH Apeldoorn, The Netherlands
Generator Tubes
Inception conditions of density-wave oscillations were determined in two sodium heated
once-through steam generator tubes, i.e., in a 44.43 m long helical coil of 0.018 i.d. and in
a tube of 0.0131 m i.d., comprised of a 9.40 m long vertical- and an 11.05 m long, V-shaped
horizontal tube. The operating conditions on the water/steam side for the experiments
were: pressure: 6-19.1 MN/m2; mass velocity 187-1020 kg/m2 s; inlet subcooling: 3.2-168
K; outlet steam quality: 1.15-2.38. The 306 data obtained and the 74 data found in the lit-
erature from sodium and electrically heated once-through steam generator tubes were
correlated within 7.5 percent accuracy for 98 percent of the time. The R.M.S. error for all
the 380 data is 3.33 percent. The observed density-wave oscillations are time-delay oscil-
lations and the length of the superheated steam region and the transit time in this region
practically govern the mechanism of these oscillations. An empirical relation has also
been established for this mechanism.

1 Introduction
A particular type of dynamic instability, i.e., Density-Wave Oscil- of secondary importance. The above phenomenon as well as the
lations (DWO), jeopardizes the safety of a sodium heated steam mechanism of the DWO observed in the test tubes were explained.
generator, a crucial component of the LMFBR cooling system. These An empirical relation has been established for the multiple regener-
oscillations cause flow maldistribution among the tubes or inside the ative feedback mechanism, which causes the DWO in the test
tubes, and thermal cycling at the upper tube plate and at the dryout tubes.
location. The DWO are low-frequency oscillations of which the period
is of the same order of magnitude as the transit time of a fluid particle 2 Experimental Apparatus and Procedure
in a steam generator tube. The DWO are due to multiple regenerative 2.1 The Loop. The flow diagram of the heat transfer loop used
feedback between the flow rate, the vapor generation rate and the is shown in Fig. 1, where the high-pressure part of the loop is drawn
pressure drop [1-2]. The DWO have been extensively studied for in heavy lines and the low-pressure part in fine lines. The main
comparatively short electrically heated tubes [1-2], but not for sodium specifications of the loop are reported in [5,12]. During high pressure
heated once-through steam generator tubes. The DWO data for the operation, the pressure is regulated by an automatic control valve and
latter are also scarce in literature. Furthermore stability models in the piston pump. The deaerator then serves as an expansion tank. The
the form of computer programs generally do not fit the data from volume of the high pressure part of the loop was approximately 0.05
sodium or gas heated once-through steam generator tubes well [3-4]. m 3 , excluding the test tube and by-pass. The sodium in the loop was
By using the 88 data of [4-10] obtained for quite a wide range of op- heated electrically. The maximum available power was about 0.5 MW,
erating conditions in six different sodium or electrically heated including the power of the preheater. Two test tubes, a Ferritic steel
once-through steam generator tubes (i.e., 1272 < Lh/d < 9502 and Test Tube (FTT) and a stainless steel Helically Coiled test Tube
X > 1), a correlation was established in [11] to predict the inception (HCT) were used in the loop described briefly above to determine the
conditions of the DWO. In that correlation, the effect of inlet inception conditions of the DWO. At the inlet and outlet of each test
throttling could not be taken into account for L^/d < 4118 due to lack tube, headers were mounted. Between these two headers, a vertical
of data. by-pass pipe of 49.3 mm i.d. was installed; thus the total pressure drop
The main purpose of the present study is to report the results of in each test tube was approximately constant. The by-pass pipe is not
the experiments carried out to determine the inception conditions shown separately in Fig. 1, but it is included in the part indicated as
of the DWO in two sodium heated once-through steam generator test section. The total length of a test tube was the length between the
tubes. One of these was a tube of total length 20.45 m and 0.0131 m inlet header and the outlet header. Flow control valves and flow me-
i.d., comprised of a 9.40 m long vertical and an 11.05 m long V-shaped ters were mounted at a test tube just after the inlet header.
horizontal tube. The other was a 44.43 m long helical coil of 0.018 m
i.d. By-pass pipes were built around both test tubes in order to keep
the total pressure drop constant in the tubes. The experiments were
carried out in such a way that the effects of all operating conditions
and inlet throttling on the stability could be determined. For this
purpose, the outlet steam quality at the inception of the DWO was
measured within about 2.5 percent accuracy for most of the tests
carried out in the first tube and within about 6 percent accuracy for
the tests carried out in the latter. The 306 data obtained and the 74
data given in [5-10] for long sodium or electrically heated once-
through steam generator tubes were correlated within about these
accuracies.
From the results of the experiments carried out in both test tubes,
it was concluded that the outlet steam quality at the inception of the
DWO is a function of pressure for a given inlet throttling whereas the
effects of all the other operating conditions on the steam quality are sodium
outlet

Contributed by the Heat Transfer Division for publication in the JOURNAL water make-up
OF HEAT TRANSFER. Manuscript received by the Heat Transfer Division Oc- system
tober 30, 1979. Fig. 1 Flow diagram of heal transfer loop

Journal of Heat Transfer AUGUST 1981, VOL. 103 / 485

Copyright © 1981 by ASME
Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 2.2 FTT. This 20.45 m long test tube of 0.01312 m i.d. and 0.0171 Both water/steam and sodium-side temperatures were measured
m o.d. was manufactured from stabilized ferritic steel, lOCrMoNb- with stainless-steel sheathed chromel-alumel thermocouples of 1 mm
Ni9.10, and consisted of a 9.40 m vertical and an 11.05 m V-shaped o.d. Maximum error in measuring these temperatures was 1.2 K. Both
horizontal tube. The sodium side surrounding the test tube was water/steam-side and sodium-side mass flows were measured with
constructed of 4 annuli. From the base, these were respectively 10.25, turbine flow meters, which had errors less than 1 percent. Water/
4,3, and 2 m long. The inner wall of each annulus was formed by the steam-side outlet pressure was measured with a dead-weight balance
test tube. The i.d. of the outer tube of each annulus was 0.0297 m. The manometer, which had a maximum error of 0.03 MN/m 2 .
annuli were connected by approximately 0.5 m long adiabatic circular 2.5 Experimental Procedure. The inception conditions of the
tubes of 0.0297 m i.d. A 0.80 m long piece at the inlet and a 0.40 m long DWO are, by definition, the operating conditions of the stable test
piece at the outlet of the test tube were not heated. The flow orien- run made just before the run in which the DWO appear. In order to
tation was upward in the water/steam side and downward in the so- generate the DWO, steam quality at the outlet of the test tube was
dium side. By means of two flow control valves mounted at the inlet increased by very small increments for most of the tests by increasing
of the test tube, the value of the inlet throttling coefficient was ad- only the sodium-side inlet temperature. For a few tests, the DWO were
justed between 54.7 and 382.9. This coefficient (i.e., total of the created by decreasing only water/steam-side pressure. The DWO were
pressure loss coefficients of the two flow control valves and a flow detected by observing the recordings of the water/steam side mass
meter between these valves) was based on the tube diameter. For given flow on the six-channel recorder. When the first diverging mass flow
openings of the valves, this coefficient was taken as the average of the oscillations appeared on the recorder, the operating conditions of the
values obtained in all tests carried out with the same openings. previous stable run were specified as the inception conditions of the
2.3 HCT. This 44.43 m long test tube of 0.018 m i.d. and 0.026 DWO. It was also ensured that the observed oscillations fulfilled the
m o.d. was manufactured from stainless steel-316 and was placed three conditions: (1) that they reach a limit cycle and thereafter stay
concentrically in another helical coil of 0.049 m i.d. The flow on the sustained (2) that they disappear when the sodium-side inlet tem-
sodium side was downward in the annulus between these coils, and perature is decreased or when the water/steam side pressure is in-
upward in the test tube. A 2.0 m long piece at the inlet and a 2.3 m long creased (3) that their period is of about the same order of magnitude
piece at the outlet of the test tube were not heated. The helix angle as the transit time of a fluid particle in the test tube.
of the test tube was 7.77 deg and the coil diameter was 1.50 m. For 41 For 82 percent of the tests carried out in F T T and 45 percent of the
percent of the tests, the value of inlet throttling was adjusted between tests carried out in HCT, the sodium-side inlet temperature was in-
223 and 665 by means of two flow control valves fitted at the inlet of creased in steps of 5 K or the water/steam side outlet pressure was
the test tube. For the calculation of the inlet throttling coefficient, decreased in steps of 0.25 MN/m 2 . For the rest of the tests, the in-
the same procedure was used as described for FTT. For the rest of the crement or the decrement in the aforesaid temperature and pressure
tests, these valves were not mounted and in this case the throttling was respectively in 10 K and 0.56 MN/m 2 steps.
of the flow was solely due to the turbine flow meter at the inlet of the Most of the tests were carried out systematically; in other words
test tube, and K = 15-27. the effects on the outlet steam quality at the inception of the DWO
2.4 Instrumentation. Both test tubes were heavily instru- of the water/steam side inlet subcooling, mass velocity, outlet pressure
mented. It is sufficient to mention here that both on the sodium side and inlet throttling, and the ratio of the mass flow in the sodium side
and water/steam side, inlet and outlet temperatures, outlet pressure, to that in the water/steam side were investigated at several pressure
inlet mass flow, and the temperature along the test tube at 11.05,15.05 levels (i.e., P = 14, 16 and 18 MN/m 2 ). For this purpose, the outlet
and 18.05 m axial positions in F T T and at 8.84, 13.59, 18.35, 23.10, steam quality at the inception of the DWO was measured for several
27.86, 32.62, 35, 37.37 and 39.75 m axial positions in HCT were mea- values of one of the aforesaid parameters, while the other parameters
sured with precalibrated instruments. Sodium-side temperatures in were kept constant and the measurements were repeated at several
F T T were also measured at 12 additional axial positions. All the pressure levels. At the DWO-runs, the water/steam side outlet tem-
measurements were collected with an on-line data acquisition system perature did not fluctuate and was constant. For these runs, this
and processed by a Hewlett-Packard-2116 B computer. Outlet tem- temperature was considerably higher than the saturation temperature
peratures both on the sodium side and on water/steam side, and the at system pressure. The above implied the presence of pure steam at
inlet temperature, inlet mass flow and outlet pressure on the water/ the outlet of a test tube. Therefore, the outlet steam quality (i.e.,
steam side, and the pressure drop in the by-pass pipe were simulta- thermodynamic steam quality) at the inception of the DWO was de-
neously recorded with a six-channel line recorder. termined with measured outlet temperature and pressure, using a
Steam Table.

^Nomenclature-
A = cross-sectional area of steam generator Pr = reduced outlet pressure (i.e., P divided superheated steam region (m)
tube (m2) by critical pressure) p = average density in a heat transfer region
a = transit time of a fluid particle (s) AP = pressure drop (N/m 2 ) (kg/m 3 )
a+ = dimensionless transit time Q = power at the inception of DWO (W)
D = coil diameter (m) Subscripts
Re = Reynolds number at the inception of
d = tube inside diameter (m) a = adiabatic length
DWO
G = mass velocity at the inception of DWO b = boiling region
r = latent heat of evaporation for water (J/
on water/steam side (kg/m 2 s) h = heated length
i' = acceleration of gravity (m/s2' kg) i = inlet condition
K = inlet throttling coefficient T = temperature at the inception of DWO on M = measured
h = enthalpy at the inception of DWO on water/steam side (K) n = sodium side
water/steam side (J/kg) t = temperature at the inception of DWO on P = predicted
AH = dimensionless inlet subcooling en- sodium side (K) p = preheat region
thalpy (1 - hi/hKt) AT su b = inlet subcooling on water/steam side S = beginning of the superheated steam re-
L = length (m) (K) gion
1 = length of a heat transfer region (m) W = mass flow at the inception of DWO s = superheated steam region
N = summation of pressure loss coefficients (kg/s) sat = liquid phase at the state of saturation
of bends along a straight tube or a serpen- X = outlet steam quality (i.e., thermody- t = total length or total power developed in
tine namic quality) at the inception of DWO, superheated steam region
P = outlet pressure at the inception of DWO based on thermal equilibrium v = vapor phase at the state of saturation
(N/m 2 ) y = axial coordinate; y = 0 at the begin of w = water/steam side

486 / VOL. 103, AUGUST 1981 Transactions of the ASME

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 Table 1 R a n g e of o p e r a t i n g c o n d i t i o n s for t h e t e s t s
Test P G Hi Number
tube MN/m 2 kg/m 2 s WJWW K kJ/kg X of data
7.6-13.5 439-1020 10.4-19.9 54.7-382.9 898-1257 1.37-1.68 16
13.6-14.5 486-910 8.8-19 54.7-196.6 768-1416 1.47-1.79 46
FTT 14.6-15.5 498-951 8.8-18.2 54.7-122.9 990-1440 1.58-1.83 26
15.6-16.5 493-890 10.1-18.1 54.7-106.9 980-1475 1.66-1.94 27
16.6-17.5 507-648 12.5-17.6 54.7-69.9 981-1471 1.78-1.95 22
17.6-19.0 526-683 10.6-16.9 54.7-107.9 826-1472 1.97-2.20 23
6.0-8.0 225-678 11.4-23.7 16-223 856-1083 1.15-1.49 8
8.0-10.0 187-731 8.6-22.6 16-665 845-1387 1.29-1.73 22
HCT 10.1-12.0 246-644 9.5-28.3 15-450 669-1421 1.37-1.83 29
12.1-14.0 260-726 9.9-29.5 15-450 720-1431 1.49-1.96 35
14.1-16.0 262-755 8.8-26.4 17-450 857-1470 1.58-2.04 28
16.1-19.1 294-768 11.7-22.2 15-223 862-1487 1.82-2.38 24

Demineralized water with an oxygen content of less than 15 ppb, -^~-—iM

a conductivity of less than 0.5 jiS/cm and a pH between 8.5 and 9 was x,o
ep
used during the tests. <--rN.Q s (y)
In total 2240 tests were carried out. In 160 tests with F T T and in Qt Data taken in FTT ; K = 54.73
146 tests with HCT, the DWO were observed. The range of operating P G iT. n sym.
2
MN/m2 kg/fn s /W„
conditions for the tests are summarized in Table 1, and are fully given
in [12]. The by-pass ratio varied between 0.1 and 13 for F T T tests and 1.51 14.0 577 101.3 17.9
between 2.2 and 12 for H C T tests. At the inception of the DWO, the ts(y)-Ts(y) afrftfc
sodium-side inlet temperature was only a few degrees higher than the ts-Ts Q| O 1.74 14.3 910 156.9 11.6
water/steam side outlet temperature.
1.72 14.0 544 152 8.9
The period of the observed mass flow oscillations was between 1.3
and 7 s in F T T and between 2.8 and 13.5 s in HCT. Peak-to-peak 1.63 14.3 572 122.6 11.9
amplitudes of these oscillations varied from 30 to 200 percent of the 0.1
mass flow recorded at the least stable test run. The water/steam side
outlet temperature did not oscillate. n/.s'yl:Ts'yl^>p.u.3f)
In the tests carried out in F T T with high inlet subcoolings, the
outlet pressure also oscillated. The peak-to-peak amplitude of these
oscillations was up to a few percent of the outlet pressure and their
period was equal to that of the mass flow oscillations. There was a
phase difference between these two oscillations. In the tests carried
out with low inlet subcoolings, no pressure oscillations were observed,
nor in the tests carried out with high inlet subcoolings and very small
by-pass ratios. The tests carried out in HCT were less susceptible to
pressure oscillations than those carried out in FTT. The generation
of pressure oscillations is explained as follows: The amount of highly 0.01
superheated steam produced in a test tube varies periodically. This 0.2 0.4 0,6 0.8
y/
superheated steam condenses completely in the cold water in the —» lc
upper header coming from the by-pass. This causes variation of the Fig. 2 Profiles of the dimensionless temperature difference between the
pressure, since the volume of the high-pressure part of the loop is sodium and water/steam side and the dimensionless power along the su-
small, as mentioned before. perheated steam region

During all the stable test runs for which X > 0, mass flow oscilla-
tions with small irregular periods and amplitudes were always present for given outlet pressure and inlet throttling. The effect of the outlet
in both test tubes. These irregular mass flow oscillations were assumed pressure on this length is minute and that of inlet throttling is con-
to be caused by the noise created by the flow in the rather small di- siderable but not of paramount importance. A similar result is also
ameter by-pass pipe. Because of these irregular mass flow oscillations, obtained from HCT tests with the value of Ci in equation (1) equal
inlet mass flow in the water/steam side was measured at one-second to 0.688. With the exception of 17 data taken mostly at low pressures
intervals for 10 seconds and the arithmetic average of the measure- when the irregular mass flow oscillations were comparatively large,
ments was considered. equation (1) predicts the length of the superheated steam region at
the inception of the DWO also within 12 percent accuracy for the data
3 Discussion of the Results obtained in HCT. The aforesaid length is considerable. For example
3.1 T h e Mechanism of the DWO. The observed DWO in both for P = 14 MN/m 2 , it is equal to about 35.6 m if K = 223 and to 33.3
test tubes are time delay oscillations. The length of the superheated m if K = 20 in HCT and to 14.3 m if K = 54.7 in FTT. Ct in equation
steam region and the transit time in this region are of vital importance (1) appears to be dependent on the tube length.
for the mechanism of the DWO. In order to show this, the following In all the tests at the inception of the DWO, the difference between
relation between the reduced outlet pressure, inlet throttling coeffi- the sodium side and water/steam side temperatures decreased ap-
cient and the ratio of the length of the superheated steam region to proximately exponentially along the superheated steam region, as
the total tube length is established from the data taken in F T T at the illustrated in Fig. 2 for the few test runs carried out in FTT. This is
inception of the DWO: due to the behavior of the specific heat of steam in the vicinity of the
saturation temperature at high pressures. The power developed along
h/Lt = Ci(l + K ) 0 0 2 8 (0.95 + 0.079P r ) (1) the superheated steam region is also given in Fig. 2, which shows that
where Ci = 0.626. The data reduction methods used throughout the in the last 50 percent of this region or beyond the about 15 m point
paper are given in [12-13]. Equation (1) predicts \JLt within about along the test tube almost no heat is transferred to the water/steam
12 percent accuracy from these data. So the length of the superheated side. During most of the tests only the sodium-side inlet temperature
steam region at the inception of the DWO is approximately constant was increased by small increments. This resulted in a gradual increase

Journal of Heat Transfer AUGUST 1981, VOL. 103 / 487

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 in the length of the superheated steam region and a decrease in the
(-^•TJTHWK)0'028
lengths of the preheat and boiling regions. Prom the above, the sig- 0. 900
nificance of the length of the superheated steam region for the -J L»_L 160 Data from F T T
a
mechariism of the DWO is obvious: In order to trigger the multiple a,* °? b*
regenerative feedback mechanism which generates the DWO in a test 0.880 - + K equal to 5 4 . 7
tube when outlet pressure and inlet throttling are kept constant, the O K greater than 54.7 and
length of the superheated steam region has to reach a constant value. X less t h u n 383
0.860 h
Thereafter the DWO appear and the multiple regenerative feedback " \. O
mechanism begins to operate. Inlet throttling delays this mechanism \ 0

in accordance with equation (1); i.e., the length of the superheated

steam region required to initiate DWO increases with increasing inlet
0.840 \°\P(X +
throttling coefficient.
V
o\ \
In order to determine the relation for the mechanism of the DWO, 0,820 S
the transit times of a fluid particle in all the different heat transfer
\ \
v\ + +1
+
regions in a steam generator tube have to be considered. The transit
0.800 h X 0 \
time of a fluid particle in a heat transfer region is
X \
x
a = Ip/G (2) 0 .780 \. + +x
X X + + °4 \ 5% line
This transit time is transformed into a nondimensional form as fol-
lows: 0.760

+ 2 1/2 1/2
a = tea /l) = (lpW) (3)
0.740
The term in parentheses in equation (3) is the reciprocal of the Proude
number and shows the ratio of gravity to inertia force.
With the data taken in FTT, the following empirical relation has 0.720
been established between the inlet throttling coefficient, outlet
pressure and the transit times at the inception of the DWO in the
preheat, boiling and superheated steam regions:
0,700 X N^
„ 5 % line ^ \
(l/qs+) + (l/ap+) < r
•• C 2 (l + K)-°- 0 2 8 (0.855 - 0.15 Pr) 0,680
1 1 1
i . 1 1
1

(l/a s +) + (l/a p +) + ( l / a 6 + ) 0. 2 0.4 0.6 0.8 1.0

(4) Fig. 3 Verification of the mechanism of the DWO in accordance with equation
(4)
where C<i = 1. Equation (4) is shown in Fig. 3 for the data from FTT.
This figure shows that the L.H.S. of equation (4) is practically con-
stant at the inception of the DWO for given outlet pressure and inlet 3.2 The Outlet Steam Quality at the Inception of the DWO.
throttling. Inlet throttling delays the DWO in accordance with This steam quality appears to be almost solely a function of outlet
equation (4) i.e., the length of the superheated steam region increases pressure for a given inlet throttling in both test tubes. The effects of
with increasing inlet throttling coefficient, as explained in [13]. The all the other operating conditions on the aforesaid steam quality seem
use of different two-phase flow models for the evaluation of at in to be of secondary importance. In the works reported in [6, 14], a
equation (4) did not practically alter the value of C2. similar conclusion has been drawn.
Equation (4) has only been checked with 16 data taken in HCT for The reason why the outlet steam quality at the inception of the •
the pressure range 13.9 < P(MN/m 2 ) < 14.3. For these data G = DWO is constant, can be explained as follows: Average density in the
262-675 kg/m 2 s; AT a u b = 20.4-168 K; X = 1.49-1.96. If the value of superheated steam region at the inception of the DWO was solved
C2 in equation (4) is taken as 0.921, this equation fits the data within from equation (4), and was correlated within about 12 percent accu-
6 percent accuracy. racy for 129 data taken in FTT, for which K = 54.7, with the equation
below
From the above, it is concluded that the observed DWO are
"time-delay oscillations," that the inlet throttling delays the mech- : 3.62 + 82.76 Pr (5)
anism of the DWO and that C 2 in equation (4) seems to be weak
function of (Lt/d). Assuming that the density is equal to that calculated using the average
Equation (4) gives the mechanism of the DWO in FTT and in HCT temperature in the superheated steam region, the outlet steam quality
if these tubes operate in the once-through mode. It follows that in at the inception of the DWO was determined within 0.4 percent ac-
order to initiate the multiple regenerative feedback mechanism which curacy by the use of equation (5) and a Steam Table, as given
generates the DWO in these tubes, the condition expressed by below
equation (4) has to be fulfilled.
X = 0.721-0.764 ln[(l - Pr)Pr0S (6)
For a steam generator tube for which X < 1, a.,+ in equation (4)
must not be considered. In this case, this equation was checked with With four exceptions, equation (6) fits the aforesaid data within 8
the 8 data reported in [14] for a very short electrically heated tube of percent accuracy. The R.M.S. error for all the data is 4.2 percent. From
Lt/d = 153. These data were taken for 4.1 < P(MN/m 2 ) < 8.3 and this, it is concluded that outlet steam quality (or temperature) at the
0.087 < AH < 0.131. The value of C 2 in equation (4) for the data were inception of the DWO is practically a function of pressure for given
0.44 with an accuracy of 10 percent. inlet throttling and geometry, and that the effects on the inception
Provided that the R.H.S. of equation (4) is determined for a rea- of the DWO of other operating conditions are of secondary impor-
sonable amount of data taken in tubes of different geometries, the tance. For the length of the superheated steam region, the same
equation may serve a stability model since it includes the mechanism conclusion was also drawn. It follows from the above that the super-
of the DWO. Both equation (1) and equation (4) were established to heated steam region dominates the mechanism of the DWO in both
show some phenomena related to the DWO in two once-through steam the test tubes and equation (4) seems to give this mechanism well.
generator tubes. Thus they are not for general use, and their appli-
cation should be restricted to the conditions similar to those used in 4 C o r r e l a t i o n of D a t a
the present experiments. The present data and the data available from the literature are

488 / VOL. 103, AUGUST 1981 Transactions of the ASME

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 correlated using the results given in Sections 3.1 and 3.2 in two dif- where
ferent ways for practical applications. For this purpose, dimensional
C 4 = 0.07-0.017.10- 3 Lh/d for Lh/d < 4118 (7.1)
analysis is used. The dimensionless numbers which fully characterize
the dynamics of two-phase flow are numerous and it is practically C4 = 0 forLA/d>4118 (7.2)
impossible to take all of them into account when correlating the data The R.M.S. error for all the 306 data is 5.44 percent. Xc in equation
[15]. Furthermore, as indicated in [16], the use of the Buckingham and (7) is the outlet steam quality at the inception of the DWO given in
Rayleigh methods of dimensional analysis will give dimensionless [11], and the effect on this steam quality of inlet throttling could not
groups, but their physical significance is not evident. The use of dif- be taken into account for Lhld < 4118 due to lack of data. The cor-
ferential equations allows one to interpret physically the dimen- relation reported in [11] predicts the power at the inception of the
sionless groups thus derived, but provides no information on the DWO within 6.5 percent accuracy from the 88 once-through steam
fundamental mechanism of the process [16]. Thus, the experiment generator data considered in [11]. It has to be noted here that it is
is the only safe basis for the selection of the dimensional variables for sufficient to know only two operating conditions, i.e., P and T; to
the correlation of the DWO data. As concluded from [17], the logical determine equation (7) for given (Lh/d) and K.
grouping of these variables into dimensionless groups can then be 4.2 Second Correlation. Since 5.44 percent R.M.S. error is not
accepted without hesitation. In this case, however, the physical sig- considered sufficiently accurate for once-through steam generator
nificance of these dimensionless groups may not be evident [16]. applications, the present data and the 74 data reported in [5-10] were
The present experiments clearly demonstrate that the most sig- recorrelated. The evaluation of these 74 data is given in [12-13]. The
nificant dimensional variables for the inception of the DWO in a range of geometries and operating conditions for these data are
once-through steam generator tube are total or heated tube length, summarized below: Geometry: circular straight tubes, a coil, a ser-
outlet temperature, outlet pressure and pressure drop due to inlet pentine and a test tube consisting of a 10 m long vertical and an 8.84
throttling. For the present experiments L/, s Lt . These variables are m long, V-shaped horizontal tube; arrangement of the test tube: tubes
transformed into dimensionless numbers such as Lid, the ratio of tube of large capacity steam generators; tubes around which a by-pass is
length to tube diameter, X, outlet steam quality, Pr, reduced pressure built; two or three tubes arranged in parallel; heating conditions:
and K, inlet throttling coefficient. The importance of the first men- electrically or sodium heated; L, = 10-223.3 m; d = 7.86-29.9 mm;
tioned three dimensionless numbers on the mechanism of the DWO D = 0.8 m; A T s u b = 23.9-245.9 K; G = 118-1365 kg/m 2 s; X =
is due to fact that they characterize the length of the superheated 1.01-1.69; P = 4.3-17.3 MN/m 2 ; K = 0-2882.
steam region and the transit time of a fluid particle in this region. As For the correlation of the data, the effect on the outlet steam quality
demonstrated in the preceding sections, these are of vital importance at the inception of the DWO of each operating condition, namely, P,
for the mechanism of the DWO in a once-through steam generator Wn, Ww, AT su b and G, and inlet throttling was determined separately
tube. The importance of the inlet throttling coefficient on this as expressed by equations (8.1) to (8.6) with the aid of the very accu-
mechanism is due to distribution of the pressure losses along the tube: rate data obtained in F T T . For this purpose, these operating condi-
The total of these pressure losses is constant and the DWO are tions were used in dimensionless form, i.e., Pr, Wn/Wu:, AH and Re„,
time-delay oscillations. The experiments of Quandt [14] carried out respectively. Thereafter a correlation was established for all the data
in a steam generator tube in which no superheated steam was pro- obtained in FTT. By comparing this correlation with the data from
duced show that the inlet subcooling also influences the inception H C T and the data reported in [5-10], the effect of inlet throttling
conditions of the DWO. This is logical, since subcooling characterizes coefficient on the tube length and the effect of the friction factor on
the length of the boiling (or preheat) region in this tube. The inlet the steam quality at the inception of the DWO were determined by
subcooling can be nondimensionalized as AH, the dimensionless inlet expressions given by equations (8.7) to (8.9). The term in the de-
subcooling enthalpy. nominator of equation (8.9) is practically identical to Blasius's friction
It follows from the above that it seems sufficient to consider five factor for straight tubes and the first two terms in the numerator are
dimensionless numbers, i.e., Lid, X, Pr, K and AH for the correlation the friction factor for helical coils [18]. The last term in the numerator
of the DWO data. Therefore, a correlation (i.e., equation (7)) was first represents loss coefficients of the bends along a straight tube or a
established using the above dimensionless numbers. This equation serpentine. The final correlation obtained is
applies both to a steam generator tube in which no superheated steam
is produced and to a once-through steam generator tube. This cor- X = AiA2A3A4A!,Ae (8)
relation should be very useful for many practical applications since
where
it includes a few operating conditions. It is also an accurate correla-
tion.
Ai = 0.131-0.135 In [(1 - Pr)Pr02] (8.1)
For some special applications in engineering practice, improve-
ments in the accuracy are considered very useful for predicting the A2 = 1-0.005 WJWW (8.2)
inception conditions of the DWO, especially for a once-through steam
generator tube. This can be achieved if the second-order variables are A 3 = 1 + 0.13 AH (8.3)
also considered for the correlation of the DWO data. These variables,
0125
in their nondimensional form, are Wn/Ww, the ratio of sodium-side A4 = Re„ (8.4)
mass flow to water/steam side mass flow and Re„, the Reynolds
number for flow in the superheated steam region. The first dimen- As = ll+KI(Lt/d)™->]cs (8.5)
sionless number influences the length of the superheated steam region
C 5 = 0.11 for F T T (8.6)
and the latter accounts for the variation of water/steam-side mass
flow. Thus a second correlation (i.e., equation (8)) was established C 6 = 0.175-4.2.10- 5 (L ( /d) forL,/d<4167 (8.7)
considering the aforesaid five dimensionless numbers, i.e., Lid, X,
Pr, AH, K, and WnIWm and Re„. The correlation thus obtained is C6 = 0 for LJd > 4167 (8.8)
indeed more accurate than the first correlation, but applies only to
once-through steam generator tubes. A6 = |[A 7 + 0.029 (d/D)0-5 + (NdlLt)1M]/A7\°™ (8.9)
4.1 First Correlation. The previous correlation of the author
[11] predicts the power at the inception of the DWO from the present A 7 = 0.304 Re„-°- 25 (8.10)
data well, i.e., within 8 percent accuracy for 90 percent of the time if
the effect of inlet throttling is taken into account using the formula If the steam quality at the outlet of a once-through steam generator
given below tube exceeds the outlet steam quality predicted by equation (8), the
DWO appear in the tube. The power at the inception of the DWO is
X = Xc(l+K)c< (7) expressed by the equation

Journal of Heat Transfer AUGUST 1981, VOL 103 / 489

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 100 (Q P -Q M )/Q M

+ <$>
a <!> 6 % (me
_4>—J= -
PTtYw "
a w*.-^
-2

fc - 6
-6
6 % line
-8
160 Data from FTT; L , / d = 1559 220 data
RMS error = 4.01
+ K equal to Si.7
© K greater than 54.7 and less than 383 Symbol Li / d
RMS Error = 2.04
-10
• 0
A
1272
1535
<!> 10150

t
Y
X
1279
2529
2397
10 12 H 16 20
2468
Pressure ( MN/m I +
10 12 U 20
Fig. 4 Errors in predicting power in accordance with equation (9) for the data
Pressure ( MN/m*)
taken in FTT
Fig. 5 Errors in predicting power In accordance with equation (9) for the data
taken in HCT and data of [5-101

GA(hsal-hi + rX) (9)

5 Summary and Conclusions
where X is given by equation (8). The results of comparison of the • DWO were investigated in two sodium heated once-through
data from FTT, and data from HCT and the data reported in [5-10] steam generator tubes of Lt/d = 1559 and 2468. The observed DWO
with equations (8) and (9) are given in Figs. 4 and 5, respectively. are time-delay oscillations. For the mechanism of these oscillations,
Equation (9) predicts the power at the inception of the DWO within an empirical relation has been established. The length of the super-
7.5 percent accuracy for 98 percent of the time. The R.M.S. error for heated steam region and the transit time of a fluid particle in this
all the 380 data is 3.33 percent. region are of vital importance for the aforementioned mechanism.
As mentioned, 82 percent of the data obtained in F T T were very • The 306 data obtained and the 74 data reported in the literature
accurate. For these data, equation (9) predicts the power at the in- have been correlated to predict the power at the inception of the DWO
ception of the DWO within 2.5 percent, as shown in Fig. 4. It is then within 7.5 percent accuracy for 98 percent of the time. The R.M.S.
concluded from equations (8) to (8.4) that the effect of the operating error for all the 380 data is 3.33 percent. The established correlation
conditions on the steam quality at the inception of the DWO is small is recommended for sodium and electrically heated once-through
with the exception of pressure for long once-through steam generator steam generator tubes of 1272 < Lt/d < 10150 for the following range
tubes. The data analyzed in Fig. 5 and equation (6) also confirm this of operating conditions: P = 4.3-19.1 MN/m 2 ; G - 118-1365 kg/m 2
conclusion. s; A T s u b = 3.2-245.9 K; X = 1.01-2.38; K = 0-2882.
The effect on X of inlet throttling seems to decrease with increasing • This study shows that the inception conditions of the DWO can
(Ltld) and disappears beyond Lt/d > 4167; in other words this effect be predicted with simple equations, thus eliminating the use of
stays within the experimental accuracy for Lt/d > 4167. For the data time-consuming complicated stability correlations in the form of a
reported in [9-10] from a very long tube of Lt/d = 10150, the inlet computer program for a wide range of operation conditions and
throttling coefficient varies between 237 and 2882. These data are geometries, as considered here. Further verification of the work pre-
shown in Fig. 5. sented is recommended.
The second correlation is recommended for sodium and electrically
heated once-through steam generator tubes of 1272 < Lt/d < 10150 Acknowledgment
for the following range of operating conditions: P = 4.3-19.1 MN/m 2 ; This study is a government sponsored work carried out by TNO.
G = 118-1365 kg/m 2 s; AT s u b = 3.2-245.9 K; X = 1.01-2.38; K = The author wishes to express his gratitude to Messrs. K. A. War-
0-2882. With the exception of coil, D - « in A 6 . schauer, A. R. Braun and M. L. G. van Gasselt for their encouragement
As discussed in [11], a stability model in the form of a long computer during the preparation of this work. The loop was operated by Mr.
program can not, in general, predict the power at the inception of the A. J. C. Spierings.
DWO better than to about 10 percent. It is known that the accuracy
of stability models presented so far in the literature can be much less References
than 10 percent. It must be remembered, however, that an empirical 1 Boure, J. A., Bergles, A. E., and Tong, L. S., "Review of Two-Phase Flow
correlation lacks generality and is, as a rule, invalid outside its range. Instability," Nuclear Engineering and Design, Vol. 25,1973, pp. 165-192.
Nevertheless, if the mechanism of the DWO described before is con- 2 Proceedings of the Symposium of Two-Phase Flow Dynamics, EUR-
4288e, Vol. I and Vol. II, 1969.
sidered, it seems that the present correlations are not much suscep- 3 Fallows, T., Hitchcook, J. A., Jones, R. C, Lis, J., and Northover, E. W.,
tible to the aforesaid invalidity. "A Study of Oscillatory Instabilities in the Parallel Channels of A High Pressure

490 / VOL. 103, AUGUST 1981 Transactions of the ASME

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 Once-Through Boiler Rig," Proceedings of the International Conference-Boiler Two-Phase Group Meeting, Erlangen, 31st May-4th June 1976.
Dynamics and Control in Nuclear Power Stations, The British Nuclear Energy 10 Deam, R., Personal communication.
Society, 1973, pp. 14.1-14.8. 11 Unal, H. C , "Correlations for the Determination of the Inception Con-
4 Waszink, R. P., and Efferding, L. E., "Hydrodynamic Stability and ditions of Density-Wave Oscillations for Forced and Natural Circulation Steam
Thermal Performance Test of a 1-MWt Sodium-Heated Once-Through Steam Generator Tubes," ASME JOURNAL OF H E A T T R A N S F E R , Vol. 102,1980, pp.
Generator Model," ASME Journal of Engineering for Power, Vol. 96, 1974, 14-19.
pp. 189-200. 12 Unal, H. C , "Some Aspects of Two-Phase Flow, Heat Transfer and
5 Unal, H. C., "An Investigation of the Inception Conditions of Dynamic Dynamic Instabilities in Medium and High Pressure Steam Generators," Ph.D.
Instabilities in Sodium Heated Steam Generator Pipes," in Two-Phase Flows Thesis submitted to Technological University of Delft, The Netherlands (to
and Heat Transfer, edited by S. Kakac and T. N. Veziroglu, Vol. 3, pp. be published on March 18,1981).
1425-1443, Hemisphere Publishing, 1977. 13 Unal, H. C , "Density-Wave Oscillations in Sodium Heated Once
6 Unal, H. C., Van Gasselt, M. L. G., and Ludwig, P. W. P. H., "Dynamic Through Steam Generator Tubes," CTI-TNO Report, Ref. Nr. 79-06976, July
Instabilities in Tubes of a Large Capacity, Straight-Tube, Once-Through So- 1979.
dium Heated Steam Generator," International Journal of Heat and Mass 14 Quandt, E. R., "Analysis and Measurements of Flow Oscillations,"
Transfer, Vol. 20,1977, pp. 1389-1399. Chemical Engineering Progress Symposium Series, Vol. 57, Nr. 32,1961, pp.
7 Brasz, J., "Hydrodynamic Instability and Performance Degradation 111-126.
15 Ishii, M., and Jones, O. C , Jr., "Derivation and Application of Scaling
of a Straight-Tube Sodium Heated Steam Generator," Proceeding of Con-
Criteria for Two-Phase Plows," in Two-Phase Flows and Heat Transfer, edited
densed Papers of the Meeting—Two-Phase Flow and Heat Transfer Sympo-
by S. Kakac and F. Mayinger, Vol. 1, pp. 163-185, Hemisphere Publishing
sium- Workshop, Ft. Lauderdale-Fla, Oct. 18-20,1976, edited by T. N. Vezir-
Corporation, 1977.
oglu, University of Miami, 1976, pp. 183-189.
16 Knudsen, J. G., and Katz, D. L., Fluid Dynamics and Heat Transfer,
8 Sano, A., Kanamori, A., and Tsuchiya, T., "Operating Experiences with McGraw Hill, New York, 1958, p. 143.
1 MW Steam Generator," Paper Presented at IAEA Study Group Meeting on 17 McAdams, W. H., Heat Transmission, McGraw Hill, New York, 1954,
Steam Generators for LMFBR's, Bensberg, October 14-17,1974. p. 135.
9 Deam, R., and Murray, J., "The Prediction of Dynamic Stability Limits 18 Ito, H., "Friction Factors for Turbulent Flow in Curved Pipes," ASME
in Once-Through Boilers Using DYMEL," Paper Presented at European Journal of Basic Engineering, June 1959, pp. 123-134.

HTD BEST PAPER OF THE YEAR

Nominations are now being accepted for the Heat Transfer Division Best Paper of 1980. The award is given annually
to the authors of the best paper published by an ASME member in the ASME archival literature in the year immediately
preceding the award. The award is open to all papers which have at least one author who is an ASME member.
Any person may nominate a paper for this award. The nomination shall be made in the form of a letter formally nom-
inating the paper and stating why the paper should be considered for this distinct honor. Nominations for the award
should be addressed to the chairman of the Heat Transfer Division Honors and Awards Committee, c/o Senior Technical
Editor, Journal of Heat Transfer. The nomination deadline is December 31; however, nominations may be made at any
time. All candidate papers will be considered by the Heat Transfer Division Honors and Awards Committee.

Journal of Heat Transfer AUGUST 1981, VOL. 103 / 491

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 Application of Mixing Length Theory
S. Levy
J. M. Healzer to Waiy Turbulent Liquid—Gas
S. Levy, Inc.,
Campbell, Calif.
Interface1
A fully developed and adiabatic two-phase annular model with liquid entrainment is de-
rived for flow in a pipe with negligible gravity effects. The model is based upon applica-
tion of the single phase mixing length theory to a wavy liquid-gas interface. The model
subdivides the flow cross section into three regions: a liquid film, a gas core of constant
density, and a transition wavy layer between them. The combination of a constant veloci-
ty and a density varying exponentially with distance from the wall is employed in the
transition layer. This approach plus appropriate logarithmic velocity distributions in the
liquid film and gas core make it possible to specify the two-phase pressure drop, volume
fraction, wave velocity, and thickness of the liquid film and transition layer. The liquid
entrainment is obtained in terms of the exponent of the density profile in the transition
layer, and interface stability considerations are used to express this entrainment parame-
ter semiempirically in terms of an apparent Weber number and density ratio. Compari-
sons of the model are made with air-water and steam-water test data, and the results gen-
erally are satisfactory over a wide range of conditions and for all the important character-
istics of this flow pattern.

Introduction Core. The central core extends from the edge of the transition
Wavy liquid-gas interfaces are found in many practical two-phase layer to the center of the flow passage. The core velocity, uc, varies
flow applications involving annular flow patterns. A very large number from the velocity, Ut, at the edge of the core, Yc, to a maximum value
of models have been proposed in the literature to deal with such an in the center of the flow passage. The density in the core, pc, is taken
interface. One large family of solutions range from the classical hy- as constant.
drodynamic treatment of surface, gravity and wind waves [1] and The advantage to dividing the annular flow pattern into the above
extends to the most recent application of the nonlinear Korteweg-de regions is that both the core and the liquid film can be treated using
Vries equation to describe small but finite amplitude longwaves [2]. single phase mixing length turbulent flow concepts. The transition
Another large group of solutions relies upon empirically based for- layer can be modeled with a relatively simple extension of these same
mulations to describe the drag coefficient of the waves or the increased concepts.
friction at the interface [3-6]. In this paper, it is proposed to deal with For steady-state turbulent flow, neglecting gravitational forces, the
the turbulent structure of the flow at the liquid-gas interface and to momentum equation reduces to the familiar Reynolds expression for
extend the accepted single phase mixing length turbulent theory to shear stress, T. In a fully developed flow of variable density, the shear
describe it. stress can be written as [7]
The characteristics of a turbulent liquid-gas interface are depicted dud(pu)
in Fig. 1(a) for an annular flow pattern in a pipe of radius R. It consists lul„ (1)
dy dy
of a liquid film at the channel wall with a central core of gas containing
liquid droplets. The liquid film and gas core are separated by a tran- where the mixing lengths l„ and lp , are assumed to be functions of
sition layer of liquid waves. For the present model, the liquid-gas y. If they are taken to be such that
arrangement is simplified as shown in Fig. 1(6). At any position z
along the direction of flow, the elements of the flow pattern are de- lulpu = FHy) 1 - (2)
scribed as follows: R,
Liquid Film. The liquid film thickness is Yf. The velocity in the there results for T = TW (1 - y/R)
liquid film is uf and varies from zero at the wall to Uf at the edge of dud(pu)
the liquid film. The density in the film is taken as liquid density, TW = FHy) (3)
Pi- dy dy
Transition Layer. The transition layer extends from the edge where rw is the wall shear stress.
of the liquid film to the edge of the core, Yt. The velocity in the For a constant density fluid, equation (3) reduces to the well-known
transition layer, ut, varies from the velocity at the edge of the liquid Prandtl expression for wall shear stress if we assume that [8]
film to that at the edge of the core. In this model, the density in the
FHy) = k*y* (4)
transition layer is assumed to decrease exponentially across the layer,
starting at the liquid density, PL, at the liquid film and decreasing to where k is the mixing length constant.
the core density, p c , at the outer edge of the core. As shown by the later For application to the present model, equation (3) is solved for each
application of the mixing length theory, the transition layer velocity of the three separate flow regions depicted in Fig. 1(6) (liquid film
is found to be a constant and is taken as the liquid wave velocity, extending to Yf, transition layer between Yf and Yt, and the core
Uw. spanning from Yt to the center of the pipe). The solutions will yield
not only the pressure drop and the volume fractions occupied by the
gas and liquid, but they will specify the amount of liquid entrained
1
This work was sponsored by the Electric Power Research Institute (EPRI) beyond the liquid film. This quantity is most important in many
under Research Project 1380-1 and their support is gratefully acknowledged. practical applications such as the prediction of critical heat flux. Also
Contributed by the Heat Transfer Division for publication in the JOURNAL
OF HEAT TRANSFER. Manuscript received by the Heat Transfer Division July predictions will be developed for the liquid and gas velocity and
2,1980. density distributions, for the liquid and transition layer thicknesses,

492 / VOL. 103, AUGUST 1981 Transactions of the ASfWE

Copyright © 1981 by ASME
Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 and for the transition layer constant velocity (taken to be the wave
velocity).
Before employing equation (3) to obtain such solutions, it is useful
to first review typical approaches which have been used in the past
and discuss their modeling assumptions for an annular flow pat-
tern.

Previous Interface Assumptions and Solutions

Many different interface assumptions have been employed previ-
WAVE
ously. They tend to fall into three classes:
• discontinuous density interface and continuous velocity inter-
face. GAS CORE
• discontinuous density and velocity interface.
• empirical interface relations.
The solutions have dealt with varying types of two or three fluid
systems and include: FLOW

Two Fluid Systems. A liquid film and a gas core with no liquid Fig. 1(a) Annular flow pattern
drops; or a liquid film and a gas core with entrained liquid drop-
lets.
Three Fluid Systems. A liquid film, entrained liquid drops in the
gas, and a gas core; or a liquid film, a transition layer, and a gas core
with entrained liquid droplets.
In the simplest case of two completely separate fluids and a dis-
continuous density interface, the liquid and gas velocity are assumed
equal at the interface. If we assume that F2(y) = k2y2 as in single
phase turbulent flow, there results

: +f//+ ysYf
"L1
H - (5)
u G + = - l n — +£//+• Y
k Yf ' /J
where Y/ is the liquid film thickness and u and Uf+ are shear stress
+

velocities.
Fig. 1(b) Annular flow pattern model

7- "''""N't (6)

The subscripts G and L are used to indicate that the gas and liquid
properties and density are used in equations (5) and (6). culate Ua+ values well below 5.2 Since a value of shear stress velocity
Equations (5) have been integrated by many investigators to predict below 5 is associated with laminar flow (8), the assumption of equal
two-phase annular flow a?id in several instances they have been liquid and gas velocity at the interface implies the hardly practical
modified to take into account the effects of gravity [9, 10, 11]. The possibility of a laminar gas condition (UG + < 5) atop a turbulent
solutions have been found to do an adequate job on the liquid film but liquid layer (Uf+ > 14). The gas interface will, in fact, be turbulent
to do poorly in predicting the gaseous pressure drop or interface and the gaseous pressure drop should be expected to exceed the pre-
friction factor. This is not too surprising since the gaseous shear stress dictions from equation (5).
velocity at the interface, Ua+ is equal to One way to correct this situation is to assume not only a discon-

UG* (7)

1
Since in most experiments performed PG/PL « 1 one would cal- For example, for PO/PL = 0.0025 and C//+ = 15, UG+ = 0.75.

.Nomenclature.
Ce = liquid concentration beyond base liquid Uf = velocity at edge of liquid film p = density
film Ut = velocity at edge of core a = surface tension
D = pipe diameter Uw = wave velocity T = shear stress
E = entrainment cross flow rate u = local velocity TW = wall shear stress
F = function describing mixing length u+ = nondimensional shear velocity (if+ = Subscripts
fis = interface friction factor U/VTW/P)
G — mass flow per unit pipe area c = core
Y/ = thickness of base liquid film
K = function defined by equation (31) e = entrained liquid
Yt = distance to outermost edge of transition
ke = cross flow entrainment coefficient f = liquid film
layer
k = mixing length constant G =gas
Yw = distance to center of gravity of transi-
ks — equivalent surface roughness L = liquid
tion layer
lu = velocity mixing length TPF = two-phase friction
y = distance from wall
lPu = mass rate mixing length t = transition layer
y+ = nondimensional distance (y + = y
NRe = Reynolds number for gas flow only wf = just beyond liquid film
V^Jp/v) wt = just short of core
(GG DIIXG) z = distance along pipe
P = pressure /3 = density function exponent Superscripts
R = pipe radius ix = viscosity — = average over

Journal of Heat Transfer AUGUST 1981, VOL 103 / 493

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 tinuous density but also a discontinuous velocity at the interface. In creasing with distance from the wall. The distribution just described
this case is that of annular flow with liquid entrainment, except that it includes
a transition layer within which the density was allowed to vary con-
+ +
UG = -ln~+UG , UG+»Uf+J-PG_ (8) tinuously from the liquid to the gas core value. To the authors'
Yf PL knowledge, such an approach with a decreasing u+ in the transition
It is interesting to note that if as in single phase flow one assumes layer has not been considered in the past. In order to more specifically
Ua+ = 5.5, equation (8) becomes with ks = 3.32 Yf illustrate the preceding comment let us assume beyond the liquid
film
1 y
uG+ = - l n — + 8.5 (9)
« Ks

With the constant 8.5, equation (9) corresponds to single phase flow (13)
in a rough pipe with a sand roughness ks [8]; and, in two-phase flow
with Ua+ = 5.5 the gas flow tends to behave as in single phase flow
in a rough pipe with a sand roughness ks equivalent to 3.32 times the where /? is a positive exponent and k is the so-called mixing length
liquid thickness. This is not very different from the Wallis empirical constant. Equation (11) becomes:
approach of using four times the liquid film thickness [5]. It is im- du+ 2
) _(YfY 1 1 2
portant to note, however, that the velocity discontinuity at the in- — +-u + «2 (14)
terface is very substantial since dy/Yfj
\dy/Y„ \y\ k2 4
and
PL_
(10) du-> .^P dy/Yf
Uf~Uf Pa (15)
V " + 2 + (2/fe/?)2 2 y/Yf
and it is doubtful that such conditions will exist and preserve the as-
sumed density discontinuity. Equation (15) can be integrated with u+ Uwf+ just beyond the in
For that reason, several investigators have recognized the need to terface, 3
break the liquid film into two regions. A continuous liquid base layer
is employed which is in contact with the wall and it supports an in-
u+ + V M + 2 + (2/fep1)2 y_fi (16)
termittent, wavy layer of variable density. The model developed by Uwf+ + VU^TWkffl Yf!
Moeck and Stachiewicz [12] is typical of such derivations and it em- Let us now consider two types of density profile. For /? > 1 and pref-
ploys a two-dimensional film and a one-dimensional droplet laden erably /? » 1 we have a rapid decrease of the density beyond the in-
gas core. Others [13] have chosen to break the core into the gas itself terface, which is typical of annular flow with liquid entrainment. If
and a separate liquid droplet field. In all formulations of this type, u+ is kept turbulent or u + at least exceeds 5, u+ 2 » (2/fe]6)2 with k
several empirical correlations are necessary to describe the interface. taken close to the accepted value of 0.4, equation (16) simplifies ap-
They include relations for the base liquid film thickness, the interface proximately to
friction factor, the droplet size and a droplet deposition coefficient.
Of particular interest are the semi-empirical works of Wallis [5], Levy
(17)
[4], and Moeck [6], which showed that the interface friction factor I V \Yf)
depends only upon the liquid film thickness. In this final class of so-
lutions for annular flow, the predictions can be expected to be more Using the density distribution of equation (13), the corresponding
accurate but they cannot deal with the details of the flow and density velocity ratio is for the positive root,
structures and must concentrate upon properties averaged over large
portions of the flow cross section.
It is the purpose of this investigation to deal with the detailed flow
or for the negative root,
u
(t) (18)

distributions and to limit the number of empirical correlations nec-

essary to predict liquid entrainment.
u
=1 (19)
Proposed Interface Solution
It is proposed in this section to examine the velocity and density If one utilizes the positive root, it would mean that u+ increases as
discontinuity at the interface. By introducing equation (6) into (y/Yf)P/2 in the transition layer and that the velocity slip between
equation (3) and nondimensionalizing, there results liquid film and gas core would be greater than that calculated from
/ du+ \2 equations (5) and (8), Also, examination of the velocity data of Cohen
plpi
(11) [14] obtained for turbulent air flowing over a horizontal liquid film
\dylYf) F2 \ dy
dylYf reveals that u+ is decreasing in the transition layer. 4 On this basis,
du+
dy/Yf V
+ IT?
F2
1
4
d I n plpi\2
dy/Yf
(12)
the annular flow with liquid entrainment can only satisfy equation
(12) with a minus sign and there results

u+ + V ^ + 2 ' + (2/A/3)2
The first major difference from single phase flow is that equation
(20)
(12) can admit to the two roots, + and - in two-phase flow. The reason Uwf+ + sJUwf+2+(2lkP)2 Uuf+ \Y,I V PL
is that in single phase flow u+ and the fluid velocity always increase
away from the wall. In two phase flow, the density p is decreasing away Under such conditions, it is worthwhile to note that the fluid velocity
from the liquid interface. We can, in fact, visualize two types of dis- is relatively constant in the transition layer and equal to a velocity
tribution. The first one is such that p decreases slowly beyond the Uu> beyond just the interface.
interface and u and u+ are going up beyond the liquid interface. This It is also worthwhile to note that the same result is obtained without
type of distribution is characteristic of bubbly flow. The second type assuming the form of F2 as long as
of distribution would be one in which the density decreases rapidly
in a transition layer from the liquid interface to a gas core and the
3
shear stress velocity u+ would be decreasing in the transition layer. In this definition, liquid properties are used because of closeness to interface,
Within the gas core, the density can be assumed to be relatively con- which accounts for subscript /.
4
stant and is equal to pc and the shear stress velocity distribution In reference [14], at a given y+ the measured values of u+ in the air layer
fall below the values obtained from equation (30) which apply to the liquid
within the core would be characteristic of turbulent flow and in- film.

494 / VOL. 103, AUGUST 1981 Transactions of the ASME

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 Id In plpL\2 It is important to note that in this model the liquid in the transition
Yf*
u^
l

^*i ted
Equation (11) then leads to the only acceptable solution in the tran-
zone or wave zone is assumed to be entrained. This is a reasonable
approach since this liquid is moving at velocities well in excess of those
in the liquid film and close to the gas velocity. The total gas flow per
sition layer, or unit pipe area Ga is equal to the sum of the gas flowing in the core and
«+ (P/PL)1/2 = constant, 0 > 1 the transition zone. It is equal to

Let us next consider the value of Uw. At the edge of the gas core the
velocity is equal to Uw and the corresponding shear stress velocity
Uwt+ is
so that
Ga =
Mfr'Mti PL
Pa
-pa
(28)

^ = = \[B±UW,+ (21)
i V pc \ PLI
-VTW/PL » PL PL\ + Gt (pL PL
If the velocity Uw at the edge of the transition layer is taken to be Pe Gc \Pa Pt
(29)
that of the liquid at the edge of the liquid film or Uf (Uwt+ = Uf+
VPC/PL)> o n e could again be faced with the anomaly of Uwt+ falling
GG
\BL _ A + .G< lpL
below 5 because pcI.PL « 1 and the gas core would be starting with a \Pc Gc \Pt
laminar interface. Because this does not make sense physically Uw Equations (23, 25, 27) and (29) mean that the ratio GJGG is a
must be taken greater than Uf and the transition layer with a con- function of the density ratios PL/PG, PL/PC, the liquid film thickness
tinuous density distribution and a constant velocity Uw requires a step ratio Yf/R and the parameter /3. Finally, the solution in the liquid film
jump in velocity at the edge of the liquid film to achieve a turbulent proceeds as in single phase flow. It is subdivided into three regions:
interface. a liquid laminar region extending to yL+ = 5, a liquid transition region
The interface equations and mixing length turbulence behavior are, extending from yL+ = 5 to yL+ = 30, and a turbulent liquid region
therefore, suggesting presence of a transition layer of constant velocity extending beyond y t + = 30 [8]. The velocity UL+ is given by [8]
with a rapid decrease in density from liquid to gas core and with a step
jump in velocity over that of liquid film. This, in fact, is the repre- "L+ = yL+ yL+ ^ 5
+
sentation that one would get of the wavy layer above the liquid if av- uL = -3.05 + 5 1 n y L + 5syL+s30
(30)
eraged over long periods of time. It has been found that disturbance uL+ = 2.5 In yL+ + 5.5 37, + g 30
waves can travel above the liquid film at velocities much greater than
where y/, + is a nondimensional distance equal to y t \prjp~il PLI PL
those in the liquid film and that they protrude several times above
with PL being the liquid viscosity. The liquid film mass flow rate per
the liquid film thickness. The equations developed herein attempt
unit area of pipe, Gf is obtained by integrating equation (30) so
to describe such a layer but averaged over long periods of time. that
Solutions for A n n u l a r F l o w with Liquid Entrainment
As shown in Pig. 1, let us assume that the flow cross section can be
broken down into three regions: a liquid film of thickness Yf, a tran-
sition layer extending from Yf to Yt, and a gas core from Yt to the pipe
GflPL

lot
R+2
K (Y>+, R+) = 2
I yf+/R+ y+\ p + , [y
1
R+IPL
— u+d (3D

PL
center with a constant density pc. As previously noted, in the transi-
tion layer, the velocity ut is constant so that where

ut UW = UW Yf<y<Yt (22) K(Yf+, R+) =-R+Yf+2 - - Yf+S, for Y/+ < 5

The mass flow rate in the transition region per unit area of pipe, Gt K(Yf+, R+) = 12.51A+ - 10.45 - 8.05A+ Yf+ + 2.775 Yf+2

-r
is obtained from equations (22) and (13) and + 5R+Yf+ In Yf+ - 2.5 Yf+ In Yf+, for 5 < Yf+ < 30
Gt/PL K{Yf , R+) = ZR+Yf+ - 63.9R+ - 2.125 Yf+ - 1.25 Yf+2 In Yf+
+ 2
= 1Vm+4
V i + 2.5R+Yf+ In Y/+ + 573.21,

—m
TW/P, for Yf+ > 30
Yf with
i- (23)
„. Ry/rw/pL
In the gas core, the velocity distribution is obtained from a single R+ = PL
P-L
phase type relation or
The preceding equations make it possible to predict annular flow

>fcH
and the mass flow rate in the gas core per unit pipe area Gc is given
(24) with liquid entrainment. If the parameters k, Uwt+, /?, UL+, PL/PC,
PL/PG are specified, the solution can proceed as follows:
1 Assume a value of Yf/R, calculate Yt/Yf from equation (26) and
by
Gt/Gc from equation (23) and (25). The entrained ratio Ge/Ga follows
2
Gclpc 2 from equation (29).
U,+ 1 +-• 0.75 • (25)
VruJpZ k R 4\fl/ 2 Yt. 2 The ratio Ga/Gc is calculated from equation (28) and the term
Gclpchwlpc obtained from equation (25). An interface friction factor
Also, because of equation (13), there results
results such that
Pc = lYA-e (26)
PL \Yfl
GG/PG (Ga/Gc) (GJpc) V pa
Because the velocity u is relatively constant in the transition layer (32)
\/TW/PG \/TW/PC lis
we can assume that the liquid and gas velocities are equal in that layer.
The same assumption can be made in the gas core where the density where fis is the interface friction factor based upon the superficial gas
was taken as constant. The total amount of liquid entrainment per velocity in the pipe, Gal pa- Equation (32) makes it possible to cal-
unit pipe area, Ge is equal to the sum of liquid entrained in the gas culate the wall shear stress TW, i.e. the corresponding two-phase
core and liquid in the transition layer. It is equal to frictional pressure drop (dp/dz)TPP for a given value of Ga or

Ge M- -^)
L \PG Pel
+ GM-
\Pa "I Ptl PL
Pa
- pa
(27)
dp|
dz J TPF
2TU,

R
1 (Ga)2
(33)

Journal of Heat Transfer AUGUST 1981, VOL. 103 / 495

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 3 The average fluid density is obtained by integrating the density © STEAM-WATER AT 50 BARS IN UPFLOW FROM WURTZ
over each of the three regions in the pipe [17]. X AIR-WATER AT 1 BAR IN HORIZONTAL FLOW FROM
4 Finally, for a given value of the total gas flow rate GG, the ZANELLI -HANRATTY
+ AIR-WATER AT 1 BAR IN HORIZONTAL FLOW FROM
non-dimensional liquid film thickness Y/ + is obtained from MIJA-WOODMANSEE-HAN RATTY
B AIR-WATER AT 1 BAR IN DOWN FLOW FROM TELLES
•J—pL
V PL YfRGa^/Tw/pG
V 'f-
in, R HL GG/PG
^ (34)

and the liquid film flow rate Gf is obtained from equation (31).
5 The calculations are repeated with different values of Yf/R and
one can generate the desired values of liquid entrainment, friction and
head losses for any specified value of gas and liquid flow rate and core
density.
Specifications of Parameters k, Ut + , Uwt+
Before predictions can be made, it is necessary to select appropriate
values for the parameters k,Ut +, a n d Uw. W e shall t a k e k = 0.4 5 a s
in single phase flow. T h i s is a reasonable assumption since a t t h e limit
of PL - Pa,k m u s t a s s u m e t h a t value. Similarly Ut + will be taken as
Ut+ = 5.5 to again obtain the correct single phase profile at the limit
of PL = PG- The parameter Uw corresponds to the constant velocity
to be employed in the transition layer. As pointed out before, it is also
taken to be the wave velocity in this transition zone. It is specified as
follows: ***
1 The position Yw of the center of mass of the transition layer is
found by writing
0 1 2 3 4 5 6
MEASURED WAVE VELOCITY , m / SEC
(35)
Yf U Fig. 2 Measured and predicted wave velocities
where pt is the mean density in the transition layer.
2 The shear stress velocity at the edge of the liquid film is Uf+ and bility relations have been developed by several authors [18] and they
the shear stress velocity at the edge of the transition layer is Ut+. By yield
assuming a straight line between U/+ and Ut + , the shear stress ve-
o- (PL + Pc)
locity at the position Yw can be calculated and it is taken to be the (37)
wave shear stress velocity Uw+. There results PL Pc (uc ~ Uf)2
where A is the interface wave length, tr the surface tension and uc and
Uf+ - 5.
Uw + = VJyfTjft = Uf+
F" 1
fit"') (36) Uf are the average core and film velocities. One can visualize the
possibility of employing equation (37) together with previously de-
fined relations to specify /? or pc. In this study, equation (37) will be
employed in a much more narrow way, that is to define nondimen-
and the corresponding wave velocity Uw is obtained.
sional parameters upon which /3 might depend. If one assumes that
The validity of equation (36) can be checked against experimental
uc » Uf, equation (37) becomes with uc = Gc/pc
data for the wave velocity for which Yt, Yf, and the pressure drop have
been reported. From the ratio YJYf one can calculate a value of ff £c_
from equation (26). The values ofptlpL and Uf+ can also be computed 1 +
which specify the wave velocity Uw. All the measured and predicted PL. Pc (38)
wave velocities are shown in Figure 2 and the correspondence between Pc_ PL \GC
tests (15, 3,16) and predictions is good. PL
Correlation for Parameter $ As suggested by Tippets [18], the wavelength A of the disturbance can
The parameter |8 is most important to the proposed solutions. In be taken proportional to the mixing length. It is also expected that
the original study performed for E P R I [17], constant values of the the wave length A decreases as Yt/Yf increases and goes to infinity
parameter /3 were first employed together with and without liquid when Yt = Yf or PL - PG, and across the transition layer, one can
present in the gas core. The solutions revealed that: write
• the model had many of the correct characteristics observed in
tests
• that the simplified assumption of no liquid in the core had a
minor impact upon the prediction
A d k Yf

so that with the approximation Gc/p,

/I" ' GG/PG there results
(39)

• that the parameter /3 which describes overall liquid entrainment

needed to vary with gas flow rate and liquid surface tension.
It was then surmised that the value of ft is controlled by interface Pc/pL
I + Pel PL
GH
Xl
Yf/R
[IPG]2
•
\pLj
\ -
<* PL
kRGa2
(40)
stability consideration. In other words, for a given gas and liquid flow
rate, the amount of liquid entrained and its distribution in the core Equation (40) contains a new non-dimensional parameter (o/kR)
should be such that the liquid-core interface is stable. Interface sta- (PL/GG2)(Yt/Yf - 1). It was pointed out after equation (29) that

5
Plots of uc+ versus y + from some velocity measurements in the gas core have (41)
given a straight line with a slope varying from 0.4 to 0.1 (See Chemical Engi- GG \PL R PL
neering, Vol. 19, pp. 665-682) and this has been interpreted that k in equation
(24) should vary accordingly. It should be noted that equation (24) used here and compatibility between equations (40) and (41) suggests
expresses uc + in terms oty/Yt and that Yt depends upon /?, the gas and liquid
density and the wall shear stress, thus allowing for the variation of the slope '° PL
observed in test data when plotted in terms of u c + . '-^HYA
[kR G \Y, G
2
(42)

496 / VOL. 103, AUGUST 1981 Transactions of the ASME

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 Physically, it is expected that /3 should decrease with a/kR PL/GQ2
(Yt/Yf ~ 1).
PREDICTIONS
The exact form of equation (42) was developed empirically and it
.0 = 3.15 , | > G " 1 - 5 1 K « ' m 3 , G G » 1 3 5 K,/m 2 s«c
was found that with Yt/Yf = (PL/PG)11^ the relation
, 3 r 4 , p G = 1.41 K9/m3,GG=:74Kj/m2i.c

via--
(3 = 1 + .
2<T

kDGG2
PL

gave satisfactory agreement with the entrainment data of Gill and

(43) £ = 5 , p G = 1.28

vS=8,f>G=1.2
K9/m3,GG=45Kg/m2nc

K 9 / m 3 , G G = 18 K 9 / m 2 l e t

Hewitt [19].
It should be pointed out that in developing equation (43) the liquid COMPLETE ENTRAINMENT
entrained Ge in the proposed analytical solutions with pc = pa was
set equal to the measured values for simplification purposes. As
pointed out in [17] making pe ^ p c , while keeping the total amount
of entrained liquid constant, does not have a major effect upon the
results and the simplification is warranted for ease of calculations.
It should also be noted that no great effort was expended to optimize
the form of equation (43) in view of some of the approximations in-
troduced; yet, the proposed model gives superior results to previously
published correlations.
G I L L - H E W I T T DATA

B pG=>1.51K9/m3,GG=10?Kj/m2
Comparison to Test Results
A f> G =1.41K 9 /m3,G G =.79 K a / m 2
Of particular interest is liquid entrainment because it is the pa-
0 f>G = 1 . 2 8 K , / m 3 , G G » 4 8 K ) / m 2
rameter least satisfactorily correlated to-date. This parameter is
+ PG"1-2 Kg'm3,GG=l6K9/tr
sometimes expressed as a cross flow rate perpendicular to the flow
direction. Such an entrainment cross flow rate E is proportional to
the liquid concentration Ce beyond the liquid film withE = keCe and
Ce = Ge/(Ga/pL + Gelpi,). Two types of correlations have been
proposed to-date for the entrainment cross flow rate. Paliev and 0 0.2 0.4 0.6 0.8 1.0
GAS WEIGHT FRACTIONS
Filippovich [20] employed the parameter (P-LGG/PG^)2 (PG/PL)- This
Fig. 3 Comparison of entrainment model to Gill-Hewitt data
grouping was shown by Styrikovich, et al. [21] not to be acceptable
for their steam-water tests. The second approach developed at Har-
PREDICTIONS
well relies upon the grouping rwYf/a which gives the ratio of the
^G =174 Kg/m2lic
disruptive shear forces on the liquid film to the stabilizing surface G

^GG=1 4 0 K , / m 2 t . c
tension forces. An alternate proposed by Whalley, et al. [22] is to re-
place the liquid film thickness Yf with the effective roughness height ,.G G = 70Kg/m2itc
) K g / m 2 s»c
ks. Such correlations have been found to be preferable but they still
leave much to be desired because data scatter is still substantial. COMPLETE ENTRAINMENT

The results obtained for the entrainment rate Ge from the solutions
proposed herein together with equation (43) are shown in Figs. 3 to
7. Figure 3 shows the predictions together with the data of Gill and
Hewitt [19]. The correspondence is very good with the deviation at
most being 50 percent and generally being within 20 percent. As can
be observed from Fig. 3, the ratio GB/GG increases with gas flow rate
GG and approaches the homogeneous curve at very high gas mass
rates. At very high quality the data tend to fall below predictions. For
such points, Y/ + falls well below 30 and equation (38) is not valid for WHALLEY-HEW ITT-HUTCHINSON
Yf+ < 30. An increase in /8 is necessary below Y/ + = 30 and the liquid O GG > 1 4 0 K 9 / m 2 i t c
entrainment can be expected to decrease when the liquid film is not X 70 < G r . < 1 4 0 K « / m 2
fully turbulent [17].
+ GG < 70Kg/m2iec
Let us next consider the prediction of the thickness of the transition
AIR/WATER
layer or height of the waves. In line with the approximation developed
pL = 1000 K g / m 3
to correlate liquid entrainment one sets pc = pa so that
PG= 3.3 K 9 / m 3
Yt = [PL W" (44)
Yf \PGI

For air-water mixtures, PL/PG » 1, P tends to be large (between 4 and

8) so that the density decreases rapidly in the transition layer. The
0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
base liquid film Yf is, therefore, not very different from the measured
GAS WEIGHT FRACTION,X
mean liquid thickness and one can use the measured mean liquid
Fig. 4 Comparison of entrainment model to Whalley data
thickness for Y/ to compare the measured ratio of Yt/Yf to the pre-
dicted one from equation (44). For the Harwell data, most measure- of 2.1 and less would result for Yt/Yf which compare favorably with
ments have been taken with air mass flow rates between 50 and 100 the experimental results of 2.0 to 3.30. The slow increase in Yt/Yf
kg/m 2 s and the corresponding values of /S are between 3.5 and 5.0. noted in the Hanratty experiments reinforces the view that an in-
Equation (44) gives ratios of Yt/Y; of 3.8 and 6.8 which agree very well creasing j8 parameter in the region of Uf+ below 30 would be more
with the data given in reference [12] of 4.0 to 6.5. In the case of Han- appropriate. In the case of Telles [15] data, Yt/Yf was reported to be
ratty and co-workers the air flow was set constant at 90 cm/s. This between 2.5 and 2.6 versus the prediction of 3 and less depending upon
corresponds to a value offiof 9.3 for a gas density of 1.2 kg/m 3 . Many whether Y/ + is above or below 30. Finally, Charvonia developed a plot
of the runs were, however, performed with very thin films so that Y/ + of Yt/Yf with values ranging from 2.8 to 3.4. The corresponding model
< 30 and the corresponding /3 value would be in excess of 9.3. Values predictions are 3.7 and less and they again bracket most of the data.

Journal of Heat Transfer AUGUST 1981, VOL. 103 / 497

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 PREDICTIONS

PREDICTIONS
,/GG = 145 Kj/m2itc G0 = 780 Kg/m2stc
Gr, = 2 3 3 Kg/m2sec
/G G = 60 Kg / m 2 stc

COMPLETE ENTRAINMENT

WHAL LEY-HEWITT-HUTCHINSON
t t t < 60Kg/m2ttc
x 60 < G G < 1 4 5 K g / m 2 i e c
WURTZ DATA FOR I 0 M M TUBE
O 145< G G < 2 2 0 Kg/m2,tc
+ GG < 109 K i ) / m 2 stc

TRICHLOP.OETHANE/AIR x 233 < GG < 109 K g / m 2 s t c

2 3 3
PL = 1300 K g / m 3 O 760< GG < Kjlm^sit

PG = 3.8 Kg/m3 a G G > 780 K g / m 2 s e c

J = 29X103N/m

STEAM/WATER AT 7 0 BAR

0.25 0.3 0.4 0.5 0.6 0.7 0.8

GAS WEIGHT FRACTION,X
0.2 0.3 0.4 0.5 0.6
Fig. 5 Comparison of entrapment model to Whalley data GAS WEIGHT FRACTION,X

Fig. 6 Comparison of entrapment model to Wurtz data

In the case of high pressure stream-water mixtures, the parameter

/3 is much lower. For example at 70 bars /3 tends to vary between 1.3
and 2.0 and the density decreases slowly in the transition layer,
making it difficult to differentiate between the base liquid film PREDICTIONS
thickness Yf, the mean liquid thickness and the edge of the transition 'GG«543K,/m2„c
layer, Yt. This is illustrated in Pig. 8 where the data of Subbotin [23] -G G = 1 6 5 K , ( m 2 i i c
are reproduced for the minimum film and maximum film thickness. -GG- 77Kg/tn2[,c
The ratio of maximum to minimum film is shown on the same plot.
Prom equation (43), one can calculate approximate values of /3 and
the predicted ratio of Yt/Yf is shown. At high quality the film thick-
ness may be small enough to possibly fall below Y/+ = 30 and the ratio
Yt/Yf decreases.
The trend and absolute values are seen to be satisfactory. Similar
acceptable comparison was obtained with the data of Wiirtz. In
Wiirtz's case YJYf ranged from 6.5 to 17 and tended to slightly exceed
the model predictions. This is not unexpected because as pointed out
before the correlation for /3 was based upon measured liquid en-
trainmenta and should lead to Yf values in excess of the minimum or WUFTZ DATA FOR 20 MM TU»E
base liquid film thickness. '<• G G < 7 7 K g / m 2 ««c
Let us next consider the wave velocity. The model predicts the ratio « 77 <GG<165Kg/m2<ec

of Uw to Galpa for given values of fi and gas weight fraction. By again © 165<GG<343 KgWs.c

using the simplifications of pc = pa, values of the wave velocity Uw a GG>543K,/m2i«c

were calculated to match a few of the test results of Whalley, Hewitt

and Hutchinson and Wiirtz. The results are shown in Table 1. For S T E A M / WATER AT 7 0 BAR

air-water tests the model is 10 to 20 percent high while for steam-water

the model is 30 to 50 percent high. Part of this discrepancy might be
explained by realizing that a much more accurate value of Uw would
have resulted from

Uw
I Y,/R
Yf/R
pu\X ly-
R
(45)
0.2
GAS WEIGHT
0.3 0.4
FRACTION,X
0.5 0.7

cYt/R (i-zL^ Fig. 7 Comparison of entrainment model to Wurtz data

JYf/R P\ Rj R
Equation (45) would give smaller values of Uw than those obtained The correlations of Wallis, Moeck and Levy have been shown to be
by employing the velocity at Yw from equations (35) and (36) and this satisfactory over a wide range of test conditions and it is proposed to
would be particularly true for low /3 values. compare the present model with such empirical correlations. In Fig.
In the case of frictional pressure drop, several good and approximate 9 the ratios of interface friction factor to single gas phase friction factor
correlations have been developed for the interfacial friction factor. predicted for /3 = 3.15 and 4 is plotted. The results correspond to air

498 / VOL. 103, AUGUST 1981 Transactions of the ASME

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 Table 1 Prediction of Wave Velocity
Measured Uw Predicted Uw
Ga (kg/m 2 s) GL (kg/m 2 s) pL (kg/m 3 ) pa (kg/m 3 ) a (N/m) (m/s) (m/s)
142.4 15.9 1000 3.26 73.1 X 10- 3 3.74 3.54
126.6 31.8 1000 3.29 73.1 X 10- 3 3.57 4.45
78.9 79.9 1000 3.26 73.1 X 10~ 3 2.90 3.82
62.7 94.8 1000 3.30 73.1 X 10" 3 2.60 3.38
100.0 400.0 742 36.00 17.7 X 10" 3 1.80 2.30
150.0 350.0 742 36.00 17.7 X 10" 3 2.20 3.20
200.0 300.0 742 36.00 17.7 X 10- 3 2.50 3.60
250.0 250.0 742 36.00 17.7 X 10- 3 2.80 4.10
300.0 200.0 742 36.00 17.7 X 10~ 3 3.10 4.60
350.0 150.0 742 36.00 17.7 X 10" 3 3.20 4.70
150.0 600.0 742 36.00 17.7 X 10- 3 2.60 4.00
225.0 525.0 742 36.00 17.7 X 10~ 3 3.20 4.80
300.0 450.0 742 36.00 17.7 X 10" 3 3.60 5.30
375.0 375.0 742 36.00 17.7 X 10- 3 3.90 6.10
450.0 300.0 742 36.00 17.7 X 10- 3 4.30 6.50

40
-
PRESENT M0DEL-^_^^
FROM DATA
1
PREDICTED 20 fL=1000 Kg/m3
WITH /B f>G = 1.51 K g / m 3
^ V
GG=135 K 5 / m 2 s e t
1.0 /' '1
£ 10 /f h/ / / /1/
E
01" 0.6
6 // /
A / /

- ^ ' ^ ^ - ~ P R E S E N T MODEL

=
f-1
^ ===x!^-^-7 PL=1000 Kg/m3
PG = 1.28 K g / m 3
WALLIS^
G G =74 Kg/m 2 s*c
. ^ MOECK
0.1 _
—
1
0.001 0.003 0.006 0.01 0.03 0.06 0.1
LIQUID FILM THICKNESS/ PIPE RADIUS,Yy/R

Fig. 9 Comparison of present model to empirical friction correlations for

air-water
0.03
STEAM/WATER AT 7X106 N / m 2
G = 1 0 0 0 Kg/m 2 sec

0.01 PRESENT MODE

0 0.2 0.4 0.6 0.8
GAS WEIGHT FRACTION

Fig. 8 Wave height prediction for steam-water

78=1.7

2
flow rates of 135 and 74 kg/m s by Gill and Hewitt in a pipe of 0.0318
m in diameter. The proposed model is seen to bracket the available
correlations and even slightly exceeds the Wallis correlation at high
flow rates. A similar comparison is given in Fig. 10 for steam-water
mixtures at 70 bars. Plots are given for /3 values of 1.31 and 1.70 cor-
responding to mass flow rates of 90 and 542 kg/m 2 s in a 20 mm tube.
0.001 0.003 0.006 0.01 0.03 0.06 0.1
Again the analyses bracket the previously proposed correlations and LIQUID FILM THICKNESS/ PIPE RA0IUS Y//R
would give good agreement with the steam-water data of Subbotin
which were found to match the Wallis correlation. Fig. 10 Comparison of present model to empirical friction correlations for
steam-water
An interesting facet of Figs. 9 and 10 is that they show that as the
fluid density ratio decreases the parameter /3 decreases thus bringing As a final check of the model let us compare the predicted film
the interface frictional curves together. Another interesting facet thicknesses to measured values. Plots are given in Fig. 11 for air-water
would be that the present model would predict increased friction mixture while Figure 12 deals with steam-water mixtures at 70 bars.
factor with fluids of reduced surface tension, a result noted by several Here again the assumption is made that Yf corresponds to liquid film
two-phase flow investigators. thickness. The comparisons are satisfactory especially when one

Journal of Heat Transfer AUGUST 1981, VOL. 103 / 499

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 + G = 60Kj/m2».c WURTZ STEAM/WATER DATA AT 70 BAR
6 5 = 1 5 7 K9/ m 2 l i e e G = 500 K)/m2i«c
+ 0 = 750K,/m2l.c
X G = 1000 Kj/ m 2, ( t

4 G = 2000 K . / m 2 i « c

m2 i t c

E
^T 0.0 20

Fig. 12 Comparison of liquid film model to steam-water data, in 20 mm

0.3 0.4 0.5 0.6 0.7 O.J 0.9 tube
GAS WEIGHT FRACTIONS

Fig. 11 Comparison of liquid film model to air-water data

6 Moeck, E. 0., "Annular Dispensed Two-Phase Flow and Critical Heat
realizes t h a t t h e simplification h a s b e e n m a d e t h a t t h e core c o n t a i n s Flux." Atomic Energy of Canada Limited-3656,1970.
no liquid droplets. 7 Levy, S., "Prediction of Two-Phase Pressure Drop and Density Distri-
I n s u m m a r y , t h e p r o p o s e d m o d e l is seen t o give good a g r e e m e n t bution From Mixing Length Theory," ASME Journal of Heat Transfer, Vol.
w i t h available d a t a over a w i d e r a n g e of p a r a m e t e r s a n d fluids. 85, 1963, p. 1973.
8 Schlichting, H., Boundary Layer Theory, McGraw-Hill, New York, I960,
pp. 488, 508, 523.
9 Anderson, G. H., and Mantzouranis, B. G., "Two-Phase (Gas Liquid) Flow
Simplified Model Phenomena—I. Pressure Drop and Hold-Up for Two-Phase in Vertical Tubes,"
A simplified version of t h e p r e c e d i n g m o d e l is d e v e l o p e d in refer- Chemical Engineering Science, Vol. 12,1960, p. 109.
ence [17]. I t relies u p o n simplified a n d s e m i e m p i r i c a l expressions for 10 Dukler, A. E., "Fluid Mechanics and Heat Transfer in Vertical Falling
Film Systems," Chemical Engineering Progress Symposium Service, Vol. 56,
t h e t w o - p h a s e friction factor a n d for t h e e n t r a i n e d flow r a t e in t e r m s
No. 30,1960.
of t h e p a r a m e t e r /3. T h e simplified m o d e l gives good a g r e e m e n t w i t h 11 Hewitt, G. F., "Analysis of Annular Two-Phase Flow. Applications of
the data. the Dukier Analysis to Vertical Upward Flow in a Tube," Atomic Energy Re-
search Establishment, R-3680,1960.
Conclusions 12 Moeck, E. O., and Stachiewicz, J. W., "A Droplet Interchange Model
for Annular-Disperse, Two-Phase Flow," International Journal of Heat Mass
1 A new m o d e l h a s b e e n d e v e l o p e d t o p r e d i c t t w o - p h a s e a n n u l a r Transfer, Vol. 15,1972, p. 637.
flow w i t h l i q u i d e n t r a i n m e n t . T h e m o d e l r e q u i r e s a single e m p i r i c a l 13 Saito, T., "Multi-Fluid Modeling of Two-Phase Flow and Heat Transfer;
e n t r a i n m e n t p a r a m e t e r /? w h i c h is o b t a i n e d from t e s t d a t a . T h i s p a - Application to CHF Prediction for BWR Conditions," Ph.D. Thesis, University
r a m e t e r /3 defines t h e density variation above t h e base liquid film a n d of Wisconsin, 1977.
14 Cohen, L. S., Interaction Between Turbulent Air and Flowing Liquid
is specified in t e r m s of n o n d i m e n s i o n a l g r o u p i n g i n f e r r e d from in- Film, Ph.D. Thesis, University of Illinois, 1964.
terface s t a b i l i t y c o n s i d e r a t i o n s . 15 Rae, J., "A Model for Interface Waves in Two-Phase Flow," Atomic
2 T h e m o d e l e x h i b i t s good a g r e e m e n t w i t h t e s t d a t a w h e n c o m - Energy Research Establishment, Harwell T P 611, HTFS RS 156.
p a r e d t o o t h e r a v a i l a b l e c o r r e l a t i o n s a n d is c a p a b l e of h a n d l i n g a 16 Wurtz, J., "An Experimental and Theoretical Investigation of Annular
Steam-Water Flow in Tubes and Annuli at 30 to 90 Bar," Riso National Lab-
variety of fluids including variation in surface tension, fluid p r e s s u r e , oratory, Denmark, 1978.
and tube diameter. 17 Levy, S., and Healzer, J. M., "Prediction of Annular Liquid-Gas Flow
3 T h e m o d e l gives good a g r e e m e n t w i t h t e s t d a t a for m a n y o t h e r With Entrainment—Cocurrent Vertical Pipe Flow With No Gravity," EPRI
NP-1409, May 1980.
p a r a m e t e r s , i n c l u d i n g l i q u i d film t h i c k n e s s , friction p r e s s u r e d r o p ,
18 Tippets, F. E., "Critical Heat Fluxes and Flow Patterns in High Pressure
interface wave h e i g h t a n d velocity. Boiling Water Flows," ASME Paper No. WA-162,1962.
4 B e c a u s e t h e m o d e l is b a s e d u p o n d e t a i l e d velocity a n d d e n s i t y 19 Gill, L. E., and Hewitt, G. F., "Further Data on the Upwards Annular
d i s t r i b u t i o n s it could b e a p p l i e d t o m o r e c o m p l e x g e o m e t r i e s . Flow of Air-water Mixtures," Atomic Energy Research Establishment, R-3935,
1962.
20 Paleev, I. I., and Filipovich, B. S., "Phenomena of Liquid Transfer in
References Two-Phase Dispersed Annular Flow," International Journal of Heat Mass
1 Lamb, H., Hydrodynamics, Dover Publications, New York, 1932. Transfer, Vol. 9,1966.
2 Banerje, S., and Chan, A. M. C , "Separated Flow Models—I. Analysis 21 Styrikovich, M. A., et al. Wave Pattern and Liquid Film Burnout with
of the Averaged and Local Instantaneous Formulations," International Journal Cocurrent Vapor Flow, International Heat Transfer Conference, Toronto,
of Multiple Flow, Vol. 6, No. 1-2,1980. Canada, 1978.
3 Miya, M., Woodmanseee, D., and Hanratty, T., "A Model for Roll Waves 22 Whalley, P. B., Hewitt, G. F., and Hutchinson, P., "Experimental Wave
in Gas-Liquid Flow," Chemical Engineering Science, Vol. 26,1971. and Entrainment Measurements in Vertical Annular Two-Phase Flow," Atomic
4 Levy, S., "Prediction of Two-Phase Annular Flow with Liquid Entrain- Energy Research Establishment, R-7521,1973.
ment," International Journal of Heat Mass Transfer, Vol 9,1966. 23 Subbotin, V. I., et al., Integrated Investigation Into Hydrodynamic
5 Wallis, G. B., "Annular Two-Phase Flow, Part 1: A Simple Theory," Characteristics of Annular-Dispersed Steam-Liquid Flows, International Heat
ASME Journal of Basic Engineering, 1970, pp. 59-82. Transfer Conference, Toronto, Canada, 1978.

500 / VOL. 103, AUGUST 1981 Transactions of the ASME

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 A Numerical Solution to Dropwise
Ewaporation
J. J. Rizza
Professor of Engineering, A numerical solution is presented for spray evaporation on the surface of a spray evapora-
Mechanical Engineering Faculty, tor. Excess temperatures below the Leidenfrost transition temperature are considered
California State University, Fullerton,
Fullerton, Calif. 92634
along with the nonisothermal wall condition. The combination of a high thermal conduc-
Mem. ASME tivity and low heat capacity for the evaporator wall material, as well as the time-depen-
dent boundary conditions, dictates a quasi-steady-state solution to this evaporator phe-
nomenon. A numerical solution to the conduction equation and boundary conditions is
given. A dimensionless evaporation time is computed, and from this a flooding index is
developed. The flooding index is shown to be related exponentially to the droplet diame-
ter and wetting ratio. The functional correlation between the heat transfer coefficient for
the spray and the wetting ratio is verified from previously published experimental data.
Introduction
How liquid droplets impact and evaporate on warm surfaces is of the impacted droplet. The energy immediately available for evapo-
interest in a number of areas of engineering; e.g., spray evaporators, ration is equal to the change in internal energy of the surface directly
cooling of turbine blades and the cooling of hot metal ingots, to name below the impacted droplet. This internal energy change is limited
a few. The evaporation of impacted droplets has received considerable by the surface temperature change; i.e., from its initial value to the
attention in the literature for the so-called Leidenfrost phenomenon; evaporant saturation temperature. The internal energy change can
i.e., for large excess temperatures, the difference between the surface be represented by
temperature and the evaporant saturation temperature [1-3]. The
prediction of the transition Leidenfrost temperature is of current AE = cwPwAweAT (1)
importance in fast breeder-liquid metal cooled reactors and in cryo- where cw and pw are the specific heat and density of the wall material.
genic systems. The evaporation of impacted droplets below the The effective heat transfer surface area, A^, will in general be greater
Leidenfrost transition temperature also has been studied for a number than the contact surface area between the droplet and the wall. The
of years, although not to the same degrees as the Leidenfrost phe- wall thickness is denoted by £ and the local temperature depression
nomenon [4-6]. by AT. The energy required to evaporate the droplet is related to the
The configuration most often considered in experimental studies phase change energy and the sensible heat. If the droplet is given to
for the surface heat transfer is one in which there is little or no thermal be at its saturation temperature at impact, the required energy is given
resistance within the wall of the spray evaporator or evaporating by
surface; i.e., the isothermal wall condition. When the droplet diameter
before impact is of the same order of magnitude as the evaporator wall AE = p/Ad b hfg (2)
thickness, the thermal resistance between the transport fluid and the
where b and Ad are the impacted droplet thickness and surface con-
evaporating droplet can not be neglected; i.e., the nonisothermal wall
tact area. Upon combining equations (1) and (2), an expression for
condition is applicable. This phenomenon is characterized by a sig-
the local temperature depression is obtained.
nificant reduction in wall temperature below the impacted droplet.
Temperature reductions up to 30° C are reported in [5] for droplet AT = (pfhfs/cw pw) (Ad/Aw) (bl£) (3)
sizes as large as 600 fim. Similar surface temperature depressions are
reported by [1] for a nonisothermal surface made of pyrex glass. When If the volume of the droplet before impact is equated to the volume
the evaporating droplet is in this configuration, it is not only the after impact, a relationship results for the disk-shaped impacted
evaporation rate from the droplet surface that determines the rate droplet thickness
of heat transfer, but is the combined transfer from the transport fluid
b = 2d/3 a2 (4)
through the wall to the evaporating droplet which must be consid-
ered. where a represents the droplet spreading ratio. This is defined as the
The present investigation is a study of the evaporation of impacted ratio of the droplet diameter after impact to the diameter before
droplets on the surface of a spray evaporator at excess temperatures impact.
below the Leidenfrost transition temperature and for the noniso- If the effective heat transfer surface area is taken to be roughly four
thermal wall condition. The droplet evaporation times are considered, times the contact area and if equations (3) and (4) are combined, the
and from them an evaporator flooding index is developed. The following expression can be written for the local temperature de-
flooding of an evaporator surface is of importance in the overall per- pression. This area is thermally equivalent to a much larger area since
formance of the spray evaporator relative to the maintenance of the temperature is inversely related to distance.
dropwise evaporation over film wise evaporation and to the effective
AT = (pf hfe/cw pw) (d/e) (1/6 a2) (5)
utilization of the spray evaporant; i.e., reduction of evaporant
runoff. A typical value for the first ratio in the above expression for a com-
bination of water and a number of metals is approximately 800/1. A
Analysis representative value for the spreading ratio is about 4/1 [4,5], If these
Local Temperature Depressions. It is possible to estimate values are used in equation (5), the following results.
approximately the combination of parametric values that will produce
AT = 8.5 (dl£) °C (6)
the nonisothermal wall condition on the surface of a spray evaporator
by considering a simplified evaporation model for the impacted The above expression demonstrates the effect of droplet diameter on
droplet. When a liquid droplet impacts a warm surface, it begins to the wall temperature distribution. If the droplet diameter before
evaporate on contact by drawing heat from the surface directly below impact is small relative to the wall thickness, then the local temper-
ature depressions are also small; i.e., the isothermal wall condition.
Contributed by the Heat Transfer Division for publication in the JOURNAL
OF HEAT TRANSFER. Manuscript received by the Heat Transfer Division July There is sufficient sensible heat in this case within the wall to evap-
14,1980. orate the droplet. However, if the droplet diameter is equal to or

Journal of Heat Transfer AUGUST 1981, VOL. 103 / 501

Copyright © 1981 by ASME
Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 greater than the wall thickness, then the local temperature depres- becomes less; i.e., larger evaporation sites. However, the droplet
sions will be large. There is insufficient sensible heat in this case within density in the direction of the flow, i.e., the average time between
the wall to evaporate the droplet. Heat must be transferred from the impacting droplets, remains approximately constant. There is an
transport fluid, the thermal resistance between the transport fluid interdependence among the above parameters and the flooding index;
and the evaporating droplet; and the local temperature field within e.g., consider the special case when the total evaporator surface con-
the evaporator wall must be considered in the overall evaporation stitutes the evaporation site. The time between impacting droplets
process, i.e., the nonisothermal wall case. would be very short, and the surface at the point of droplet impact
Flooding Index. The continual maintenance of dropwise evap- would become flooded for most evaporant flow rates, i.e., even a small
oration on the surface of a spray evaporator is desirable in order to number of droplets can cause local flooding if the spray distribution
maximize the heat transfer rate and minimize the required evaporant is adverse.
flow rate for a given operating condition. Flooding causes a liquid film The average number of droplets generated per unit time can be
to form on the evaporator surface, thereby producing a thermal barrier represented by the ratio of the evaporant mass flow rate to the average
between the warm surface and the impacting evaporant droplets, droplet mass.
resulting in a reduction in overall performance for the spray evapo-
ration process. The ability to understand and predict evaporator N = 6m/d3 ir p (10)
flooding is useful in the design and analysis of a spray evaporator The average number of evaporation sites is given by the ratio of
system. evaporator surface area to the evaporation site area; i.e.,
The evaporator flooding index for the total evaporator is defined
as the ratio of the mass flow rate of the evaporant spray to the evap- N = A/4R2 (11)
orating mass flow rate; i.e.,
The average time between impacting droplets can be redefined from
<t> = m/(q/hfg) (7) equations (9,10) and (11) as follows.

When <t> is greater than unity, on average for the total evaporator t* = A d3 7T p/24m R2 (12)
surface, the system will become flooded. More insight on the flooding
It is convenient to define a dimensionless evaporation time for the
index can be gained if it is viewed at the local level or at a single
evaporating droplet. This is defined as the ratio of the actual evapo-
evaporation site; i.e., a small portion of the evaporator surface which
ration time to the minimum or ideal evaporation time; i.e., the time
serves as an evaporation region for a single impacted droplet at a given
required for evaporation if the site were totally wetted by the droplet.
point in time. Locally the flooding index can be defined as a time ratio;
This ideal time is obtained by taking the ratio of the droplet phase
i.e., the ratio of the average evaporation time for a single droplet at
change energy to the ideal heat transfer rate at the site based on total
a given site to the average time between impacting droplets at that
wetting.
same site.
t, = hfe (p 7T d3/6)/4h'R2 ATst (13)
<t> = t/t* (8)
where h' is the transport fluid heat transfer coefficient and ATst is
The average time between impacting droplets can be represented by
the difference between the transport fluid and evaporant temperature.
the ratio of the total number of evaporation sites on average to the
The nondimensional evaporation time becomes
number of droplets generated per unit time within the spray.

t* = N/N T = 24 t h'R2 pATstfhfg ir d3 (14)

(9)
The number and size of the evaporation sites on the evaporator sur- where T and t are the nondimensional and dimensional evaporation
face are dependent on the characteristics of the spray nozzle and its times, respectively. The flooding index can now be placed in final form
distance from and orientation to the evaporator surface. A fogging by combining equations (8,12) and (14).
nozzle will produce a large number of small droplets dispersed over :
T hfg rhlh' A ATst (15)
a wide area; i.e., a large cone angle. Thus, the evaporation sites will
be many and relatively small. As the nozzle is moved away from the where it remains to determine T as a function of the system parame-
surface, the droplet density in the direction perpendicular to the flow ters.

•Nomenclature-
a = thermal diffusivity of the evaporator N = average number of droplets generated 6 = nondimensional temperature, equation
wall per unit time (18c)
A = surface area q = heat flow rate A = nondimensional axial direction
b = impacted droplet thickness Q = nondimensional heat flow rate p = density
B = nondimensional impacted droplet q' = heat flux r = nondimensional evaporation time,
Q' = nondimensional heat flux equation (14)
thickness
r = radial direction, cylindrical coordinates tj> = flooding index, equations (7) and (15)
c = specific heat of the evaporator wall
r = impacted droplet radius
d = droplet diameter before impact R = evaporation site radius Subscripts
E = a quantity of heat t = evaporation time d = droplet
h = enthalpy T = temperature / = liquid evaporant
h' = heat transfer coefficient, transport t* = average time between impacting drop- fg = difference between vapor and liquid
fluid lets at a given location component
h * = heat transfer coefficient, evaporant 2 = axial direction, cylindrical coordinates g = vapor evaporant
k = thermal conductivity of the evaporator a = spreading ratio, droplet diameter after i = a grid point
wall impact to droplet diameter before im- / = ideal
£ = evaporator wall thickness pact s = saturation point
m = droplet mass P = wetting ratio, adllR st = difference between transport fluid state
M = nondimensional droplet mass A = difference between two values of the and evaporant saturation point
m = evaporant mass flow rate same parameter t = transport fluid
N - number of evaporation sites 1) = nondimensional radial direction w = evaporator wall

502 / VOL. 103, AUGUST 1981 Transactions of the ASME

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 Droplet Evaporation Times. Consider an impacted disk-shaped
droplet on the surface of a spray evaporator; the droplet is heated from
below by a warm transport fluid. The droplet lies at the center of an
evaporation site of radius, R. The heat conduction diffusion equation
governs the flow of heat within the evaporator wall.
d2r ldT d2T
2 (16)
dr r dr dz 2 a i>t
where T denotes the temperature at any point within the wall and a
represents the thermal diffusivity of the wall material. The associated
boundary conditions on the evaporation site are given by the following
relationships:

atr=R dT/dr = 0 (17a)

at r = 0 dT/dr = 0 (17b)
Fig. 1 Heat-flow lines and isotherms for an impacted droplet on an evaporator
surface
at z = 0 dT/dr = -h' [Tt - T(r,0,t)]/k (lie)

&tz=£ dT/dz = 0 for r>rd (t) (lid) dition by assuming that all of the heat flows through the impacted
(He) droplet thickness and that the heat transfer through the droplet
at z T = Ts for r < rd (t)
conforms to one-dimensional steady flow. A simple thermal circuit
The instantaneous impacted droplet radius, rd(t), is time-depen- can be setup which includes the transport fluid boundary layer con-
dent and varies with the rate of heat transfer. Fig. 1, a sketch of the ductance, the wall conductance and the conductance within the im-
heat flow lines for the above configuration at a given point in time, pacted droplet for both convection and conduction. The amount of
reveals that the flux lines are highly concentrated at the outer radii convective mixing within the droplet is dependent on the impact ki-
of the droplet, thereby causing the droplet to evaporate from the outer netic energy, two-phase pumping and natural convective currents.
radius inward toward the droplet center. This is in addition to the A nominal value of unity is taken for the Nusselt number. This ap-
evaporation from the upper surface of the droplet. Equation (16) is proximate one-dimensional model indicates that the conductances
made dimensionless by the following relations and equation (14). within the impacted droplet is 50 times larger than the conductance
within the transport fluid boundary layer for the case of a stainless
r, = r/R (18a) steel walled evaporator and for the spray evaporation parameters
consider in this paper. For an aluminum walled evaporator it is about
\ = zie (18b)
3.5 times larger. However for a copper wall the approximate model
6=[Tt-T (r,z,t)]/ATst (18c) predicts a value of only 1.5 times as large. It should be noted however,
that for the actual three-dimensional unsteady model a large portion
The diffusion equation now becomes
of the heat is not transferred through the droplet but at the outer
d2g 1 d6 IR\2 d20 _ (R2\ d0 radius, as can be seen in Fig. 1. In addition at the beginning phases
(19) of evaporation, when the one-dimensional model predicts the lowest
dij 2 v dy \e) dX2 \ot/j dr droplet conductance, the droplet is in a state of excitement due to the
The boundary conditions transform as follows: impact kinetic energy. Finally, the conductance of the droplet is
continually increasing due to the reduction in droplet diameter and
at r\ = 1 dfl/d?/ = 0 (20a)
thickness.
at i/ = 0 d0/d\ = 0 (20b)
Thus, for a very high wall thermal conductivity, such as copper, the
at X = 0 d8/dz = h£ 8(r,0)/k (20c) relative thermal conductance through the impacted droplet and
transport fluid boundary layer may be such that the interface tem-
at X = 1 dd/dz = 0 for r) > rd(t)/R (20d) perature would be more than a few degrees above the saturation
temperature as might be expected for the general case of the noniso-
atX = l 8 = 1 tor ri <rd(t)/R (20e)
thermal wall condition. This is dependent on the level of convective
The convective heat transfer between the transport fluid and droplet activity within the impacted droplet caused by the impact kinetic
occurs through the evaporator wall. Thus, the conventional choice energy, two-phase pumping and natural convective currents. Since
of driving potential for the transport fluid convective process is the the saturation temperature represents a more general approach this
temperature difference between the inner wall temperature, i.e., T value has be selected. For special cases a pseudo-saturation temper-
(r,z,t) = T (r,o,t), and the transport fluid temperature, where 2 = 0 ature can be defined to include the increase in interface tempera-
denotes the position of the inner wall in contact with the transport ture.
fluid and z = £ denotes the outer wall position in contact with the The model was first developed in three-dimensional rectangular
impacted droplet. The convective heat transfer coefficient is taken coordinate system (x ,y,z) but soon after changed to a two-dimensional
to have a constant value. The variation in h' due to changes in the fluid cylindrical coordinate system (r,z). The incremental increase in ac-
properties caused by the nonisothermal wall would be small. curacy for the three-dimensional space did not justify the longer
It is demonstrated above that for the nonisothermal wall condition computer run times, particularly in view of the errors associated with
the temperature below the impacted droplet is depressed by the matching the boundary conditons for the disk shape droplet in a
evaporating droplet; i.e., there is insufficient sensible heat within the rectangular space and the computer roundoff error associated with
wall to evaporate the droplet and heat must be transferred directly the longer run times. However, it should be noted that the ideal
from the transport fluid. evaporation time is defined by equation (13) in terms of the square
The thermal contact conductance between the impacted droplet based evaporation site rather than a cylindrical site as used for the
surface in contact with the evaporator wall and the internal thermal computer solution since there was no need to make this approximation
conductance of the impacted droplet are taken to be large compared for the normalization of the evaporation times and flooding index.
to the conductance within the thermal boundary layer of the transport An order of magnitude study of equation (19) reveals that the dif-
fluid. Thus, the temperature at the interface is taken to be approxi- fusion equation reduces to the Laplace equation. The combination
mately equal to the saturation temperature. of a high thermal conductivity and low heat capacity for the evapo-
A worse case approximation can be developed for the above con- rator wall material indicates that changes in the storage energy, i.e.,

Journal of Heat Transfer AUGUST 1981, VOL. 103 / 503

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 wall internal energy changes, are small when compared to the droplet Q = E(iI?«-U?)fli (26c)
phase change energy. The magnitude of the coefficient on the time Now, having defined the heat transferred during a given time in-
derivative term (the energy storage term) in equation (19), i.e., R2/atj, crement, it is possible to describe the incremental mass evaporated
causes that term to be small when compared to the remaining three during this time increment.
terms of the equation. This can be demonstrated formally by rear-
ranging the terms in equation (19) and estimating the order of the Am = q At/hfg (27a)
magnitude of each term. This is done by noting that each of the
If equations (14, 226) and (266) are combined, a dimensionless in-
nondimensional variables, i.e., r], X, T and 0, has an order of magnitude
cremental droplet mass is obtained.
of unity and the first and second derivatives have orders of magnitude
of unity and 1/e, respectively, where e denotes a small incremental AM=QAT (276)
value. A review of the isotherms and heat flow lines sketch, i.e., Fig.
The numerical solution is terminated when the sum of the incre-
1, supports this position. Prom equation (19),
mental dimensionless masses is equal to unity; i.e., when the droplet
i>28 i d0 my m le has been completely evaporated. Equation (276) gives the incremental
2 :0 (21a) time associated with each incremental decrease in mass based on the
di]2 r] di) \£) dX [at dr
total heat transferred in each time increment. As the droplet evapo-
o (1A) + od) + (^)2 0(l/e) - (- 0(1) = 0 (21b) rates, the droplet radius decreases and less surface area of the evap-
\atil oration site is wetted, thereby reducing the heat flow to the droplet.
2 u
2
a 0 1 d0 IR\2 d 0 For each incremental change in droplet radius, a new steady-state
2 •0 (21c) temperature field must be computed, as well as an updated value of
_ + H Ae) ax
2
dij r) di) the heat flow. If either the incremental time or droplet mass is selected
where 0( ) denotes order of magnitude. Typically, £2/atj would vary as the independent incrementing parameter, the droplet radius, in
between 0(1) and 0(0.1) for a nonisothermal wall evaporator, de- nearly all cases, will fall between two grid points. Therefore, it is far
pending on the evaporator wall material and thickness and droplet more conveneint to select the droplet radius as the independent in-
size. crementing parameter and have it change in nonfractional increments
However, it should be noted that the solution to equation (21c) is of the grid spacing, usually one grid space at a time. The incremental
still time-dependent, since the boundary conditions, equations (20d) time and mass now become dependent parameters.
and (20e), are time-dependent. The solution to the above set of
Equation (24) shows that the dimensionless droplet mass is de-
equations does lend itself most readily to a numerical/computer so-
pendent on both the droplet radius and thickness. As indicated above,
lution. However, before this can be accomplished, equation (20) must
the heat flux lines are highly concentrated at the outer droplet radii;
be defined in more detail. however, in the center or middle regions of the droplet, the flux lines
The instantaneous droplet radius is dependent on the droplet mass are quite uniform and nearly equally spaced. Thus, there is a reduction
and thickness at any given point in time. The initial mass of the im- in droplet thickness in the middle region of the droplet which con-
pacted droplet can be written in terms of the droplet radius and tributes to the droplet mass reduction, in addition to the mass evap-
thickness. orated in the outer radii; i.e., the region of highly concentrated flux
m(0)=T/-2i(0)b(0)p (22a) lines. The incremental change in droplet thickness in the middle re-
gions of the droplet is determined by evaluating the average heat flux
where m(0), rd(0) and 6(0) are the initial droplet mass, radius and at the interface between the droplet and the evaporator surface.
thickness, respectively. The initial droplet mass can also be described The incremental droplet thickness is given by
in terms of the droplet diameter before impact.
Ab = q' At/hfg p (28a)
m(0) = 7rd 3 /6 (226)
where q' is the average heat flux in the droplet middle region and p
The instantaneous droplet mass during the evaporation process is is the droplet density. However, it should be noted that the droplet
represented by mass and energy are conserved by equations (27) and (24) and not
m(t) = TV r\ {t) b (t) p (23) solely by equation (28a). If equation (28a) is divided by the initial
droplet thickness, a nondimensional thickness is defined.
It is convenient to define an instantaneous nondimensional droplet
mass by taking the ratio of equations (22a) and (23). AB = q' Atlhfg p 6(0) (286)

M(t) = (rd(t)/R)2 (6(t)/6(0)) (1//32) (24) It is convenient to nondimensionalize the heat flux and incremental
time in the above relation by making use of equations (186) and
The wetting ratio, (3, is defined as the ratio of the impacted droplet (18c).
diameter to the diameter of the evaporation site; i.e.,
q' = (k ATje) a0/dX (29a)
! = a d/2R (25)
and be rearranging the above, a dimensionless heat flux is defined
The incremental droplet mass evaporated within a given time in-
crement is determined from the heat transferred at the interface be- Q' = q'/(k ATs/e) (296)
tween the inner wall and the transport fluid. Once the steady-state
temperature field is determined for a given time increment, the heat or in more compact form
transfer along the wall can be found by summing the incremental heat Q' = a0/ax (29c)
flow for each spacial grid point.
When equations (14, 25, 28) and (29) are combined, the dimensionless
q-T,h' Ai (Tt - Ti) (26a) droplet thickness becomes
where Ti and A; are the temperature and heat transfer surface area
AB = Q' AT P2/(h' £/k) (30)
at a grid point. It is helpful in the development of the numerical so-
lution to define the above heat flow in terms of a nondimensional The assumption of approximate uniform droplet thickness is made
surface area and temperature. above even though there exists a large nonuniformity in the droplet
heat flux relative to droplet radius. The combined fluid action of the
9 = I'i'[irfi2(l?+1-1?)]ATI1 (266)
liquid droplet and droplet surface tension force causes the droplet to
The above relation can be rearranged to define a dimensionless heat maintain its original disk shape. The surface tension forces are rela-
flow along the wall. tively high for these size droplets, when compared to other forces such

504 / VOL. 103, AUGUST 1981 Transactions of the ASME

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 as the force due to gravitation and droplet fluid acceleration [5]. The
8 =
fluid can easily adjust to any local changes in droplet thickness for 1/10
the relative space and time increments of the droplet. The fluid ac- 3 - /
celeration required to maintain a steady, uniform droplet thickness
would be very small since the Reynolds number for this type of fluid
motion within the droplet would be about unity or in the creeping flow 2.5 - / 1/8

regime.
Numerical Solution Outline. The combination of a high ther-
mal conductivity and a low heat capacity for the evaporator surface 2
/ /
material, as well as the time-dependent boundary conditions, dictates
/ / * 1/6
a quasi-steady-state solution to this evaporation phenomenon. The
complex characteristics of the time-dependent boundary conditions
suggest that a numerical approach provides the best possibility for
a solution. The computer program proved to be long running, since 1.5 - / X yS
for each time increment a steady-state numerical solution to the La-
place equation was required. In addition, a variety of droplet sizes
relative to the evaporator wall thickness were considered and a
number of different values for the wetting ratio (see Fig. 2). The fol- 1/2
lowing list of equations represents a summary of the relationships used
sCsC-^^ ^~*-^ _____——
1
in the final solution. \mff===^r\ i. 1 • '
0 1 2 3 4 5
d 1 d<? IR\2 m ad/Jl
+ =0 (21c)
^^ b)^ Fig. 2 Dimensionless droplet evaporation times for various wetting ratios
and relative droplet sizes
at 77 = 1 iid/di) == 00
dd/dri (20a)
droplet mass. This may occur when the droplet radius becomes zero,
at j) = 0 dfl/dij = 0 (206)
i.e., at the center grid point, or if the droplet thins out first, i.e., the
at X = 0 d0/dz = h £ 8 {rfl)/h (_20c) instantaneous droplet thickness becomes zero.
The numerical procedure used for solving Laplace's equation was
at X = 1 dd/dz = 0 for r\ > rd(t)/R (20d) to transform the differential equation into a difference equation. This
atX = l 6 = 1 for 7] <rd(t)/R (20e) was accomplished with the use of first central-difference approxi-
mations for the first and second partial derivatives of the nondi-
Q = T. (vhi - Vf) Bi (26c) mensional temperature with respect to the nondimensional distances
in the r and z directions. In a like manner, the partial derivatives of
AM = QAT (276)
the boundary conditions were transformed. The error associated with
Q' = dd/dX (29c) the first central-difference approximations is related to the truncated
terms of a Taylor series and is approximately equal to the square of
AB = Q' AT 82/(h' £/k) (30) the grid size dimension. The nondimensional grid size used in the final
2 s solution varied between 1/10 and 1/40 for the r direction, depending
T = 24 t h'R pATst/hfg ir d (14)
on the size of the impacted droplet relative to the evaporation site,
n = r/R (18a) and 1/5 in the z direction. Grid sizes as small as 1/100 and 1/10 for the
r and z direction, respectively, were tried in the initial checkout phase
\ = z/£ (186) of the program. The worst case truncation error would be 4 percent,
0=[Tt-T (r,z,t)]/ATst (18c) based on the largest grid size of 1/5 in the z direction. The number of
iterations required for convergence varied between 200 and 500; the
<t> = T hfgrii/h' A ATst (15) condition placed on convergence was a less than 1 percent change on
all the temperatures within the field. Since the solution is on a closed
8 = a d/2R (25) domain, roundoff errors are not a factor, except on the last iteration
Equation (21c), the dimensionless Laplace equation, was placed of the solution.
into a difference equation form, along with the associated boundary Figure 2 presents the dimensionless evaporation times versus the
conditions given in equation (20). A particular droplet diameter, relative impacted droplet diameter for various values of the wetting
evaporation site radius and spreading coefficient were selected to ratio. This family of curves represents the complete numerical solution
generate a single point for Fig. 2. An appropriate grid size was selected, to the above system of equations. The two parameters associated with
and the time-dependent boundary conditions were defined based on the evaporant are the wetting ratio 8 and the relative impacted droplet
the parameters selected. The initial temperature field was determined diameter ad/£, A total of 25 conditions were selected based upon
from equations (20) and (21c). With this information, equation (26c) typical spray evaporation conditions. The parameters /3 and ad/£
is used to determine the heat flow rate between the transport fluid varied as follows: 1/10,1/8,1/6,1/4,1/2 for 8 and 1, 2, 3, 4, 5 for ad/£.
and the evaporator inner surface. The droplet was reduced by de- The incorporation of equation (30) into the numerical solution re-
creasing the droplet radius by one grid spacing. The incremental quires the selection of a value for the dimensionless spray evaporator
change in mass, resulting from the radius reduction, is combined with parameter h'£/k. This parameter is similar in appearance to the
the calculated heat flow rate in order to determine the time increment Nusselt number but different in its function. Here k is the thermal
from equation (27). The time increment is expanded to account for conductivity of the wall and not of the transport fluid, and £ is not a
the change in mass due to the change in droplet thickness as deter- characteristic dimension for the transport fluid flow but the wall
mined in equation (30). A running sum is maintained on the evapo- thickness of the evaporator. The value selected for this parameter is
ration times and on the evaporated droplet mass. For each incre- 0.036, which is typical for spray evaporators.
mental change in droplet radius, a new temperature field is computed
using the Laplace equation and the updated boundary conditions; Results and Discussion
i.e., the time-dependent conditions equations (20d) and (20e). A new
The fraction of heat transfer surface area covered by the impacted
heat flow rate and time increment are computed for each new radii.
droplets on the evaporator surface is relatively small; e.g., in [4] this
The solution continues until the evaporated mass is equal to the initial
fraction was measured experimentally and reported to vary from 0.015

Journal of Heat Transfer AUGUST 1981, VOL 103 / 505

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 to 0.196 or in terms of the wetting ratio /3 from 0.122 to 0.443. The -I 1 1 1 T-
quantity of heat transferred from the transport fluid directly below
the inner radii or droplet base compared to the heat transferred to
the outer radii or the peripheral heat transfer is approximately pro-
portional to the fractional wetted surface area, except as the droplet
radius becomes very small. When the initial droplet diameter is nearly ln(h*)
equal to the evaporation site diameter, the ratio of peripheral to base
heat transfer, as well as the unwetted to wetted surface area ratio, is
near unity. As the droplet evaporates, more of the surface becomes
unwetted and the heat transfer ratio increases. The calculated values
for this ratio ranged between 3 and 6. As the droplet radius becomes
very small, temperature gradients in the radial direction cannot be
a L series: LI, L2 and L3.
maintained to support the large variations in heat flux and the ratio
again approaches unity.
The assumption of a time-dependent impacted droplet radius was
made in the above model development based on the calculated high
heat fluxes in the outer droplet radii. Under a certain set of conditions,
a constant radius model can be assumed [4]. However, it should be
noted that the presence of a high heat flux a t the outer radii of the
evaporating droplet is more a result of the solution rather than an Fig. 3 Functional correlation for the wetting ratio and the spray heat transfer
assumption, even if a constant radius model is selected. However, how coefficient
the solution is finally carried out does depend on the droplet evapo-
ration dynamics model selected. A constant radius model requires the h* = (q/A)/ATst (336)
assumption that the droplet will thin out independent of a high Combining equations (32) and (33) for 4> equal to unity, i.e., for the
non-uniformity of heat flux and independent of a strong droplet maximum heat flux value just before flooding occurs, gives the fol-
surface tension action. As the droplet diameter becomes small relative lowing functional correlation for the heat transfer coefficient and the
to the evaporation site diameter, the concentration of heat flux in the wetting ratio, j3,
outer radii becomes less. The average droplet sizes considered in [4]
are relatively small, ranging between 51 to 99jt, so that the non-uni- ln(h*) o= (-1//? 1 - 34 ) (34)
formity in heat flux would be less pronounced. In addition, the A proportionality relationship results since the wetting ratio pa-
evaporator surface geometry selected by [4] makes the surface less rameter functional validity in equation (32) is being verified and not
of a nonisothermal wall than a typical spray evaporator surface, fur- the validity of the combined relative impacted droplet diameter and
ther reducing the nonuniformity in heat flux. wetting ratio parameters. Fig. 3 demonstrates the validity of equation
Figure 2 demonstrates the strong dependency of the droplet di- (34) based on experimental data presented in [4].
ameter on the evaporation times and on the flooding index defined The ordinate axis values for Fig. 3 are computed by taking the
in equation (15). Typically, ad/£ would vary between 2 and 3 for a natural log of the values in column 7, titled h (kW/m 2 K), presented
spray evaporator and spray nozzle generating droplets of about 300 in Table 2 of [4]. The abscissa values are developed from the values
/xm in diameter. However, these values can easily be outside the 2 to in column 11, titled e Meas. In [4] e Meas. denotes the measures value
3 range for droplets less than 200.U on the low end and for droplets of the fractional heat transfer area covered or wetted by the droplets.
greater than 600ft on the high end of the range. The droplet sizes Since /3 represents the wetting ratio or the ratio of the impacted
considered in [4] would place the value of ad/£ below 2 for most spray droplet diameter, ad, to the evaporation site diameter, 222, f is equal
evaporators. The wetting ratio, /?, is spray nozzle-dependent and to f}2. The square root of the values in column 11 are taken and then
varies not only with position within the spray but with the distance raised to the 1.34 power. The abscissa values are computed by taking
between the nozzle and the evaporator surface. Typical values for /5 the reciprocal of these last values.
of 0.122 to 0.443 are reported in [4] for a variety of nozzles. If the above The 28 data points reported by [4] are plotted in Fig. 3 with the
values are considered, the dimensionless evaporator tirnes, from Fig. exception of run number Cl (as identified by [4]), coordinates (16.7,
2, would vary between 1.05 and 1.50; i.e., a 5 to 50 percent increase in 2.77) which are off the graph in Fig. 3. This data run represents a very
evaporation time due to the wall and transport fluid thermal resis- light spray. The heat flux for Cl was reported to be 47.8 kW/m 2
tance. compared with the average heat flux of 782 kW/m 2 for all 28 runs. The
The family of curves in Fig. 2 can be represented, within 5 percent, second lowest or the next highest value of heat flux to Cl was run
by a single relationship for values of ad/£ between 0 and 2.5. A log-log number Al, 75.9 kW/m 2 or 58 percent greater than C l . Run Al does
plot of the average slope of each curve within the above range indicates correlate well with the other data, coordinates (11.8, 2.72); see Fig.
a linear correlation among the slope values or an actual exponential 3. If the analytical value of the wetting ratio computed by [4] is used
relationship among the slopes. Thus, the dimensionless evaporation for run number Cl rather than the experimental value, the correlation
times in the above range of ad/£ are given by is improved by a significant factor. The coordinates for C l would now
be (14, 2.77); the correct correlation should be (14,2.20) based on the
T = exp [(0.0075 ctdie)/^} (31) wetting ratio.
Upon combining equations (15) and (31), a relationship for the Also worth noting is the correlation of the complete L series, i.e.,
flooding index results. runs LI, L2 and L3, which lie more off line than do the remainder of
<*> = h fg m exp [(0.0075 ad/£)/0^*]/h' A ATst (32) the other data points; see Fig. 3. The L series represent a combination
for values of adit from 0 to 2.5 only. of a high wetting ratio and a low total volume flow rate from the nozzle.
Therefore, for a given evaporator, the flooding index is a function The wetting ratios for LI, L2 and L3 were 0.392,0.434 and 0.402, re-
of only two major parameters; i.e., the wetting ratio, /5, and the relative spectively, compared to a maximum of 0.443 and an average value of
impacted droplet diameter, ad/£. The functional validity of equation 0.308 for all 28 runs. The liquid volume flow rate for the type L nozzle
(32) relative to the wetting ratio parameter can be demonstrated with was 0.68 cm 3 /s, the lowest flow rate of the 14 nozzles considered. The
the use of the experimental data reported in [4]. cone angle for the L nozzle was 45 deg. The average flow rate for the
14 nozzles was 1.5 cm 3 /s. The second lowest or next highest flow rate
The heat flux on the evaporating surface is given by
to the L nozzle was the type C nozzle at 0.89 cm 3 /s or 31 percent
qlA = hfe rh/A (33a) greater than the L nozzle. The cone angle for the C nozzle was 80 deg.
and the heat transfer coefficient for the spray is defined as The combination of these two nozzle parameters, high droplet density

506 / VOL. 103, AUGUST 1981 Transactions of the ASME

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 on the evaporating surface and low nozzle flow rate produced a highly JOURNAL OF HEAT TRANSFER, Vol. 95, No. 2,1973, pp. 166-172.
concentrated spray with the narrow 45 deg cone angle and may have 2 Pedersen, C. O., "An Experimental Study of the Dynamic Behavior and
produced some local flooding of the evaporation surface. Heat Transfer Characteristics of Water Droplets Impinging Upon a Heated
Surface," International Journal of Heat Mass Transfer, Vol. 13, 1970, pp.
369-381.
Acknowledgment 3 Wachters, L. H. J., Smulders, L., Vermeulen, J. R., and Kleiweg, H. C,
"The Heat Transfer from a Hot Wall to Impinging Mist Droplets in the Sphe-
This paper presents the results of research performed under a roidal State," Chemical Engineering Science, Vol. 21,1966, pp. 1231-1238.
NASA-ASEE Summer Faculty Fellowship Program at the Johnson 4 Bonacina, C, Del Gindice, S., and Comini, G., "Dropwise Evaporation,"
Spacecraft Center in Houston, Texas. ASME JOURNAL OF HEAT TRANSFER, Vol. 101,1979, pp. 441-446.
5 Rizza, J. J., "Triple-Point Pressure Evaporation of Impacted Droplets,"
Journal of Spacecraft and Rockets, AIAA, Vol. 16, No. 4,1979, pp. 210-213.
References 6 Toda, S. and Uchida, H., "A Study of Mist Drops on a Heated Surface
1 Baumeister, K. J., and Simon, F. F., "Leidenfrost Temperature—Its at High Temperatures and High Heat Fluxes," Fourth International Heat
Correlation for Liquid Metals, Cryogens, Hydrocarbons and Water," ASME Transfer Conference, Paris-Versailles, 1970.

NOMINATIONS FOR HEAT TRANSFER DIVISION DISTINGUISHED

SERVICE AWARD
The ASME Honors and Awards (K-3) Committee is soliciting nominations for the Heat Transfer Division Distinguished
Service Award. The nominations for the award, to be given at the November 1982 ASME Annual Meeting, are due on
or before November 1,1981. The selection procedure and the guide to nomination preparation are now available from
the K-3 Committee, c/o Professor J. R. Welty, Department of Mechanical Engineering, Oregon State University, Corvallis,
Oregon 97331. Nominators are urged to obtain a copy of the guide before submitting their nominations.
This award has been established to reward unusual, outstanding, and sustained distinguished service to the Heat
Transfer Division of the ASME and to recognize such service in a manner that is comparable to the recognition presently
given to other achievements in the profession.

Journal of Heat Transfer AUGUST 1981, VOL. 103 / 507

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 Cooling of Freely Falling Liquid
Yasuhiko H. Mori
Kyuichiro Nagai 1
Hitoshi Funaba 2
Drops with a Shell of an Immiscible
Kazunari Komotori
Volatile Liquid
Department of Mechanical Engineering,
Keio University, An augmentation of evaporative cooling of liquid drops falling freely through the atmo-
3-14-1 Hiyoshl, Kohoku-ku, Yokohama 223 sphere has been attempted by attaching them with some amount of immiscible volatile
Japan liquid. An experimental study was performed for a water/n-pentane system, the water
either being pure or containing a small amount of surfactant. The cooling rate was found
to be strongly augmented provided the volatile liquid {n-pentane) engulfed almost com-
pletely the liquid which was to be cooled (water). The dependences of the cooling rate on
the initial drop temperature, volume ratio of the volatile liquid to the other in each drop,
air temperature and humidity were examined.

Introduction
The cooling of liquid drops while falling freely through an atmo-
sphere of lower temperature is the most commonly used operation
in an evaporative cooling system. The cooling rate will increase [•- Liquid A --]
strongly if some amount of volatile liquid is so attached to each drop
as to envelop it thoroughly or partly. This technique may be applicable ^Liquid B/

in some limited situations in which, for example, some means are Liquid B-
available to recover economically the evaporated volatile species or I 1 I
the evaporated volatile species are permitted to be discarded into the
Fig. 1 Possible configurations of composite drops
atmosphere from both economical and ecological points of view. A
crowd of such "composite drops" each of which consists of two im-
II The drop of liquid A and the drop of liquid B connect to each
miscible liquids may be formed by, for example, an impingement of
other so that a line bounding the three phases is formed.
liquid jets generated by a couple of adjoining nozzles. Gibson, et al.
III The liquid A-liquid B interface cannot form spontaneously;
[1,2] made a spray of composite drops by making a pair of brine jets
i.e. the drops remain separated by the gas without forming a composite
impinge at an acute angle upon a Freon jet locating in between, the
drop.
freeze desalination being the object of their study. Though they
evaluated experimentally the overall performance of such a novel IV The drop of liquid B is entirely enclosed by the liquid A, op-
spray freezer, the characteristics of heat and mass transfer from each posite to case I.
drop have been analyzed neither experimentally nor theoretically. We assume that the configurations are determined solely by the
three interfacial tensions; cr^ (liquid A-gas), as (liquid B-gas) and
This paper provides an experimental examination on the cooling
O"AB (liquid A-liquid B). The effects of shear on falling drop surfaces,
of single composite drops falling freely in an air which contains in its
drop oscillations, etc. are neglected. Thus, it can be shown, through
bulk a negligible amount of the volatile species evaporating from the
a similar deduction as described by Mori [5], that the above configu-
drops. The cooling rate is primarily controlled by the evaporation rate
rations are provided respectively under the following conditions.
of the volatile liquid, which in turn should depend on the configura-
tion of drops, i.e. how the volatile liquid is situated in each composite I SA < 0, SB > 0, SG < 0,
drop. The possible configuration for a specific combination of liquids II SA <0,SB< 0, Sa < 0,
can readily be predicted from the surface and interfacial tensions as (1)
described below. The dependence of the cooling rate on the configu- III SA <0,SB< 0, SG > 0,
ration has been confirmed in the experiments and will be shown IV SA >Q,SB <0,SG <0-
later. where SA, SB, So are the spreading coefficients defined as

P r e d i c t i o n of C o n f i g u r a t i o n s SA = <sB — (aA + oAB),-\

Here we consider the interaction of two immiscible liquid drops SB — aA - (aB + aAB),
brought into contact in a gas phase. The problem is similar to those (2)
SG = BAB — (aA + as)-
dealt with by Torza and Mason [3, 4] and Mori [5], the former being
concerned with the interaction of two immiscible liquid drops in a
third immiscible liquid and the latter with the configuration of gas- Some of the inequalities given in (1) can be eliminated, since they are
liquid two-phase bubbles in a second immiscible liquid. For conve- satisfied automatically if some other inequalities are satisfied. Fur-
nience, the less volatile liquid which is to be cooled is designated as thermore, we know that SG is negative for any combination of actual
liquid A and the volatile liquid is designated as liquid B. Possible liquids; this suggests that the case III is unreal. In other words, once
configurations after the contact of the two liquids are shown sche- drops of different liquids contact each other, they will never separate
matically in Fig. 1. These are: spontaneously. Thus the criteria for the configurations of composite
I The drop of liquid A is entirely enclosed by the liquid B. drops become
I SB>0,
II SA <Q,SB< 0,
1
Present address: Hitachi Consumer Products Research Center, Yoko- (3)
IV SA > 0.
hama
2
Present address: Ina Seito Co., Ltd., Tsukuba
Contributed by the Heat Transfer Division for publication in the JOURNAL
OF HEAT TRANSFER. Manuscript received by The Heat Transfer Division It is readily presumed that the configuration I is the most favorable
September 22,1980. for the purpose of cooling liquid A. On the contrary, the volatile liquid

508 / VOL. 103, AUGUST 1981 Transactions of the ASiWE

Copyright © 1981 by ASME
Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 B hardly contributes to the cooling of liquid A in case of configuration
IV. The configuration II is situated between the above two limiting
cases.
The exact prediction of configuration for a specific combination
of liquids is not always easy because of the difficulty to know the ac-
curate values of interfacial tensions under such a condition that each
liquid is saturated with the other and particularly at elevated tem-
peratures. Following are typical examples in which rather reliable
predictions are possible based on the criteria (3): (i) The combination
of water (,4) and a light hydrocarbon (B) in which the equilibrium
value of SB is generally close to zero will yield either configuration I
or configuration II in which only a small portion of the surface of liquid
A is not covered by the liquid B; (ii) a drop of a liquid metal (A) will
be completely enclosed by the shell of water (S), i.e. configuration I
will be available; (iii) a nonvolatile mineral oil (A) will completely
engulf a drop of water (B) resulting in configuration IV. In short, the
proposed process is favorable for the cooling of liquids with large
surface tensions.
Air
supply
In the experiments described below, the combination of water (A)
and rc-pentane (B) was used. A small amount of surfactant (sodium
oleate) was added in some runs to the water phase to change the
configuration of composite drops. In case of no surfactant in water
phase, a shell of pentane covered almost the whole surface of water Fig. 2 Schematic of experimental setup
phase, though we could not confirm whether it was truly configuration
I or not. An addition of the surfactant to the water phase reduced the
fraction of water surface covered by re-pentane resulting in the com-
posite drops of typical configuration II.

Experimental Equipment and Procedure

The experimental setup is schematically shown in Fig. 2. It consists
of a column containing the conditioned air, an air conditioning
equipment, a water feeding unit, are-pentane feeding unit, and a drop
catcher that could be positioned at any elevation to catch the drop
and to measure its temperature. The column was 170 mm X 170 mm
in cross section and 3 m in height, being sheathed by another square (Dimensions in mm)
column. Through a gap between the columns the air whose temper-
ature had been regulated was circulated to insure a uniform temper- Fig. 3 Construction of nozzles
ature distribution inside the inner column. The air whose temperature
To and humidity r/> had been regulated to specified values (TV. 20 ~ of composite drops. The water temperature was controlled by an
31 °C, <j>: 20 ~ 100 percent) flowed downward in the inner column at electronic system in which a silicon transistor was applied as a heater
about 60 mm/s to maintain a constant environment (humidity and and a sensor simultaneously [6]. The temperature monitored by a
vapor pressure of re-pentane).3 The temperature of the air measured sheathed copper-constantan thermocouple inserted in the nozzle at
by copper-constantan thermocouples at three heights in the column 3 mm from its tip was regarded as the nominal initial temperature of
never deviated over ±0.3 K around a specified level during each run. liquid A, TAI; the temperature in the region closer to the nozzle tip
The humidity was measured at the entrance of the column by use of was found to fluctuate due to the evaporation of pentane contacting
a wet and dry bulb hygrometer, which was removed before each run with the water so that it could not be employed as the standard of the
to prevent the addition of water vapor from its wet bulb to the air initial water temperature. TAI was regulated to either 50 °C or 31 °C
flowing into the column. Water deionized and distilled and n-pentane with a fluctuation less than ± 0.2 K during each run; the former being
of reagent grade were fed to the nozzles facing each other as shown higher and the latter lower than the saturation temperature of pen-
in Fig. 3 at specified flow rates, they came into contact with each other tane under the atmospheric pressure, 36.1 °C. The pentane feeding
at nozzle tips, and then fell down in the column one by one as a form unit was so carefully designed that the pentane never contacted with
any material other than glass and Teflon. All joints in the unit, e.g.,
those at the inlet and outlet of the flowmeter, were immersed in pools
3 of water to prevent tiny air bubbles from being entrained into the
The air velocity ( a 60 mm/s) was much smaller than the fall velocity of drops
except at very small falling distances («10 mm), and thus its effect on the drop flowing pentane through minute gaps in those joints due to capillarity.
cooling must have been negligible with a good approximation. The temperature of pentane was not controlled artificially. The

^Nomenclature-
A = liquid which is to be cooled TAf* = ideal value of TA( given by equation A;, Ao = indices of cooling augmentation
B = volatile liquid (4) defined by equations (6) and (7)
CPA = specific heat capacity of liquid A TAI = initial temperature of liquid A ft = volume flow ratio of liquid B to liquid
H = falling distance from nozzle tips TAO = lim#^oT,4(H) A
Ah„B = specific latent heat of vaporization of TA' = adiabatic saturation temperature of PA, PB = densities of liquid A and of liquid
liquid B B
air-liquid A system
SA,SB,SG = spreading coefficients defined " A , OB, OAB = surface tension of liquid A
in equation (2) TB' = adiabatic saturation temperature of
saturated with liquid B, surface tension of
TA = maximum temperature inside a drop air-liquid B system
liquid B saturated with liquid A, and in-
TAJ = temperature of liquid A in a composite T„ = air temperature terfacial tension of liquid A-liquid B in-
drop when the attached liquid B has just r = cooling efficiency defined in equation terface
vanished (5) <t> = relative humidity of air

Journal of Heat Transfer AUGUST 1981, VOL 103 / 509

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 temperature measured by a thermocouple locating in the nozzle at Drop
10 mm away from the tip varied between 8 and 18 °C depending on
the flow rate of pentane, while the flow rate of water and TAI were held
constant, because a part of pentane crept up the outside wall of the
nozzle made of Teflon and evaporated there, resulting in the cooling
of the nozzle. However, the effect of the variation in pentane tem- Acrylic plastics
1
perature on drop cooling was estimated to be sufficiently small and .tes
plates
was neglected in this study.
The volume fraction of pentane in each composite drop at the
moment of its departure from the nozzles could not be specified ex-
actly because some fraction of pentane evaporated before the de- Thermocouple
parture. Thus, we use the ratio of pentane flow rate to water flow rate, junction
ft, as the nominal value of initial volume fraction of pentane. The
experiments were designed to provide data for varying ft while water
volume in each composite drop and the frequency of drop departure
were held almost constant; this was accomplished by shifting slightly
the positions of two nozzle tips relative to each other correspondingly (Dimensions in m m )
with the variation of pentane flow rate under a constant water flow Fig. 4 Drop catcher detail
rate. The equivalent spherical diameter of water phase in a composite
drop and of a water drop was 3.5 ± 0.1 mm throughout the experi- the level of peaks on the oscillogram. Such a preheating was found to
ments except a special run in which water drops of 3.2 mm dia were hardly affect the level of peaks themselves: this indicates that the level
used for the convenience of comparison of the result with Yao and of peaks represented with enough accuracy the maximum tempera-
Schrock's result [9] with 3 mm dia drops. The frequency of drop de- ture inside drops irrespective of the preheating.
parture was 1.04 to 1.20 Hz. Further details of the experimental equipment and procedure are
To measure properly the variation in a certain representative described in references [7] and [8].
temperature (e.g., the maximum temperature or mean temperature)
of water in drops with the falling distance was a crucial problem in Results and Discussion
this study. The method of Yao and Schrock [9], i.e., collecting drops Results are shown in Figs. 5 to 9 in the form of TA versus falling
in a small dewar—which could be positioned at any elevatipn in the distance from the tips of nozzles, H. Except Fig. 9, the results are for
column—to measure the mean temperature, was not applicable for pure pentane-water system without surfactant. In every figure data
composite drops. Thus we attempted to measure the maximum on composite drops (ft > 0) are given together with the corresponding
temperature by the following method. data on water drops (ft = 0) for the convenience of comparison. The
The drop catcher as illustrated in Fig. 4 was mounted on a specially data on water drops agree well with those of Yao and Schrock [9] who
designed carrier that enabled us to move the catcher in three direc- measured the mean drop temperature instead of the maximum
tions independently by drawing wires from outside of the column. A temperature TA inside drops. The fact that little difference is found
copper-constantan thermocouple of 30 ftm dia was stretched rigidly between TA we measured and the mean temperature Yao and Schrock
at the entrance of a center hole (of 3 mm dia, a little smaller than the did, may be ascribed to a mixing of water due to oscillation of drops.
drop diameter) of the catcher into which drops were to be swallowed. Composite drops are apparently characterized by a large temperature
It was confirmed with such a catcher configuration that the water drop till their departure from the tips of nozzles due to a partial
phase in a composite drop contacted the junction of the thermocouple evaporation of attached pentane as well as by the cooling rates which
with little time delay when the drop entered the catcher almost con- are much higher than those of water drops, while falling through the
centrically, and then it slowly drained off the catcher through the air, due to the evaporation of residual pentane.
center hole. When a drop entered the catcher off its center, there was Dependence on Initial Volume Fraction of n-pentane, ft,. The
a larger time delay of the contact between the drop and the thermo- sheath of pentane in a composite drop must have some thermal re-
couple junction; this should have resulted in a lower temperature sistance to heat flow from the hot water core to the surrounding de-
exhibited by the thermocouple. The thermo voltage was amplified pending on its thickness. The calculation by assuming a concentric
by a d-c amplifier (Model 6L3-1, San'ei Instruments Co., Ltd.) and pentane shell around a spherical water core and a radial conduction
then recorded by an electromagnetic oscillograph (Visigraph FR-201, predicts the resistance of the pentane shell being dominant in the
San'ei Instruments Co., Ltd.). Thus, the periodical peaks on the os- present heat transfer process at least in an early stage of the fall of the
cillogram were expected to represent the maximum temperature TA drop. In practice, however, ft did not affect the cooling rate, as shown
of water in the drops at a falling distance H measured from the nozzle in Figs. 5 and 6, except that the cooling curve for ft = 0.3 began to
tips to the thermocouple junction, if the drops entered the catcher deviate increasingly from those for larger ft's at about 0.2 m of the
concentrically and if the response rate of the thermocouple was high falling distance probably due to the complete consumption of the
enough. In practice, the heights of peaks were scattered to some extent pentane. This fact indicates that the thermal resistance of the pentane
because of the eccentricity of impingement point to the center of the layer was negligibly small.
catcher being different from drop to drop; and thus the maximum of
A high-speed cine-photography [8] revealed that the major part of
a large number (160 to 620 depending on H) of peaks sampled suc-
the pentane formed a convexo-concave lens covering a limited area
cessively at a specified H was assumed to represent TA at that H. The
of the surface of water core while the residual surface area was covered
response rate of the thermocouple was considered to be high enough
with a thin film of pentane. The lens revolved quickly round the water
based on the following facts.
core while falling, probably resulting in a continual supply of pentane
By increasing the feeding speed of oscillopaper up to 1 m/s so that to the film region where the pentane was quickly consumed by
the detailed shape of each peak could be revealed, it was found that evaporation. Such a configuration and a motion of the pentane as
the thermo voltage reached a maximum asymptotically in about 4 ms described above were presumed to be the cause that minimized the
after the start of its rise, which was enough short compared to the resistance of the pentane layer.
duration of the thermocouple junction staying in the water phase in It is noticed in Fig. 5 that the cooling rate for ft = 0.3 in the range
a drop, and then fell slowly relative to the rate of rise up to the H > 0 . 2 m agrees with that for water drops with nearly the same in-
aforementioned maximum. The responding characteristics were also stantaneous TA- This indicates that the attached pentane had
examined by preheating the thermocouple with a power transistor evaporated almost completely till H = 0.2 m as described before and
attached to the catcher till the level of thermo voltage while the that the cooling of drops in the further range resulted from the
thermocouple junction was not in contact with drops came close to evaporation of water. The cooling rates for ft = 0.43 and 0.74 agree

510 / VOL. 103, AUGUST 1981 Transactions of the ASME

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 with each other and are appreciably higher than those for water drops
T» = 20.7'C 5, "fci/C with similar TA 'S throughout the whole range of H in the present
50 f =35 •/. s 0 49.5 experiments; this indicates that the cooling was due to evaporation
• 0 40.6 of pentane throughout the whole range of H.
A 0 36.1
Dependence on Humidity and Temperature of Air. Figures 7
a 0.30 49.7
and 8 show the variations of cooling curves with the relative humidity
o 0.43 49.7
4) and the temperature T» of the air. The cooling curves for composite
45 A 0.74 49.7
drops were not affected by 0 and T . whenever some amount of pen-
-*-~ tane remained in the drops, but the cooling rate for water drops was
appreciably reduced with an increase of <j> from 35 to 100 percent (Fig.
7) and was slightly reduced with an increase of T„ from 21 to 31 °C
(Fig. 8) due to the suppressions of mass transfer (i.e., latent heat
transfer) and of sensible heat transfer respectively. This fact clearly
indicates that the cooling of composite drops is dominated by the
latent heat transfer due to the evaporation of pentane.
Effect of Drop Configuration. The effect of an addition of sodium
oleate to water phase is demonstrated in Fig. 9. The addition of so-
dium oleate resulted in a decrease of cooling rates for both water drops
and composite drops. The decrease of cooling rate for water drops may
be ascribed to the suppression of oscillation of drops while falling. The
high-speed cine-photography showed that the addition of sodium
30 _l_ _i_ oleate caused a 20 percent reduction of frequency and a little increase
0.5 1.0 1.5
Falling distance H , m
of amplitude of the oscillation of water drops in the range of H < 0.2
m; the effect on the oscillation of composite drops was less remarkable.
Fig. 5 Dependence on initial volume fraction of pentane (TM '• 50 °C) The decrease of cooling rate for composite drops due to the addition
of sodium oleate is much more drastic. Such a decrease seems to be
primarily ascribed to the change in the configuration of composite

T.= 20.7*C ti TAi.-C

35 T.,'C f . ' l . 5, \,X
f = 35 V. • 0 30.9
D 0.30 30.8 35 • 20.7 35 0 30.9
O 0.43 31.0
A 31.0 22 0 30.9
A 0.74 30.8
O 20.7 35 0.43 31.0

30 K. A 30.5 22 0.43 31.0

20
0.5 1.0 1.5
Falling distance H , m 20,
0.5 1.0 1.5
Falling distance H , m
Fig. 6 Dependence on initial volume fraction of pentane (TAI ^ 31 °C)
Fig. 8 Dependence on air temperature

?,"/. I, TAi , * C Surface active agent i, \X

T.= 20.7*C
35 o 35 0 30.9 35 9 - 0 30.9
A 100 0 30.9
A Sodium oteate(0.rW 0 30.9
O 35 0.43 31.0
O
- 0.43 31.0
A 100 0.43 30.9
A Sodum oteate(0.r/.) 0.43 30.9

30
'**-*.-__^__

;
20.7*C
• 35 °U
e 25

20
20 0.5 1.0 1.5
0.5 1.0 1.5
Falling distance H , m Falling distance H , m

Fig. 7 Dependence on relative humidity Fig. 9 Effect of addition of a surfactant to water phase

Journal of Heat Transfer AUGUST 1981, VOL. 103 / 511

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 drops as described before, i.e. from configuration I (or configuration
II with a very small surface area not covered by pentane) to the typical ;, TA, ,'C V.'C
D 0.30 49.7 33.0
configuration II, resulting in disappearance of the surface area covered
1.5 o 0.43 49.7 258
with a thin film of pentane where the internal resistance to heat
A 074 49.7 86
transfer must be very small. Further, the cine-photography suggested H 0.30 30 8 14.3
that the mobility of the pentane lens round the water core was ap-
© 0.43 31.0 7.3
preciably suppressed, probably being related to the presence of the A L 0.74 30.8 -10.0
three-phase contact line. Hence, it is reasonably presumed that the
uJl-0
cooling rate was remarkably reduced due to an increased internal
resistance of the pentane layer which was thicker and less mobile as
well as a decreased area of pentane-air interface providing an ap-
preciably larger resistance to the diffusion of pentane vapor.
Cooling Efficiency. The absolute cooling limit for water drops is
given by the adiabatic saturation temperature for air-water system
(i.e., wet bulb temperature), TV, while that for composite drops is
given by the adiabatic saturation temperature for air-pentane system,
TB'. The former is, for example, 12.3 °C under the condition that T„
= 20 °C and <j> = 35 percent, and the latter was roughly measured to 0.5 1.0 1.5
be as low as —15 °C under the same T - and the vanishing pentane Falling distance H
concentration in the free-stream air. In practice, the temperature TAJ
Fig. 10 Cooling efficiency
which the water temperature attains in a composite drop when the
attached pentane has just vanished is restricted by f; and may be
higher than TV in some cases. If T&f > TV, the drop will be further
cooled gradually in the succeeding fall in the air by the evaporation
of water; and if T V < TV, the drop will be heated up to TV- Hence,
TA; may desirably be predicted from the practical point of view. If the
pentane supplied to the nozzle is completely used to cool the water, 25
TA/ is equal to TAJ* given by T-=20.7"C t; TA,.*C
f = 35 •/. s 0.30
TAr = TAi_kmA!hB. (4) 8 0.43 -50
PA CpA A 0.74
20 a 0.30
In practice, TA/ must be somewhat higher than TAf*, because some
fraction of pentane evaporates during the formation of a composite
0 0.43 -31
A 0.74
drop cooling the nozzle and the air, and because a much smaller
fraction may be lost due to the sensible heat transfer from the air to 4..
pentane surface if the temperature at pentane surface becomes lower
than To. in the course of the fall of the drop. Here we define the cooling
efficiency T as

T-
TAJ ~~ TA If 5 - * «
(5)
TAI - TAf*
which is plotted against the falling distance in Fig. 10. It is presumed
that the pentane attached to each drop vaporized completely during
its fall till H o± 0.2 m for the case f; = 0.3 as described before. It is seen
J 0.5 1.0 1.5
in Fig. 10 that at this falling distance
Falling distance H , m
r = 0.3 ~ 0.4 : TAi =* 31 °C and Fig. 11 Total cooling augmentation
r = 0.6 ~ 0.7 : TAi =* 50 °C.
It should be noted that T depends on the condition of formation of
composite drops. Higher flow rates of pentane and water shorten the
period of drop formation, and thus will result in a decrease of the
initial loss of pentane. Hence, the cooling efficiencies as shown here
are expected to be appreciably raised in practical cooling systems 15
operated at much higher flow rates of liquids. T»=20.7'C <;, WC
Evaluation of Cooling Augmentation. Though the cooling rate f = 35 % B 0.30
9 0.43 -50
of composite drops being appreciably higher than that of water drops
A 0.74
can readily be recognized in Figs. 5 to 9, the extent of augmentation
is not quantitatively specified there. Since the cooling of composite J,o, D 0.30
O 0.43 -31
drops depends on many factors, no definite method of quantitative A 0.74
evaluation of augmentation can be presented. Under each condition
(a combination of liquids, TAI , T«, ,</>, f; and drop size are specified),
however, the augmentation may be evaluated by use of specific tem-
perature fall {TAi — TA) I (TAI — T„) of composite drops to that of
water drops with almost the same TV at each falling distance H. The
dependence of *• I
U^i ^A/ix^Ai •*• o>) composite drop
A; (6)
( T A ; — TA)l(TAi — T „ ) w a t e r drop as 1.0 1.5
Falling distance H , m
on H is shown in Fig. 11 for the runs presented in Figs. 5 and 6. Here
the cooling augmentation of drops at the nozzle tip and that during Fig. 12 Cooling augmentation of drops during their free fall

512 / VOL. 103, AUGUST 1981 Transactions of the ASME

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 their free fall are evaluated in the gross. The latter may rather be history of each drop while the volatile liquid is remaining on the drop
evaluated by using surface but it determines the total temperature fall attained when the
volatile liquid has vanished.
. _ (TAP ~ TA)/(TAO — T<,)compoaite drop .
(TAO — TA)/(TAO ~ ^«)water drop Acknowledgment
where TAO = lim^—oTxCff) is obtained by the extrapolation of TA The authors greatly appreciate the help and advice afforded by Dr.
versus H relation in each run, and is lower to some extent than TAi Y. Tochitani in constructing the experimental setup. K. Higeta is also
for composite drops and practically the same with TAI for water drops. acknowledged for his help, particularly in the high-speed cine-pho-
In Fig. 12 Ao is evaluated for composite and water drops with almost tography. H. Mitsuyoshi, Y. Ueno and K. Ishimitsu partially assisted
the same TAI, though it should be done, in principle, for those with the experiments. This work has been sponsored in part by the Grant
same TAO- A proper evaluation of Ao for composite and water drops in Aid for Scientific Research of the Ministry of Education of Japan
with same TAO would give higher values than those given in Fig. 12. under Grant 375113.
The extent of augmentation represented by such an index as A; or
Ao becomes more remarkable as TAI (or TAO) decreases, causing a References
suppression of water evaporation. It should be kept in mind that A; 1 Gibson, W., Emmermann, D., Grossman, G., Johnson, R., Modica, A., and
(or Ao) goes to infinity as TM (or TAO) approaches TA', the adiabatic Pallone, A., "Spray Freezer and Pressurized Counterwasher for Freeze Desal-
saturation temperature. If TAI {TAO) decreases further beyond TA', ination," Proceedings of the Fourth International Symposium on Fresh Water
from the Sea, Heidelberg, Sept. 1973, Vol. 3, pp. 343-345.
A; (Ao) becomes negative and loses the meaning as an index of cooling 2 Johnson, W. E., Fraser, J. H., Gibson, W. E., Modica, A. P., Grossman,
augmentation. G., and Stepakoff, G. L., "Spray Freezing, Decanting, and Hydrolysis as Related
to Secondary Refrigerant Freezing," Office of Saline Water, Research and
Development Progress Report No. 786,1972.
Concluding Remarks 3 Torza, S., and Mason, S. G., "Coalescence of Two Immiscible Liquid
Drops," Science, Vol. 163,1969, pp. 813-814.
The possibility of augmentation of the evaporative cooling of falling 4 Torza, S., and Mason, S. G., "Three-Phase Interactions in Shear and
drops by attaching them with some amount of an immiscible, more Electrical Fields," Journal of Colloid and Interface Science, Vol. 33,1970, pp.
volatile liquid has been considered and examined experimentally. 67-83.
Dependences of cooling rate on various factors and the extent of 5 Mori, Y. H., "Configurations of Gas-Liquid Two-Phase Bubbles in Im-
miscible Liquid Media," International Journal of Multiphase Flow, Vol. 4,
available augmentation were exemplified for water drops of 3.5 mm 1978, pp. 383-396.
dia sheathed by rc-pentane. Following are principal findings that seem 6 "Temperature Limiter for Small Variations," Electronic Engineering,
to be not specific for such a system but of general nature. Vol. 49, No. 590, Apr. 1977, p. 23.
The cooling is intensely augmented on the condition that the 7 Nagai, K., "Cooling of Freely Falling Liquid Drops with a Shell of an
Immiscible Volatile Liquid," (in Japanese), M. Eng. thesis, Keio University,
spreading coefficient of the volatile liquid on the other which is to be Yokohama, March 1979.
cooled is positive or at least around zero, so that the volatile liquid 8 Funaba, H., "Cooling of Freely Falling Liquid Drops with a Shell of an
encloses almost completely the other. The layer of the volatile liquid Immiscible Volatile Liquid," (in Japanese), B. Eng. thesis, Keio University,
hardly adds to thermal resistance even if its thermal conductivity is Yokohama, March 1980.
9 Yao, S.-C, and Schrock, V. E., "Heat and Mass Transfer from Freely
fairly low because of its motion round a drop. Consequently the initial Falling Drops," ASME JOURNAL OF HEAT TRANSFER, Vol. 98, 1976, pp.
volume fraction of the volatile liquid does not affect the temperature 120-126.

Journal of Heat Transfer AUGUST 1981, VOL. 103 / S13

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 S. J. Palaszewski
A Three-Dimensional Air-Vapor-
Bell Laboratories,
Member of Technical Staff,
Whlppany Road,
Droplet Local Interaction Model for
Whlppany, NJ 07981
Assoc. Mem. ASME
Spray Units
L. M. Jiji
Professor.
An approximate fully three-dimensional numerical model for predicting the detailed flow
and thermal characteristics of spray units is presented. The new model differs from all
previous analyses of spray cooling systems in that it determines the local variation in the
S. Weinbaum dry-bulb temperature, absolute humidity and the air streamlines throughout the flow
Professor. field encompassing the spray umbrella and the effect of this local variation on drop cool-
ing. The conservation equations to determine the local absolute humidity, velocity and
dry-bulb temperature of the air-vapor phase are written in Lagrangian form where the
The City College of the
City University of New York, droplets are treated as spatially varying sources of mass, momentum and energy. The
Department of Mechanical Engineering, analysis takes into consideration stable and unstable meteorological conditions, turbu-
Convent Avenue at 138th Street, lent mixing in the atmospheric surface layer and nonuniform upwind and local air-veloci-
New York, NY 10031
ty profiles. The numerical model is of intrinsic interest because it demonstrates the feasi-
Mem. ASME
bility of computing a three-dimensional, two-phase flow field without the use of excessive
computer time. Model predictions of droplet return temperature along the spray center-
line were compared with data for the Spraco 1751 nozzle. Good agreement was observed.

Introduction of the mass, momentum and energy exchange experienced by the

In recent years, spray cooling systems have received favorable at- air-vapor element from the moment it approaches the spray to the
tention from the power industry as an alternative to cooling towers time it reaches its present position. It also depends on the meteoro-
for the dissipation of waste heat from electric generating plants. Spray logical conditions in the lower part of the earth's boundary layer. The
cooling systems have a number of advantages over cooling towers. problem is further complicated by the fact that particle trajectories
These include economy of construction, ease of maintenance and the and air-vapor streamlines differ substantially, as shown schematically
ability to control the rate of heat dissipation by selectively shutting in Fig. 1, which illustrates that a vapor bubble does not follow the drop
off or turning on any number of spray modules. Factors of land as assumed by Elgawhary [3] and Frediani and Smith [4].
availability, environmental considerations and the inability of current In view of the above complexities, different approximations have
theoretical models to accurately predict their performance are the been introduced for obtaining the local wet-bulb temperature. The
major reasons why spray cooling systems for waste heat dissipation two limiting approximations are that the wet-bulb temperature is
have had very limited use to date. Ryan and Myers [1] underscore the given by upstream ambient conditions [2] or that the air is saturated
uncertainty in current theories by an example calculation where the
number of required spray units for a given design criteria varies by
a factor of 2.5 depending on the theoretical model used. A magnitude
of uncertainty of this order is intolerable in view of the substantial
capital costs of each spray module. The purpose of this paper is to
demonstrate the feasibility of solving a three-dimensional, two-phase
flow field by using a comprehensive and accurate numerical finite
difference model which requires little computer time. The model
considers for the first time the local changes in the air-vapor phase
$?%~JZ y * (LONGITUDINAL)
due to the presence of the spray and the effect of these local changes
on the energy released by the droplets.
Existing theoretical models for predicting the performance of spray
units and systems have been based in large measure on approximate
solutions to Merkel's equation for the energy transfer of an individual -im
droplet from the moment it leaves the nozzle until it returns to the
canal. The driving potential for the energy loss due to convective and
evaporative cooling at the drop surface is the instantaneous difference PLAN VIEW
in the total heat function (enthalpy of air-vapor phase minus the
liquid enthalpy of the vapor fraction at the adiabatic saturation ^DROPLET
temperature) between the drop temperature and the local wet-bulb / TRAJECTORY AIR-VAPOR
temperature. Merkel's equation neglects radiation losses from the NVERTICAL) / ' ELEMENT?
drop and assumes that its thermal relaxation time is short compared // \ z /
N
to the trajectory time. Soo [2] has shown that both are reasonable ^STREAMTUBE /
approximations for typical spray units.
The principal difficulty in evaluating Merkel's equation is that the
local wet-bulb temperature can change significantly as the air is
heated and gains moisture while passing through the spray. The local
wet-bulb temperature is a complicated function of the time history

Contributed by the Heat Transfer Division for publication in the JOURNAL ELEVATION V I E W
OF HEAT TRANSFER. Manuscript received by the Heat Transfer Division July'
14,1980. Fig. 1 Trajectory of a fluid element as it crosses the spray field

514 / VOL 103, AUGUST 1981 Transactions of the ASME

Copyright © 1981 by ASME
Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 in the spray domain [5]. Other investigators have introduced an ap- Another important consideration in predicting spray unit perfor-
propriate constant average wet-bulb temperature [6, 7], or have uti- mance is the droplet size distribution. For the 56 kW Powered Spray
lized convective-diffusion theory (incorporating an experimentally Module manufactured by the Ceramic Cooling Tower Company that
obtained spray heat release parameter NTU) to determine a dimen- is examined in the present paper, the droplets are relatively mono-
sionless interference factor f relating the local wet-bulb temperature dispersed and an effective Sauter-mean diameter is used. For sprays
to the initial droplet and ambient wet-bulb temperatures [8, 9]. The exhibiting a distribution of drop sizes, a single effective drop size may
latter models in particular provide useful engineering estimates of not be satisfactory. These cases are easily treated in the framework
dimensionless groups, denoted by NTU or SER, which characterize of the present model by dividing the droplet size distribution into
the total energy released by the droplets. No previous model has at- discrete intervals and calculating the particle trajectories for each
tempted to theoretically predict the local variation in wet-bulb tem- interval. Thus the analysis of the Spraco 1751 nozzle described later
perature or the three-dimensional streamline patterns of the air-vapor in the paper considers five different droplet sizes each with a different
phase, two important objectives of the present finite difference model. ejection angle.
The streamline patterns obtained from the solution are particularly Two complementary papers are in progress. The first of these pa-
instructive since they show how the droplet drag and buoyancy forces pers deals with the application of the basic model presented in this
affect the air-vapor circulation in the spray as a function of wind speed study to the performance and design optimization of individual spray
and atmospheric stability. units. The second paper analyzes the interaction between spray units
While the more comprehensive model developed herein eliminates in a module or a pass.
the need for empirical factors such as effective droplet diameter or
an average wet-bulb temperature, it does use the empirical correlation Model Description
equations of Ranz and Marshall [10] for the convective heat and mass At present, finite difference solution schemes do not exist for solving
transfer from the drop and an eddy viscosity model for the turbulent the Navier-Stokes equations in three dimensions, even without the
diffusion of mass, momentum and energy in the air-vapor mixture. added complications of the turbulent transport terms or the addi-
In essence, one has replaced an ad hoc empirical correction factor by tional terms due to the droplet phase. One thus needs to devise an
fluid equations with an approximation for local transport properties. approximate set of model equations which retain the three-dimen-
The important advantages of using the latter approach are that one sional character of the flow field and droplet trajectories but which
obtains a picture of the air-vapor flow field and can calculate the heat can be solved in a few minutes on a large capacity high speed com-
loss for a droplet along any particular trajectory in the spray umbrella. puter. Short computational times are essential if the computer code
Furthermore, since the exit conditions for one spray unit provide the is to be used for spray design and optimization.
entrance conditions for the next spray unit in the pass, one has The minimum requirements of the model are that it:
eliminated the need for additional empirical correction factors be- 1 accurately calculate the trajectories of the droplets released in
tween rows in a spray system. One important limitation of the model any direction with respect to the incoming wind profile taking
formulation is that the ambient air is assumed to be in motion. The into account the instantaneous local difference in velocity and
numerical calculations give reasonable results provided the air ve- direction between the air-vapor and droplet phases (see Fig.
locities induced by droplet drag and buoyancy are of the same order 1),
as the local wind speed. For typical power plant spray units this 2 determine the three-dimensional path of an air-vapor element
implies a minimum wind speed of approximately 1.0 m/s. through the spray umbrella and the change in local dry-bulb

-Nomenclature.
A = cross-sectional area of streamtube Aj = number of droplets ejected from the 6 - prescribed tolerance for convergence of
CD = drag coefficient nozzle per unit time for a given trajec- spray heat transfer from successive itera-
Cp = constant pressure specific heat tory tions
c = number characterizing the size of an air NTU = number of transfer units 6 = droplet discharge angle in the horizontal
element Pr = Prandtl number = MaCp„//ea plane (measured from the x -axis)
D = diffusion coefficient Q = rate of heat transfer from a spray unit H = viscosity
d = droplet diameter Re = Reynolds number = pad\ v — u\/na p = density
dc — droplet diameter upon canal re-entry s = distance along a streamline or drop tra- <j> = spray discharge angle in the vertical
g = gravitational acceleration jectory plane measured from the x-y plane
Ho = surface heat flux Sc = Schmidt number = lxa/(PaD) a = absolute humidity
h = droplet heat transfer coefficient SER = spray energy release
ho - droplet mass transfer coefficient T = temperature, T* = (T - Twba)/(Tdo - Superscript
h[ = enthalpy of water at saturation tem- Twb«.)
* = dimensionless variable
perature Tc = spray cold water temperature at canal
i = enthalpy of air — vapor mixture re-entry
j = unit vector in the wind direction (longi- Tdo - spray supply temperature Subscripts
tudinal) Tra = reference temperature 0 = initial value at nozzle discharge
k = thermal conductivity t = time 2m = ambient value at a 2m height
k' = von Karman's constant {/» = wind speed <=° = ambient value at any height
Kh = turbulent thermal conductivity of air u = air velocity vector a = air
Km = turbulent viscosity of air u* = friction velocity av = air-vapor mixture
Kwu — turbulent thermal diffusivity of water V = volume of a finite air-vapor element d = droplet
vapor in air v = droplet velocity vector db = dry bulb
£ = lateral dimension of streamtube Do = initial speed of the mean size drop i = droplet size in size distribution
L = latent heat of vaporization of water x = horizontal space coordinate perpendic- j = trajectory number
Lmo = Monin-Obukhov length scale = ular to the wind direction k = iteration number
- PaCpa TdbU» 3/(k 'gHQ) y = horizontal space coordinate in the wind s = saturated air
M = droplet mass direction v = water vapor
m = vertical dimension of streamtube Z, z = vertical distance above canal surface wb = wet bulb
N = total number of trajectories 20 = roughness height y = horizontal space coordinate in the wind
n = number of droplets in an air-vapor ele- PT - coefficient of thermal expansion of direction
ment air z = vertical space coordinate

Journal of Heat Transfer AUGUST 1981, VOL. 103 / 515

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 temperature and absolute humidity along this path taking into of evaporation that a droplet experiences during its time in flight, one
account droplet drag, buoyancy, forces due to thermal gradients relates the time rate of change of droplet mass to the evaporative flux
in the air-vapor phase, the local convective and evaporative across the concentration boundary layer of the droplet:
losses from the droplets and the turbulent mixing of the gaseous
dM/dt = -pahDTrd2[us(Td)-<1>}. (1)
phase,
3 include basic meteorological conditions such as the stability of The mass transfer coefficient ho is determined by an empirical cor-
the atmospheric surface winds both in the incoming velocity relation given by Ranz and Marshall [10]:
profile and in the expressions for the turbulent transport
coefficients. hD = (D/d)[2 + 0.6(Re)^ 2 (Sc) 1/3 ]. (2)
Two reasonable approximations which greatly simplify the field
Conservation of Droplet Momentum. Droplet velocities are
equations for the air-vapor phase are that the pressure field is uniform
determined by applying Newton's law for a particle. Consideration
and that turbulent transport is of the boundary layer type (only tur-
is taken of gravitational forces and the drag induced by the local
bulent diffusion normal to the earth's surface is considered). With
vector-velocity difference between the droplet and the air-vapor
these two approximations, the differential equations for the air-vapor
continuum. The resulting equation is:
phase change from elliptic to parabolic form, provided the droplets
can be treated as known sources of mass, momentum and energy. Mdv/dt = Mg - (CDPa7rd2/8)|v - u|(v - u ) (3)
Unfortunately, the droplet source terms are not known since they
depend on the local velocity and wet and dry-bulb temperatures of where [12]:
the air-vapor phase. Thus, an iterative solution procedure must be CD = 0.44 = drag coefficient for Re > 1000 (4a)
adopted in which the lowest order solution for the droplet trajectories
and temperature is based on ambient conditions. This iterative pro- CD = (24/Re)(l + 0.15 Re 0 - 687 ), Re < 1000. (46)
cedure will be described in more detail later. The important point to
be made here is that for each iteration the droplet source terms can A droplet trajectory is found from:
be computed, the air-vapor differential equations treated as a para- dsd/dt = v. (5)
bolic system and a finite difference forward marching scheme em-
ployed. Conservation of Energy for a Droplet. Taking into consider-
ation surface convection and evaporation, and neglecting temperature
The governing equations that have been adopted for the air-vapor
gradients within the droplet, the energy balance gives
phase differ from conventional boundary layer equations in several
important respects. Although the pressure field is neglected, the -MCpddTd/dt = wd2h(Td - Tdb) + ird2pahD(us(Td) - co)L.
momentum equation contains nontrivial components both normal
(6)
to the canal surface and in the lateral direction (see Fig. 1). The
transverse forces are generated by buoyancy and droplet drag com- The terms on the right hand side of equation (6) are the energy
ponents normal to the wind direction. A novel computational scheme given up by the droplet due to convection and evaporation, respec-
is used for calculating the transverse velocity components and tively.
streamline displacements and is described in detail by Palaszewski The heat transfer coefficient h is obtained from an empirical cor-
[11]. Secondly, because the droplet trajectories are distinct, the source relation for a droplet given by Ranz and Marshall [10]:
terms for mass, momentum and energy are distributed along discrete
surfaces in the air-vapor space. To calculate the number density of h = (kjd)[2 + 0.6(Re) 1/2 (Pr) 1 «]. (7)
droplets along each surface one must account for both the divergence Water Vapor Continuity Equation. To determine the variation
of the trajectories with distance from the spray nozzle and the droplet of absolute humidity along any air-vapor streamline in the spray field,
velocity variation along each trajectory. This kinematics problem species conservation along a streamline is required. Conservation of
must first be solved and the spatial distribution of the droplets stored
mass for the water vapor in a fluid element gives:
in the computer before the source terms can be calculated in the air-
vapor differential equations. Thirdly, the duration of a droplet in an paVdoo/dt = paird2nhr>[us(Td) - co] + paVKwud2(i}/dz2 (8)
air-vapor element traversing the spray envelope depends on the local
relative velocity difference between the phases. where

These particular features of the spray flow field motivate the use du/dt is the total derivative following the fluid motion.
of a Lagrangian coordinate description in which each air-vapor ele-
The first term on the right hand side of equation (8) represents the
ment is followed along a rectangular streamtube whose cross-sectional
droplets as sources of water vapor to the air-vapor elements. The
area changes with the local speed and whose side dimensions are ob-
second term is the net efflux of water vapor across the boundaries of
tained by considering the transverse momentum equations. The in-
the element due to turbulent diffusion.
teraction between streamtubes occurs through the kinematic dis-
placement of streamlines required from continuity and turbulent If an element is considered to have the shape of a rectangular par-
diffusion. In a typical calculation, a rectangular grid of 10 by 20 stream- allelpiped, the volume of an element can be characterized by an
tubes is used to describe the air-vapor flow field on one side of the equivalent cube, whose side dimension is some multiple of the initial
symmetry plane of the spray. The surfaces representing the locus of mean droplet diameter:
points of droplet trajectories are divided into anywhere from 30 to 100 V = (cd 0 ) 3 . (9)
discrete segments. This brief description covers the essential features
of the model formulation. The governing equations for the droplets Equation (8) is integrated along the gas-phase streamlines to de-
and air-vapor elements are summarized below. A more detailed de- termine to. These streamlines are obtained from the momentum,
scription of the method of solution and the finite difference procedure equation for the air-vapor element.
is given by Palaszewski [11]. The equations are written in dimensional Conservation of Air Momentum. The velocity of an air-vapor
form so that the individual terms can be easily recognized. The di- element is found from a Lagrangian form of the momentum equation.
mensionless forms of the governing equations are given in [11]. These Forces on the element are droplet drag, buoyancy and turbulent shear.
equations are useful in performing parametric studies aimed at im- The air momentum equation can be written as
proving the efficiency of spray units. paVdu/dt = (npaCDird2/8)\v - u|(v - u)
- papTVg(Tdb - Tdb«,) + paVKm(d2uy/dz2)j (10)

Conservation Equations Air Continuity Equation. The continuity equation for the air-
Conservation of Mass for a Droplet. To determine the amount vapor phase is formulated for a rectangular streamtube. The cross-

S16 / VOL. 103, AUGUST 1981 Transactions of the ASME

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 sectional area of the streamtube upwind of the spray is given by the streamlines cross, one obtains the first order corrections to the air-
frontal area of an air-vapor element. This area changes as the element vapor phase along each trajectory. One can now solve the droplet
moves through the spray domain according to: equations again, using the corrected local values of the air-vapor
properties instead of the ambient values. The resulting solutions for
PaliyA = paU„A«, (11)
droplet location, velocity, temperature and size are used to evaluate
where the cross-sectional area of a streamtube is given by: the mass, momentum and energy source terms appearing in the air-
vapor phase equations. Hence, one can obtain the second order so-
A = £m. (12) lution for the gas phase and again assign the new air-vapor properties
The lateral and vertical streamtube dimensions and velocity com- to the droplet trajectories where they cross the gas-phase streamlines.
ponents are obtained by utilizing a ratio of the lateral and vertical This completes the second order solution for the two-phase
components of the air momentum equation as detailed by Palaszewski system.
[11]. The air-vapor streamline itself is found from: The above iterative solution procedure is repeated until successive
iterations on the droplet phase yield values of the rate of heat transfer
dsjdt = u. (13) from the spray that fulfill the following convergence criteria:
Air Energy Equation. Neglecting conduction in comparison with Qk-Qk
turbulent diffusion of heat, one can write for conservation of energy
<£. (19)
along an air-vapor streamline:
A maximum of four iterations are required. The total heat transfer
paVdildt = mrd2h(Td - Tdb) + mrd2LhDpa(ois(Td) - u) from the spray is given by:
+ paCpaVKhd2Tdb/dz2. (14) , PdCpdTT N
- L njlldodoj J- cj&cj J (20)
The terms on the right hand side of equation (14) represent convective 6 j=i
and evaporative heat transfer from the droplets and turbulent dif- A numerical fourth-order Runge-Kutta technique was used to solve
fusion of heat in that order. The enthalpy i of an air-vapor mixture the coupled set of droplet and air-vapor equations. The model can be
is given by: applied at wind speeds above approximately 1.0 m/s. For lower wind
speeds, continuity cannot be satisfied since the spray-air momentum
i - Cpa(Tdb — Tra) + (x>[h/(Twb) + L(Twb) + Cpv(Tdb — Twb)].
interaction causes a reversal of the air-vapor velocities, which indicates
(15) a condition of recirculation.
After substituting equation (15) into equation (14) and dropping small
terms, equation (14) becomes:
Spray Characteristics
paVCpaudTdb/dt = PauVCpudTwb/dt + nird2h(Td - Tdb) The present model is capable of predicting the local air-vapor ve-
+ paCpaVKhd2Tdb/dz2 (16) locity, absolute humidity and temperature fields of a spray unit, in
addition to the droplet trajectories, velocities, temperatures and sizes.
where, To demonstrate this capability, spray field characteristics are shown
for a single spray of the Powered Spray Module (PSM) manufactured
^pau ~ ^pa "•" M^pv (IV) by the Ceramic Cooling Tower Company. A typical spray module is
aligned parallel to the edge of a cooling canal. The module is powered
Meteorological Considerations by a 56 kW pump. Each nozzle sprays approximately 9470 liters of
water per minute. The nozzles are located 0.91 m (3.0 ft) above the
The initial upwind profiles of dry-bulb temperature and absolute
water surface, and produce a spray pattern with a maximum height
humidity can be considered uniform since the variation of these
of 6.1 m and a radius of 6.1 m at the water surface in the absence of
quantities is small over heights typical of sprays (5 m). The wind ve-
wind. The initial droplet velocity is 11.64 m/s and the spray discharge
locity profile is given [13] as:
angle is 74deg [14].
UM) = u,/k' £n\—+ l\ + \p(z) In the theoretical model, droplets were discharged from 36 sites
(18)
zo I around the nozzle perimeter, producing a droplet trajectory every 10
deg.
}p(z) takes on different forms depending on the stability condition
of the atmosphere [13]. Values of the turbulent transport properties The spray was assumed to be monodisperse. The value of the initial
and the turbulent length scale Lmo are given in [13]. While the spray droplet diameter (do = 1.5 cm) was obtained by varying the drop di-
heat and moisture release will affect atmospheric stability, the effect ameter until the single particle momentum equation (3) yielded the
of changes in stability are assumed small compared to convection and experimentally measured trajectory [14].
evaporation in determining drop and air-vapor temperatures. Figures 2 to 5 show model predictions of the detailed air-vapor
droplet flow field of a spray unit. All results are for typical summer
ambient design conditions. The atmosphere was assumed to be ex-
Iterative Solution Technique tremely unstable under strong solar insolation.
To obtain the lowest order solution for the droplet velocities, lo- Figure 2 shows the trajectories of droplets ejected from the PSM
cations, temperatures, and sizes, one considers the local environment at initial angles of d0 = - 8 5 deg and do = 85 deg in the x -y plane for
of the drops to be that of the ambient air-vapor phase. The calculation the condition of low wind. Also shown are a number of air-vapor
of local air-vapor velocities, streamlines, temperatures and humidities streamlines traversing the spray domain near the centerline (y-axis)
can then be performed by using the droplet solutions to evaluate the of the PSM. The effect of the upwind droplets in blocking the air-
droplet source terms appearing in the air-vapor conservation equa- vapor flow is seen by the rising of the streamlines upwind of the nozzle.
tions. These first order air-vapor solutions yield a correction to the Continuity demands that the streamlines rise due to the reduced
local environment of the drops, and are utilized in the next iteration air-vapor velocity and the small lateral forces of the droplets on the
on the droplet phase as detailed below. air (the streamlines are not deflected very much laterally). Downwind
Droplet trajectories and air-vapor streamlines cross at many loca- of the nozzle, the droplets speed up the air and force the streamlines
tions in the spray domain. The velocity, temperature and humidity to descend. Figure 2 also reveals the effect of buoyancy on the
of the air-vapor element as it crosses a given droplet trajectory de- streamline pattern. Near y * = -0.39, the droplets following the up-
scribes the local environment of the droplets at this location. By as- wind trajectory have a horizontal velocity component which is small
signing the first order solutions for the air-vapor properties to loca- compared to the vertical component, since the droplets are falling
tions along each droplet trajectory where the trajectories and almost vertically. One would therefore expect little blockage of the

Journal of Heat Transfer AUGUST 1981, VOL. 103 / 517

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 0.8
T d0 = 51.7'C,T w b ( 0 = 25.6'C
TRAJECTORY
T db0D = 28.9'C
AIR-VAPOR
.— STREAMLINE Uima> - 1.164m/s
z0 = 0.04m, L m o - 2 . 5 m
0.6 - ^'—
—
WIND *• / \ h =-85°(UPWIND TRAJECTORY)
= 85°(DOWNWND TRAJECTORY)
\ So
- DIRECTION / .
\
wo
* 0.4
_
\^ S
s
s—ST """-^

-
/ /
L-—
V —
\
\
^"
\
0.2 _-_ V\ /
-Y--V
-
_A \ l~~
"Vr^~
i if i i i i i i i i \ i
-0.6 -0.4 -0.2 0.0 0.2 0.4 0.6

»§'«
Fig. 2 Air-vapor streamlines near centerline of PSM for low wind (elevation view)

-DROPLET T d 0 = 5 1 . 7 - C , T w b ( D = 25.6'C
TRAJECTORY
Tdb00 = 2 8 . 9 ' C
U2moo = 1.164m/s
0 . 0 4 m , L_„ = - 2 . 5 m

Fig. 3 Air-vapor streamlines initially at a 2m height upwind of PSM (plan view). Low wind

air-vapor flow near the lowest streamline in Fig. 2, and consequently, downwind of the nozzle. Vertical air entrainment can be seen from
little rise of this streamline near y* = —0.39. These expectations are Fig. 2 where the uppermost streamline is lower downwind of the spray
borne out in the figure. However, a t y * = —0.39 the other streamlines, than upwind, indicating the tendency of the spray to pull ambient air
which have not as yet interacted with the spray, are seen to rise sub- downward. This entrainment of ambient air is required to satisfy
stantially. This effect is due to the buoyancy force produced by the continuity. Vertical entrainment results from cooling of the air-vapor
heated air near the lowest streamline which has undergone thermal flow due to turbulent diffusion of heat from the air-vapor elements,
interaction with the droplets neary* = —0.39. as well as the acceleration of the air-vapor flow by the downwind
The streamlines at y * = 0.0 have not seen the downwind droplets. droplets. Lateral entrainment is due exclusively to momentum in-
As stated above, the effect of the downwind droplets is to speed up teraction between air-vapor and droplets. The lateral and vertical
the air and cause the streamlines to descend. However, one notices entrainment by the spray at low wind indicates that more air partic-
the streamlines descending before the air-vapor elements interact with ipates in the local mass, momentum and energy interaction with the
the downwind droplets. This results from turbulent diffusion of heat spray than initially enters the spray domain with the ambient wind.
from the hot air-vapor elements near the top of the spray to the cooler The ability of the model to predict this effect obviates the need for
ambient free stream. empirical entrainment coefficients used by other investigators.
Figure 3 is a plan view of the PSM spray field at low wind. Droplet Figure 4 shows typical absolute humidity profiles near the PSM
trajectories are plotted for half the spray domain. Air-vapor stream- centerline for low wind. Upwind of the spray, the absolute humidity
lines initially at a 2 meter height are also shown. One can readily see profile is uniform. As air enters the spray, it is progressively humidi-
the small effect of drag on the droplets at low wind since the droplet fied, and exhibits the developing profiles depicted in Fig. 4. The lower
trajectories lie essentially in vertical planes. The lateral entrainment boundary condition on the profiles is the saturated humidity at the
of air by the spray is revealed by the compression of the streamlines temperature of the canal water. Dry-bulb temperature profiles, al-

518 / VOL 103, AUGUST 1981 Transactions of the ASME

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 ~,° 0.4 -

Fig. 4 Absolute humidity and air-vapor velocity profiles near centerllne of PSM for low wind

0.8
DROPLET T d 0 = 51.7"C,T wboo = 25.6-C
TRAJECTORY
T db0D = 28.9'C
_ AIR-VAPOR
STREAMLINE UzmoD = 5.82 m/s
B0 =-85-(UPWIND TRAJECTORY) z 0 » 0 . 0 4 m , Lm0 = -13.5m
0.6

*o'9
Fig. 5 Air-vapor streamlines near centerllne of PSM for high wind (elevation view)

though not shown, exhibit similar behavior to the absolute humidity in that the droplets exhibit a pronounced size distribution and are
profiles. ejected at different angles depending on their size and discharge di-
Figure 5 shows PSM droplet trajectories and air-vapor streamlines rection (upwind or downwind). The nozzle was mounted 1.52 m (5.0
for high wind conditions. One observes the substantial distortion of ft) above the pond surface. The flow rate of water through the nozzle
the droplet trajectories at high wind whereas droplet drag on the air was 200.6 1/min (53.0 gal/min) at a nozzle pressure of 51.0 KPa (7.4
causes little deflection of the air-vapor streamlines. Streamline psig). The maximum height of the spray above the pond at this nozzle
spacing is wider downwind of the spray than upwind indicating that pressure and zero wind speed is 4.0 m, and the maximum spray radius
no significant entrainment has taken place. Absolute humidity and at the pond surface is 4.9 m. To collect the sprayed water, catch pans
dry-bulb temperature profiles develop more slowly at higher wind were placed on the pond surface along the longitudinal axis (see Fig.
speeds. The greater air flow through the spray at high wind results 1). Cold spray temperatures were measured in the catch pans for
in local conditions closer to ambient conditions. different wind speeds, ambient psychrometric conditions and droplet
discharge temperatures.
Comparisons with Data Five discrete droplet diameters from the measured size distribution,
The present model was compared with data [15] obtained for the their discharge angles, and the mass fraction of the total spray flow
Spraco 1751 nozzle in an outdoor laboratory facility located at the rate for each size, were used as input to the model [11].
Richmond Field Station, University of California at Berkeley. The In the model, droplets were discharged from ten sites around the
spray produced by this nozzle differs from that produced by the PSM nozzle perimeter. All five droplet sizes were discharged from each site.

Journal of Heat Transfer AUGUST 1981, VOL 103 / 519

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 0.8
Tdo = 46.7«C u
Zma, = l.OOm/s
T
<">a, 22.7'C *-mo = - 2 . 5 m
= 0.04 m
wbco = 18.9°C
T 'o

0.6
//

i// -a WIND

0.4 A AV /
/
/
/ DIRECTION
PRESENT THEORT
/ / LOCAL INTERACTION

t-
- °K\ O
S
/
/ t
6
NO INTERACTION
REFERENCE [ 1 5 ]
\ / O MEASURED DATA
0.2 -- \ A / A DATA CORRECTED FOR HEAT
/ TRANSFER FROM CATCH PAN
\
\
- VERTICAL ARROWS • • POINT
N ^ y
TO MODEL PREDICTIONS FOR
d 0 | * 0.254cm ( MEAN DROP
•-•.___ , DIAMETER)
i 1 1 ... 1 i i t i .... • ...._ i .— i
06 0.4 0.2 0.0 -0.2 -0.4 -0.6

Fig. 6 Centerline droplet return temperature for Spraco 1751 nozzle (U2mco = 1-00 m/s)

T
do = 45.1'C U 2 n l ( i ) - 2.88 m/s
T
1b„ = 17.2'C Lm0 = -3.5m
= 15.6"C 20 ' 0.04m

0.6
A
-
«3 WINO
DIRECTION
I-
3 i,'/
0.4 A \ / / PRESENT THEORr
I
LOCAL INTERACTION
\ / / NO INTERACTION
A o REFERENCE [ 1 5 ]
\ O MEASURED DATA
\ A DATA CORRECTED FOR HEAT
u.<:
\ TRANSFER FROM CATCH PAN
\
\ VERTICAL ARROWS • • POINT
\ S TO MODEL PREDICTIONS FOR
doi " 0.254cm ( MEAN DROP
DIAMETER I
1 i ... i i i i i 1 1 1 1
0.6 0.4 0.2 0.0 -0.2 -0.4 -0.6

»5'«
Fig. 7 Centerline droplet return temperature for Spraco 1751 nozzle (U2mco = 2.88 m/s)

The discharge velocity voi, for each droplet size is given by: ~ Twt,a)l{T,i0 — Twb°>)) of the water collected by the catch pans versus
dimensionless distance (y* = yg/vo2) along the spray centerline in the
Uo; = tWsin 0; (21) wind direction (y-axis). Additional comparisons of model predictions
where with data can be found in [11]. Also shown is the same data corrected
for heat and mass transfer losses from the water in the catch pans
Voz ** initial vertical component of droplet velocity.
[15].
The initial vertical velocity component of the droplets discharged by The no-interaction theory, depicted by the dashed curves, gives the
ramp-bottom swirl nozzles (e.g., Spraco 1751) is a constant for all maximum amount of cooling that can be experienced by the droplets
droplet sizes at a given sprayflowrate, and is found empirically. For under the given ambient air conditions and wind speed. That is, the
the Spraco 1751 nozzle discharging 200.6 1/min, voz = 8.093 m/s, as environment of the drops during their entire time of flight is ambi-
given by Chen [15]. ent-air absolute humidity, dry-bulb temperature and velocity. The
For all model predictions, the atmosphere was assumed to be un- no-interaction predictions were obtained from the lowest order
stable under strong solar insolation. The surface roughness was set droplet solutions in the computer generated results.
at zo = 0.04 m (30 cm high weeds). These assumptions were made in In general, the local interaction theory shows good agreement with
the absence of any specification of atmospheric turbulence conditions the corrected data, while the no-interaction theory predicts more
in the data. cooling than the measured data. Since the local air in the spray domain
Figures 6 and 7 are example comparisons between predicted and is progressively heated and humidified as it passes through the spray,
measured values of dimensionless cold spray temperature (Tc* = (Tc it is expected that the droplets do not experience the amount of

520 / VOL. 103, AUGUST 1981 Transactions of the ASME

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 cooling predicted by the no-interaction theory. This is indicated by Acknowledgment
the data. This work was supported by the National Science Foundation—
Equal size droplets ejected upwind and downwind from the nozzle Grant No. ENG. 78-20709. Appreciation is extended to the City
at the same discharge angle will not be cooled by equal amounts. The University of New York Computer Center, for the use of their facili-
droplet ejected upwind will experience a large relative velocity to the ties. The above work was done in partial fulfillment of the require-
wind and see air close to ambient conditions. A large relative velocity ments of the Ph.D. degree of S. Palaszewski from the School of En-
translates into a large Reynolds number and greater convective and gineering of the City College of the City University of New York.
evaporative heat transfer from the upwind droplet. Local air close to
ambient conditions also results in greater heat transfer from the up-
wind droplet. The downwind droplet experiences a smaller relative References
velocity to the local wind and substantially heated and humidified 1 Ryan, P. J., and Myers, D. M., "Spray Cooling: A Review of Thermal
local air. It therefore experiences less cooling. Since the downwind Performance Models," Proceedings of the American Power Conference, Vol.
droplets were discharged at a lower angle, one would expect even less 38,1976, pp. 1473-1481.
cooling than for equal upwind and downwind discharge angles. This 2 Soo, S. L., "Power Spray Cooling—Unit and System Performance,"
ASME Paper No. 75-WA/Pwr-8, Nov. 1975.
results from the shorter time of flight at smaller discharge angles. The 3 Elgawhary, A. M., "Spray Pond Mathematical Model for Cooling Fresh
up and down arrows in Figs. 6 and 7 show the cooling experienced by Water and Brine," Ph.D. Dissertation, Oklahoma State University, Stillwater,
a droplet of initial diameter equal to 0.254 cm (mean size drop). In Aug. 1969.
each case, the downwind drop experienced less cooling (higher T c *) 4 Frediani, H. A., and Smith, N., "Mathematical Model for Spray Cooling
Systems," ASME Journal of Engineering for Power, Vol. 99, No. 2, Apr. 1977,
than the upwind drop. It may also be noticed that the points at which pp. 279-283.
the droplets strike the pond surface are shifted to the right (against 5 Berger, M. H., and Taylor, R. E., "An Atmospheric Spray Cooling
the wind) if one takes local interaction into account. This effect is due Model," Environmental Effects of Atmospheric Heat/Moisture Releases, The
to the blockage of incoming air by the spray, slowing down the air due Second AIAA/ASME Thermophysics and Heat Transfer Conference, Palo Alto,
Calif., May 1978, pp. 59-64.
to momentum exchange with the droplets. The decreased drag on the 6 Porter, R. W., and Chen, K. H., "Heat and Mass Transfer from Spray
drops resulting from the lower air velocity allows the drops to strike Canals," ASME JOURNAL OF HEAT TRANSFER, Vol. 96, No. 3, Aug. 1974, pp.
the pond further upwind. 286-291.
7 Chen, K. H., and Trezek, G. J., "Spray Energy Release (SER) Approach
to Analyzing Spray System Performance," Proceedings of the American Power
Conclusions Conference, Vol. 38,1976, pp. 1435-1448.
A three-dimensional model that examines the detailed two-phase 8 Chaturvedi, S., and Porter, R. W., "Air-Vapor Dynamics in Large-Scale
flow field of a spray unit has been presented. The model is capable Atmospheric Spray Cooling Systems," Journal of Fluids Engineering, Vol. 100,
No. 1, March 1978, pp. 65-72.
of predicting the local properties of both the air-vapor phase and the 9 Porter, R. W., Jain, M., and Chaturvedi, S. K.( "Unit Thermal Perfor-
drops in the entire domain of the spray without recourse to design- mance of Atmospheric Spray Cooling Systems," ASME JOURNAL OF HEAT
dependent empirical factors. TRANSFER, Vol. 102, No. 2, May 1980, pp. 210-214.
The model was used to generate solutions for the spray character- 10 Ranz, W. D., and Marshall, W. R., Jr., "Evaporation from Drops,"
Chemical Engineering Progress, Vol. 48, No. 3, February 1952, pp. 141-146,
istics of a single PSM spray. At low wind (Uim„ = 1.16 m/s), signifi- 173-180.
cant entrainment of ambient air by the spray takes place, while at high 11 Palaszewski, S. J., "An Air-Vapor-Droplet Local Interaction Model for
wind (t/2m«= = 5.82 m/s) less air interacts with the spray than initially Spray Units," Ph.D. Dissertation, The City University of New York, New York,
enters the spray domain. Theoretical predictions of the model com- 1980.
12 Wallis, G. B., One-Dimensional Two-Phase Flow, McGraw Hill, New
pare favorably with experimental data on centerline droplet return York, 1969.
temperature for the Spraco 1751 nozzle. 13 Panofsky, H. A., "The Atmospheric Boundary Layer Below 150 Meters,"
The new model for spray units can be extended to determine the Annual Review of Fluid Mechanics, Vol. 6,1974, pp. 147-177.
cooling performance of a pass of sprays aligned in the wind direction. 14 Wilson, D. C, "Thermal Performance of Powered Spray Modules,"
Masters Thesis, University of Illinois, Urbana-Champaign, 1972.
The performance of an entire spray canal can be obtained by incor- 15 Chen, K., "Heat and Mass Transfer from Power Plant Spray Cooling
porating the model for a pass of sprays into a system model. Ponds," Doctoral Dissertation, University of California, Berkeley, 1975.

Journal of Heat Transfer AUGUST 1981, VOL. 103 / 521

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 1
Natural Convection from a
B. Farouk
Graduate Student.
Horizontal Cylinder—Laminar
S. i. Giigeri
Assistant Professor. Regime
Mechanical and Aerospace Engineering
Department,
A finite-difference numerical method has been adopted to generate flow patterns and
University of Delaware, heat transfer characteristics for laminar, steady-state, two-dimensional natural convec-
Newark, Del. 19711 tion around a circular cylinder submerged in an unbounded Boussinesq fluid. The ap-
proach allows the use of nonuniform as well as uniform specified temperature and heat
flux distributions over the cylindrical surface. Part of the results are generated for reverse
convective flows with recirculation zones which occur when part of the cylinder is below
the ambient temperature while the remaining part is above. The results for uniform tem-
perature boundary condition are in good agreement with the experimental data and other
solutions available in literature.

Introduction
Natural convection from a horizontal cylinder has been the subject As in the present work, no boundary layer approximations were
of several research studies. This configuration has applications over made in the finite element approach of Holster and Hale [15] and in
a wide range of engineering problems. Empirical relations, which are the finite difference approach of Kuehn and Goldstein [16]. In the
based on extensive experimental data have been presented by latter, the numerical solutions are compared with extensive experi-
McAdams [1]. Temperature fields around heated single horizontal mental data for the temperature and velocity fields as well as the
cylinders have been measured by interference fringes [2]. Earlier, Nusselt number distribution.
photographs based on the same technique, illustrating the tempera- This study is primarily concerned with the natural convection heat
ture field around horizontal cylinders in laminar flow have been ob- transfer from a horizontal circular cylinder with uniform and non-
tained by Eckert and Soehngen [3]. Attempts were also made to an- uniform boundary conditions, placed in an unbounded Boussinesq
alytically predict the local and overall heat transfer rates from the fluid. The surface temperature distribution is considered in such a
surface of heated, horizontal cylinders. Analytical studies are mostly way that the resulting buoyancy forces vary in magnitude around the
based on the boundary layer approach. In general, these studies cylinder. If this variation is large enough, flow reversal may occur.
concern the nature of the portions of such flows attached to the sur- These temperature variations are frequently observed in solar energy
face. The subsequent regions where the flow may separate from the applications such as the Thermal Wall Panel passive heating concept
surface and rise as a buoyant plume were not considered in detail. One [17,18], where the thermal storage materials are contained in hori-
of the earliest works in this area was by Hermann [4] who obtained zontal cylinders and partially subjected to solar irradiation. Due to
a solution for the differential equations under the assumption of a thin the complex flow fields, boundary layer approximations lose their
(as compared with the cylinder diameter) heated laminar layer for validity and the complete Navier-Stokes and energy equations need
diatomic gases and small temperature differences. Due to the to be solved. The special case of isothermal cylinder solutions are
boundary layer approximation, his results are valid for moderately compared with available data in literature to verify the accuracy of
high Grashof numbers (104 to 3 X 108). Merk and Prins [5] presented the approach.
analytical results of the heat transfer around horizontal cylinders and
spheres. Chiang, et al. [6, 7] used a Blasius series expansion method Problem Statement
to predict the heat transfer coefficients for laminar, natural convection The steady-state, two-dimensional laminar natural convection from
from horizontal cylinders and spheres. Nakai and Okazaki [8] studied horizontal cylinders is governed by the three conservation principles,
free convection from horizontal wires at uniform temperatures for the viz., the conservation of mass, the conservation of momentum and the
case when the thickness of the boundary layer around the cylinder conservation of energy. The equations describing these principles are
is sufficiently large compared with its diameter. The near and far
as follows:
fields around the cylinder were considered separately. A compre-
Conservation of mass,
hensive review of boundary layer flow with body forces including
nonisothermal surfaces is given by Ostrach and Moore [9]. An account dvr vr 1 dv«
+- + - = 0 (1)
is given of the boundary layer analysis of the vertical flat plate prob- dr r dd
lem with prescribed heat flux arid the vertical cylinder where the
conservation of momentum, r • direction,
surface temperature is permitted to vary linearly with the axial
coordinate. In the latter case a similarity solution becomes possible. dvr vedvr
Churchill and Chu [10] presented correlating equations for both dr dd ' or
laminar and turbulent flow conditions, based on boundary-layer
d2vr 1 du r 1 d2vr 2 duo
theory and compilation of experimental data. Kuehn and Goldstein -M + -+- • gpfi(T' - T„) cos 0
[11] gave expressions for average Nusselt numbers over a wide range or2 dr d0 2 r2 r2 dd
of Rayleigh numbers as a special case of correlating equations for (2)
natural convection between concentric horizontal cylinders.
conservation of momentum, ' - direction,
Numerical solutions of the boundary layer equations for natural 1 dp
dve vedvp vrve
convection around horizontal cylinders have been obtained by Muc-
ogulu and Chen [12], Merkin [13] and Fujii, et al. [14]. dr r dd r rdd
d2Vg 1 dvg 1 d2vg Vg 2 dvr
+ fi + gpfi(T' -T„) sinfl
1
Presently, Assistant Professor, Drexel University, Philadelphia, Pa. dr 2
r dr r2 dd2 r2 r2 d0
Contributed by the Heat Transfer Division for publication in the JOURNAL
OF HEAT TRANSFER. Manuscript received by the Heat Transfer Division, (3)
August 11,1980.

522 / VOL. 103, AUGUST 1981 Transactions of the ASME

Copyright © 1981 by ASME
Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 conservation of energy, a. Boundary Conditions on the Surface of the Cylinder.
Temperature is specified over this boundary section. For the purpose
[ dT ve dT] , fd 2 T 1 dT 1 d2T
P (4) of demonstrating the effects of nonuniform temperature distribution,
[ dr r d0\ [dr2 r dr r 2 d0 2 an artificial temperature generation function is formulated as
These equations can be put into a more convenient form by using the
definitions for the stream function and vorticity as T\R = V T o 2 ^ e 2 cos 2 0 - T „ + e sin0 (10)
With this expression, the temperature distribution can be envisioned
(5) as being proportional to the spacing between two concentric/eccentric
pr d6 p dr
circles, having radii proportional to To and T „ , respectively. When
and the eccentricity factor e is zero, the case corresponds to an isothermal
dvr cylinder. It is possible that for sufficiently large values of e, the two
-(rus) • (6) circles intersect each other, indicating that the cool side of the cylinder
dr "B7 is at a lower temperature than the ambient while the remaining sec-
Also due to variable temperature boundary condition, it becomes tion is higher.
impossible to define a meaningful Grashof number; therefore, the Since the cylinder surface is assumed to be nonpermeable, it follows
expressions are not non-dimensionalized. Substituting the expressions that for the stream function
given by equations (5) into equation (6), the following expression is
obtained: \p\R=0 for 0 < 0 < 1-K (11)
r d\p The boundary condition for the vorticity is not quite so easy to specify
(7) since it requires information on the gradients of velocities which are
dr [p dr d0 [pr d0
not known a priori. An expression for the vorticity boundary condition
The two momentum equations, i.e., equations (2) and (3) can be re- can be obtained by expanding the stream function near the wall, using
duced to a single equation by using the definitions of stream function a three term Taylor series and making use of the continuity and the
and vorticity. Following substitutions, equation (2) is differentiated no-slip condition. It is given as
with respect to 8 and equation (3) is differentiated with respect to r.
After adding and simplifying, -2$
co|B (12)
I d\l/\ d ( diMl fd j dQuio)\ p(Ar)2
co— co— - — i
\ ddl d0\ drj\ [dr\ dr ) where \p is the value of the stream function at a short distance Ar into
the fluid. An alternate boundary condition for vorticity at the wall
d /ld(/tto)y „ dT .
0 (8) which is based on the assumption that the gradients with respect to
+ d0[r cost/ H
d0 j . r sin I
0 are negligible in the vicinity of the cylinder surface is presented in
dd dr
Also the energy equation with proper substitutions becomes. the Appendix. The formula for the vorticity at the wall is obtained
by expressing its value in terms of the dependent variable values at
[M r ^)_A( T ^)MA(A^ + A(AI^l = 0 (9) and adjacent to the cylinder surface.
dr \ ddj dd \ drj\ [dr \cp drj d6 \cp r d8)\ b. Boundary Conditions at the Far Field. The necessity to limit
The advantage of the stream function—vorticity formulation is that the size of the solution domain requires that a pseudo boundary be
the pressure variable is eliminated and the number of equations to defined in the far field. To establish these conditions, a procedure
be solved is reduced to three. The solution to equations (7,8) and (9), suggested in reference [15] is followed. This approach considers the
with appropriate boundary conditions, yields the desired distribution horizontal cylinder as a pump whereby the pumping is provided by
of \p, co and T in the flow field. the buoyancy forces. Any realistic model must allow for fluid to cross
the pseudo boundary. It is assumed that the velocity component in
Boundary Conditions the circumferential direction, along the pseudo boundary is zero
Solution of elliptic equations requires that the conditions be implying that in the far field, all flow must be in the radial direction.
specified along the entire boundary which encloses the flow field. In If the domain is a large circle, the stream lines become normal to the
the present problem, two of the most important boundary sections boundary hence the boundary conditions for the stream function and
are the cylindrical wall and a pseudo boundary sufficiently far away vorticity become
from the cylinder. A third section may be the vertical symmetry plane d\p dco
which exists only for symmetrical wall temperature or heat flux = 0; — = 0 0 < 6 < 2-K (13)
dr B» dr
boundary conditions. However, in this study, even for such boundary
conditions the entire flow field is considered and the need for the and temperature of the fluid drawn into the flow field is the same as
boundary conditions on the symmetry plane is eliminated. The two the ambient temperature. However, a generalization along the entire
boundary sections are handled as follows: boundary cannot be made since the temperature distribution along

-Nomenclature-
cp = specific heat at constant pressure
e = eccentricity factor for temperature dis- R = radius of circular cylinder 0 = circumferential (angle) coordinate
tribution Ra = Rayleigh number (Gr • Pr) p. = dynamic fluid viscosity
g = gravitational acceleration R a = radius of pseudo boundary p = density
Gr = Grashof Number, 8p2gPRa(T0 - T-) T' = local temperature ip = stream function
/n2 Tm = mean temperature at the surface co = vorticity component parallel to cylinder
GrOT = mean Grashof Number, 8p2g/3Rs(Tm To = ambient temperature
T„)/p2 To = reference temperature as used in
k = thermal conductivity of fluid equation (10) Subscripts
Nu = Nusselt number T = T' - T . N,S,E,W,NE,SE,NW,SW = compass point
p = pressure vr = radial velocity component location of nodes with respect to node P
Pr = Prandtl number ue = circumferential velocity component n,s,e,w = compass point locations of the in-
r = radial coordinate /? = thermal coefficient of volume expan- tegration cell

Journal of Heat Transfer AUGUST 1981, VOL. 103 / 523

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 the section of the boundary where the plume crosses, is not known which is calculated by using a two-point difference scheme for the
a priori. It is assumed that the temperature gradient normal to the temperature gradient at the wall. An important parameter for the
pseudo boundary is zero implying that the heat carrying action of the finite difference method described herein is R«,/R, the ratio of the
fluid is done purely by convection rather than by conduction, i.e., radius of the pseudo boundary to the cylinder radius. A ratio of R^IR
= 8 was used in obtaining the results presented for variable temper-
dT
•0; 0 < 8 < 2p (14) ature boundary condition because the local and mean Nusselt num-
dr bers become independent oiR^IR when R„IR > 6.
For uniform temperature boundary condition consistent with
Solution Procedure reference [16], it was observed that for larger Rayleigh numbers the
The equations of the previous sections are transformed into dif- boundary layer becomes thinner necessitating a finer mesh close to
ference equations by using a finite-difference method in polar coor- the boundary but allowing smaller R„/R to be adequate. When R „/R
dinates, presented by Gosman, et al. [19] and recently applied to is decreased without changing the number of nodal points in the r-
forced flow across tube bundles by Launder and Massey [20]. A suc- direction, it automatically provided a finer mesh. For Ra < 104, R^/R
cessive-substitution technique is employed to solve the finite differ- = 8 was used but for larger Ra, it was gradually decreased to 4 for the
ence equations. Instead of using standard Taylor series expansions, case of Ra = 107. All results were generated by considering air as the
the finite difference equations are obtained by integration over finite fluid with Pr = .721. Figure 2 shows the distribution of the local
cells. After creating a mesh in r and 0, these cells are obtained around Nusselt number and its comparison with those obtained by other
each nodal point P through the half-distance points (Pig. 1). This researchers. A 5°C difference was used between the uniform wall
method ensures that the conservation laws are obeyed over arbitrarily temperature and the ambient temperature. The agreement is quite
large or small portions of the field and it lends itself better to physical satisfactory. Figure 3 shows the effect of the Rayleigh number on the
interpretation and hence to ease of understanding. Since the details angular distribution of the Nusselt number for cylinders with uniform
of the procedure are given in reference [19], only the integration of wall temperature. Figure 4 indicates the variation of the average
the source terms will be presented briefly. The integral of the source Nusselt number with Rayleigh number and its comparison with data
term for T is zero. For \p, following the assumption that co is uniform from [16]. Temperature and flow field plots are not presented for the
over the integration cell, isothermal case since they are given in [16] along with comparison of
experimental and numerically calculated temperature fields.
IL = up(rE - rw)(8N - ds)rp/4 (15)
For vorticity, to, the source term for equation (8) becomes Agreement of our results with those presented in [16] is satisfac-
tory.
dT
J "-«
rw J OB
n«n
1
dT
r — sin0 dd dr
dr
The main advantages of the present approach are that it can easily
take into account the nonuniform boundary conditions and flows with

rs d6
cos0 dd dr

The above equation can be integrated by parts. After integrating once,

(16)
recirculation. The results for natural convection from a cylinder with

the temperature T appears as a multiplier within the integrand; it is

then assigned the value Tp. Also noting that,

cos0„ c; (COS@N + cosdp)/2, etc.,

the discretized form of the source term integral for vorticity be-
comes

'"or = ^pKcosfljv - coB6s)l(rE + rP)(TE + TP)

8
- (rP + rw)(Tw + TP) - 2TP(rE - rw)]
- (rE ~ rw)[(cos8N + cos8P)(TN + TP)
- (cos0 s + cos0 P )(Ts + TP) - 2Tp(cos0 N - cos0 s )]| (17)
These results, along with similar derivations for the diffusive and
convective terms can be organized in such a way that each variable
can be expressed as a successive substitution formula [19]. Fig. 1 Typical node " P " with Its surrounding integration cell
In the point iteration process employed here, the grid is system-
atically scanned node by node and the successive substitution formula 0.80
is used to update the values of the variables (\p, co, T) at each node.
The iteration sweep for the region 0 < 0 < ir was carried out in the .
anti-clockwise direction and for the region w < 8 < 2ir it was carried
out in the clockwise direction. When a symmetry plane is not con- 0.60 **<•„»««
sidered in the flow domain, the above sweep directions were found >—••"""i
to give better convergence. For a typical problem, with zero initial
distributions of the dependent variables, converged solutions were
-O-
I
^ ^

.1u
obtained after about 400 iterations requiring a total of about 30 Siv®
minutes of CPU time on a Burroughs' 7700 computer.

Results and Discussion

The nonlinear coupled partial differential equations are solved by 0.20 — Presant Method Ro"IO° ,
considering a grid of 41 X 100 (r X 0) for the entire domain. The node ® Kuehn and Goldstein RaBK)5(Exparim&ntal) [ie] ^k,
spacing in the r -direction is reduced by a factor of 2 for R < r < (R A Kuahn and Goldstein Ra"IO°(Nunwte<il) [l^l ^>
+ R«,)/2 region as compared to the remaining region to obtain a better Hals and Hohtor Ra-3.6xl0 4 &*] I
accuracy in evaluating the local Nusselt number given as 0.00 . I , i , I . I , I .
30 60 90 120 I50 ISO
2R dT I Angle (degrees)
Nu(0) (18)
T\R dr Fig. 2 Distribution of local heat transfer rates for an isothermal cylinder

524 / VOL 103, AUGUST 1981 Transactions of the ASME

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 0.80 r

Ra»72xl0*

Ra» 3.6x10
Non-uniform wait temp.
Mln. wall lemp • Ambient temp.
Ro'10*

120 150 180

Angle (degrees)
Fig 3 Influence of Raylelgh number on local heat transfer coefficients for convection with prescribed nonuniform surface
an Isothermal cylinder temperature, e = 5.0, T0 = 23.85°C, T= 18.85°C

30
— Present Method
o Kuehn and Goldstein [i6]

20

10

10* io- 10" 10° 10'

Ra
Fig. 4 Average Nusselt number as a function of Rayleigh number for an
isothermal cylinder
prescribed nonuniform wall temperature are investigated with two
cases, where the degree of nonuniformity is varied by the eccentricity
factor e as introduced earlier. Figures 5 and 6 show the temperature
and stream function distribution in the case where the difference Fig. 6 Streamlines for the case of Fig. 5
between the ambient and surface temperatures vanish at the 6 = 3ir/2
point. The surface temperature distribution is given by equation (10)
with e = 5.0°C, T 0 = 23.85°C and T„ = 18.85°C. The mean Grashof
number is found to be Gr m = 50,000. Comparing Figs. 5 and 6 with
figures in [16], it is observed that the plume shifts to the right because
of higher surface temperature of the cylinder on this side. Since no
point on the cylinder surface is colder than the ambient temperature,
no recirculation was observed for this case. Recirculation was not
found for the isothermal cylinder as well. In the next example, a value
of e = 15.0°C was-used along with the same values of To and T» used
previously. The resulting surface temperature is such that reverse
buoyancy forces occur on the two sides of the cylinder. On the cylin-
drical surface the largest positive and negative temperature differ-
ences with the ambient are 20°C and - 1 0 ° C at the 8 = ir/2 and 8 =
3ir/2 locations respectively. The isotherms for this case are presented
in Fig. 7. The streamlines in Fig. 8 indicate that recirculation is pro-
nounced. For the above two examples, a meaningful definition of the
Nusselt number becomes difficult to make. The local Nusselt number
as given by equation (18) goes to infinity in regions where the tem-
perature difference between the surface and the ambient vanishes.
The Nusselt number may again vanish when d T / d r | # becomes close
to zero. The polar plots of the local Nusselt number distribution for
the above two cases as well as for the uniform surface temperature case
are given in Fig. 9 for the purpose of comparison.
With this study, the use of a finite difference technique [19] in Fig. 7 Isotherms for natural convection with prescribed nonuniform surface
solving natural convection problems is demonstrated. The results temperature, e = 15.0, T0 = 23.85°C, T = 18.85°C

Journal of Heat Transfer AUGUST 1981, VOL. 103 / 525

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 agree well with the available data and other solutions. The method
presented works efficiently and quite satisfactorily in analyzing the
regions such as the plume and recirculation zones if they exist. Fur-
thermore, nonuniform boundary conditions which occur in many
engineering problems can be accounted for. The present method also
provides far more detail about the flow than is usually obtainable in
an experimental study.

l*=.000l
Acknowledgments
^=-J0003 The authors thank Mr. Joseph Quigley for his expert help in com-
puter graphics and Mr. Mark Deshon for his carefully done draw-
ings.
References
1 McAdams, W. H., Heat Transmission, 3d. ed., McGraw-Hill, New York,
1954.
2 Kennard, R. B„ Natl. Bur. Standards J. Research, 8,787,1932, "Tem-
perature, Its Measurement and Control in Science and Industry," Reinhold,
New York, 1941, pp. 685-706.
3 Bckert, E. R. G., and Soehngen, E., "Studies on Heat Transfer in Lam-
Fig. 8 Streamlines for the case of Fig. 7 inar Free Convection with Zehnder-Mach Interferometer," USAF Tech. Report
5747, Dec. 1948.
4 Hermann, R., "Heat Transfer by Free Convection from Horizontal
Cylinders in Diatomic Gases," NACA TM 1366,1954.
5 Mark, H. J., and Prins, J. A., "Thermal Convection in Laminar Boundary
Layers, I, II, III," Applied Scientific Research, Vol. 4A, 1953-54, pp. 11-24,
195-206, 207-224.
6 Chiang, T., and Kay, J., "On Laminar Free Convection From a Hori-
zontal Cylinder," Proceedings of 4th National Congress of Applied Mechanics,
1962, pp. 1213-1219.
7 Chiang, T., Ossin, A., and Tien, C. L., "Laminar Free Convection From
a Sphere," ASME JOURNAL OF HEAT TRANSFER, Vol. 86, 1964, pp. 537-
542.
8 Nakai, S. and Okazaki, T., "Heat Transfer From a Horizontal Circular
Wire at Small Reynolds and Grashof Numbers, Pts. I and II," International
Journal of Heat Mass Transfer, Vol. 18,1975, pp. 387-413.
9 Moore, F. .,K. (ed), Theory of Laminar Flows, Vol. 4, Princeton Uni-
versity Press, 1964.
10 Churchill, S. W., and Chu, H. H. S., "Correlating Equations for Laminar
and Turbulent Free Convection from a Horizontal Cylinder," International
Journal of Heat Mass Transfer, Vol. 18,1975, pp. 1049-1053.
11 Kuehn, T. H., and Goldstein, R. J., "Correlating Equations for Natural
Convection Heat Transfer Between Horizontal Circular Cylinders," Interna-
tional Journal of Heat Mass Transfer, Vol. 19, 1976, pp. 1127-1134.
12 Mucoglu, A., and Chen, T. S., "Analysis of Combined Forced and Free
Convection Across a Horizontal Cylinder," The Canadian Journal of Chemical
Engineering, Vol. 5, June 1977, pp. 265-279.
13 Merkin, J. H., "Free Convection Boundary Layers on Cylinders of El-
liptic Cross-Section," ASME JOURNAL OF HEAT TRANSFER, Vol. 99, Aug.
1977, pp. 453-457.
14 Fujii, T., Fujii, M., and Matsunaga, T., "A Numerical Analysis of Lam-
inar Free Convection Around an Isothermal Horizontal Circular Cylinder,"
Numerical Heat Transfer, Vol. 2,1979, pp. 329-344.
Fig. 9 Polar plots of the local Nusselt number distributions for natural con- 15 Holster, J. L., and Hale, L. A., "Finite Element Simulation of Transient
vection with e = 0.0, e = 5.0 and e = 15.0 Free Convection From a Horizontal Cylinder," ASME Paper No. 79-HT-49.
16 Kuehn, T. H., and Goldstein, R. J., "Numerical Solution to the Nav-
ier-Stokes Equations for Laminar Natural Convection About a Horizontal
Isothermal Circular Cylinder," International Journal of Heat Mass Transfer,
23,1980, pp. 9717979.
17 Giiceri, S. I., and Faunce, S. F., "Modeling of a Thermal Wall Panel Using
Phase Change Material," ENERGY—The International Journal, Vol. 4, No.
4, Sept. 1979.
0.60 18 Faunce, S. F., Sliwkowski, J. J., and Giiceri, S. I., "An Application of
PCM in Passive Solar Systems," International Solar Energy Society Annual
Meeting, Atlanta, May 1979. Proceedings.
19 Gosman, A. D., Pun, W. M., Runchal, A. K„ Spalding, D. B., and
Wolfshtein, M. W., Heat and Mass Transfer in Recirculating Flows, Academic
i 0.40 - Press, London, 1969.
20 Launder, B. E., and Massey, T. H., "The Numerical Prediction of Viscous
Flow and Heat Transfer in Tube Banks," ASME JOURNAL OF HEAT TRANS-
FER, Vol. 100,1978, pp. 565-571.

0.20 -
APPENDIX
An alternative to equation (12) can be derived as boundary condi-
0.00 tion for to over the cylindrical surface by noting that the gradients near
the boundary, along the 0-direction are negligible, when compared
Angle (degrees) with those in the direction perpendicular to the surface. With this
assumption, integrating equation (9) twice, substituting into equation
Fig. 10 Comparison of the distribution of local heat transfer rates for an
isothermal cylinder with vorticity boundary conditions given by equation (12) (8) for the temperature, integrating the resulting equation twice and
and equation (20) rearranging,

526 / VOL. 103, AUGUST 1981 Transactions of the ASME

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 lEEl gpfoind n dT where
R2sin6 + R- In r
H r R r dr r 2 + fl2, r Ri- R2, R i
ot = In —I < — In - + -
dw gpPsind dT 4 R 4 2 r
+ co lnfl R2 \nR
dr dr 3r 2 + fl2, r fl3-r3
gpPR2sinf) dT y = In —I
(19) 12 R 9R
dr In equation (20) \p, to and T are evaluated at the near wall node. The
Finally, substituting for the vorticity in equation (7), performing the drawback of this formulation is that it is valid only if the angular
integration from the boundary surface to a point r adjacent in the flow derivatives of the dependent variables are negligible near the wall.
field and making use of the no-slip condition yields A comparison of the local Nusselt number distribution around the
cylinder for the case of Ra = 105 is given in Fig. 10 with both boundary
co , gpPyRhinB T - T\x , f - f\R conditions as given by equations (12) and (20).
— a _i 1
W\R •• •R n r—R p (20)
R
a — K
r-R

ERRATUM
Erratum: C. E. Hickox, "Thermal Convection at Low Rayleigh Number from Concentrated Sources in Porous Media," published
in the May 1981 issue of the Journal of Heat Transfer, pp. 232-236.
Figures 2, 5(a) and 8 were not reproduced properly for publication.

. 0. 125 • 2/70.

Fig. 2 Steady-state streamlines and isotherms Fig. 5( a) Transient streamlines for a point source Fig. 8 Transient streamlines at one year for
for a point source at the base of a semi-infinite located below a permeable boundary (Ra = 0.5, sub-seabed disposal of nuclear waste (Ra = 10~3,
region ( t = k 1/2 ) T = 0.1. L= D.yj/iX 103 = 0.25, 0.5, 1, 2, 3, 4, T = 4.8 X 1015, i. = D ; f X 10~4 = 1, 2, 3, 4, 5,
4.5) 6, 7, 8, 9, 10, 12, 14)

Journal of Heat Transfer AUGUST 1981, VOL. 103 / 527

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 ERRATUM
Erratum: C. E. Hickox, "Thermal Convection at Low Rayleigh Number from Concentrated Sources in Porous Media," published
in the May 1981 issue of the Journal of Heat Transfer, pp. 232-236.
Figures 2, 5(a) and 8 were not reproduced properly for publication.

10 10
Fig. 2 Steady-state streamlines and Isotherms Fig. 5{ a) Transient streamlines for a point source Fig. 8 Transient streamlines at one year for
for a point source at the base of a semi-Infinite located below a permeable boundary (Ra = 0.5, sub-seabed disposal of nuclear waste (Ra = 10-3 ,
region (L = k 112) = = =
T 0.1. L D,1/;1 X 103 0.25, 0.5, 1, 2, 3, 4, T = = =
4.8 X 10 15, L D; 1/;1 X 10- 4 1,2,3,4,5,
4.5) 6,7,8,9,10,12,14)

Journal of Heat Transfer AUGUST 1981, VOL. 103 / 527

Copyright © 1981 by ASME
Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 Transient Natural Confection Heat
L. Robillard Transfer of. Water with iaximum
P. Vasseur
Density Effect and Supercooling
Ecole Poiytechnlque,
Unlverslte de Montreal,
Department of Civil Engineering, Transient two-dimensional laminar convectiue cooling of water in an enclosed rectangu-
Montreal, P.O., Canada, H3C 3A7
lar cavity with a convectiue boundary condition is approached by a numerical method.
The maximum density effect and supercooling are considered in order to understand the
flow and temperature fields in a mass of water cooled to a temperature near its freezing
point. The time dependent flow and temperature fields, mixed mean, center point, maxi-
mum wall temperature and the heat transfer rate on each wall are presented. The compu-
tations are performed mainly for the case of a square cavity but the influence of the aspect
ratio of the cavity is also discussed.

1 Introduction 2 M a t h e m a t i c a l S t a t e m e n t of t h e P r o b l e m
Many studies on natural convection heat transfer related to the A geometry of the problem is given in Fig. 1. The aspect ratio of the
cooling of a mass of water near its freezing point have been performed half cavity is denoted by E = (.lb. The initial fluid temperature T; is
in the past [1-3]. Those studies were undertaken because of the im- assumed to be uniform and at time t = 0 the cavity is suddenly ex-
portance of water in atmospheric and terrestrial problems. Further- posed to a constant ambient temperature maintained at T„ < 0°C.
more, the fact that water is affected by a density anomaly near 4°C The resistance and capacitance of the cavity walls are neglected. The
[4] introduces an additional complexity in the study of the transient cavity is supposed to be sufficiently long in the direction normal to
natural convection heat transfer, the latter case being considerably the plane of Fig. 1 to assume a two-dimensional motion. It is also as-
different from that of a fluid without maximum density [5], sumed that the motion is laminar. The governing equations for the
The effect of maximum density on the free convection induced by transient natural convection, using the Boussinesq approximation
an isothermal vertical flat plate adjacent to a mass of cold water was [20] and neglecting viscous dissipation and compressibility effects,
first studied by Schechter [6]. The existence of dual motion (both up are then given in nondimensional form as:
and down) in the boundary layer was demonstrated experimentally
df2 dUQ dVSl Igb^dAp „ „„
and theoretically and it was found that the resulting heat transfer — + + = F — — - + PrV 2 fi (1)
coefficient was considerably affected by this particular flow motion. dr dX dY 2
a / £>y
Because of their importance to certain freezing processes, a number de due dve
of theoretical and experimental investigations were conducted on (2)
— + + = V20
convective heat transfer occuring near an isothermal vertical flat plate dr dX dY
[7, 8], between horizontal water layers [9-11], and inside enclosed (3)
cavities [12-17]. In all these investigations it was shown that the rates
of heat transfer by free convection in water at 4°C are usually less than (4)
dY dX
those occuring at other temperatures.
Most of the past investigations concerning the cooling of a mass of where
water have only been performed at temperatures above 0°C. However,
_at ,T_U''
it is well known [3] that water will normally supercool substantially V = ^
below 0°C before the occurrence of ice nucleation. For instance, "b2 ~~ol
measurements by Dorsey [18] on the nucleation temperature in T-T„
X: Y = y- (5)
quiescent tap water show that the typical nucleation temperature is AT b
in the range of —4 to —7°C. The growth of ice inside supercooled water
o>b2 _d_ _d_
results in the formation of dentritic ice which may be responsible, * =t V2:
under certain conditions, for the complete blockage of a water pipe dX 2 dY 2
subjected to a freezing environment [3,13], It is thus desirable to have •pie) hb
a better understanding of the natural convection processes occuring Ap = >- Pr = - Bi:
a
during the supercooling of water.
AT = Ti ~ T-
The purpose of this investigation is to study numerically the
transient two-dimensional laminar convection resulting from the and all other symbols are defined in Nomenclature.
cooling of a mass of water enclosed in a rectangular cavity subjected The initial and boundary conditions are:
to convective boundary conditions. The cooling of water from the
( 7 = V = ^ = r2 = o atr = 0
initial uniform temperature greater than 4°C passing through the
maximum density point and finally in the supercooling range is con-
sidered. The present study is an extension of the works of Cheng, et U=V=*=0 atX = 0
al. [13] and Gilpin [16] concerning cooling in water pipes. Numerical
studies of natural convection inside a rectangular cavity, in the ab- atX = E (6)
sence of maximum density effect, have been carried out in the past d!7
by many investigators; that of Newell and Schmidt [19] contains a = V = V = SI = 0 ^ =0 atY = 0
£>Y' dY
comprehensive bibliography.
(7= V = ^ = o — = - B i 6W atY=l
dY
Contributed by the Heat Transfer Division and presented at the 19th It is noted in the boundary conditions that use has been made of
AIChE/ASME National Heat Transfer Conference, July 27 to 30,1980, Orlando,
Fla. Revised manuscript received by the Heat Transfer Division November 14, the symmetry of the problem with respect to a vertical plane passing
1980. Paper No. 80-HT-74. through Y = 0. The particular case with U = V = 0 in energy equation

528 / VOL. 103, AUGUST 1981 Transactions of the ASME

Copyright © 1981 by ASME
Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 Jx <pT- : Cll»L) dY (10)
THERMAL Jo \dXlx=E
f~ INSULATION
• / / / / / / / 7-/ / ' / / / / / , in which 4>T ' s the dimensionless heat flux relative to that boundary.
Similar results may be obtained for the side and bottom boundaries
4>L and <I>B-
The sum of the heat fluxes across boundaries is related to dw, the
dimensionless temperature averaged over the boundary and to 6, the
dimensionless temperature averaged over the cavity, by:
CONVECTIVE
COOLING WITH
d
h • CONST. [<I>T +<PB + E0L} = Bi (2 + E)BW = \-E (11)
dr
COLD
ENVIRONMENT
Thus the sum of the dimensionless average wall heat fluxes can be
Too evaluated once the average wall temperature 6W is known.
/ / / / / / /JJL
3 Numerical Solution
T o = AMBIENT TEMPERATURE < 0°C
The coupled vorticity transport and energy equations (1) and (2)
Ti = UNIFORM INITIAL WATER
TEMPERATURE > 4° C may be solved by different numerical methods such as standard ex-
Fig. 1 Coordinate system and boundary conditions plicit method, alternating implicit Dufort-Frankel method and others
[24]. In this study, the alternating direction implicit (A.D.I.) method
(2) corresponds to the unsteady heat conduction in an infinite rec- is employed because of its simplicity and unconditional stability re-
tangular slab for which the exact solution is well known [21]. Fur- gardless of time step AT for linear parabolic equation. The first and
thermore the dimensionless time T defined in equation (5) is similar second derivative were approximated by central differences and the
in form to Fourier number in unsteady heat conduction. time derivative by a first order forward difference. The finite differ-
The density-temperature relationship of water can be approxi- ence forms of the equations were written in conservative form for the
mated by the following equation [22, 23]: advective terms in order to preserve the transportive properties.
The specification of computational boundary conditions greatly
Po (7) affects the accuracy of the solution of the vorticity equation. Thus to
Po ;=i
be consistent with the accuracy of the scheme utilized in this study
where the density at 0°C is p0 = 0.9998396 (gr/cm 3 ). The values of the the wall vorticity was specified by a formulation similar to that de-
constants are given in reference [13], for the temperature ranges 0 ~ veloped by Woods [25]. Furthermore it should be noted that, in con-
20°C and - 1 0 ~ 10°C. junction with Woods's form, a consistent formulation of the velocity
The mean temperature T of water is defined as field for the grid points nearest the wall requires the use of a three
AT p rBt points special form obtained by a cubic equation [26]. The velocity
T = To. + - tdXdY (8) components computed in this way extrapolate smoothly to the
E Jo Jo wall.
It is noted that the quantity (T - T „ ) / A T represents the ratio of The elliptic equation (3) was solved by a successive over-relaxation
excess energy based on the initial state. It is also seen that the ratio (S.O.R.) procedure. For the present problem it was found that a re-
(Ti - T ) / A T represents the total heat loss with respect to the initial laxation factor of 1.8 was an optimum value. The iterative procedure
heat content arid attains a value of 1 when the water is cooled to was repeated until the ratio of the maximum change in stream func-
T„. tion occuring in the field as a result of one approximation to the
By assuming a perfect mixing of the fluid at any time, the mixed maximum value of the stream function was smaller than 5 X 10~4. In
mean temperature Tj is given by: most of the calculations presented in this paper it was observed that
the number of iterations required decreases rapidly from 30 imme-
Tb = T«, + AT exp[-TBi(l + 2/E)] (9)
diately after the start of the Cooling to 2-5 for most other time steps.
The time-dependent heat transfer rates on each wall of the cavity It was also found that the number of iterations reaches as high as 20
are of practical interest. The local wall heat flux q can be computed when the maximum density effect appears.
by considering the wall temperature gradient (dd/dN)ff=o, where N The effect of grid size on the accuracy of the present numerical
is dimensionless normal to the boundary. The dimensionless local wall solution has been studied for different mesh sizes. As noticed by Rubel
heat flux, defined as Nu = (qb)/(kAT) = ~(d0/diV)jv=o, may thus be and Landis [27] indications that a given mesh size will become inad-
integrated over a given boundary. For example, integration over the equate as Grashof number increases are most easily given by the ap-
top boundary gives: pearance of local temperature peak and by the unexpected rapid rise

-Nomenclature-
b = enclosure width x ,y = cartesian coordinates 8 = dimensionless temperature
Bi = Biot number bh/h X,Y = dimensionless cartesian coordinates
E = aspect ratio of the half cavity £/b a = thermal diffusivity Superscripts
Gr = Grashof number (gb3/v2) (p4 — pi)fp j8i. . . $4 = constants (see equation (7) * = pure conduction case
h = constant heat transfer coefficient v = kinematic viscosity — = mean temperature condition
k = thermal conductivity p = water density
£ = enclosure height Ap = (p - p)lp Subscripts
Nu = Nusselt number AT = Ti - T„ i = initial condition
Pr = Prandtl number via T = dimensionless time °° = ambient air temperature
q = local heat flux by unit area <I>T,<I>L,<I>B — dimensionless average heat w = wall condition
T = water temperature transfer at the top, side and bottom wall b = mixed mean water condition
t = time \p = stream function c = the center of the cavity
u,u = velocities in x a n d y directions ^ = dimensionless stream function max = maximum value
U,V, = dimensionless velocities in X and Y o) = vorticity min = minimum value
directions 12 = dimensionless vorticity 4 = condition at 4°C.

Journal of Heat Transfer AUGUST 1981, VOL. 103 / 529

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 of the average heat transfer. The mesh size used in this study ranged,
depending on the aspect ratio E of the cavity, from 1/10 to 1/15 in the
Y direction and E/20 or E/30 in the X direction. Numerical results
have shown that the use of a finer grid mesh made relatively little
difference to the final results although increasing sensibly the com-
puter time. In particular the vortex formation was not affected by a
mesh refinement. It was also verified that the vortex formation was
not influenced by the sweep direction in the S.O.R. solution as it does
for instance in a bottom-heated cavity. Typical values of the dimen-
sionless time step were 0.0001,0.0005 and 0.001. The total time steps
ranged from 1600 to 2500, and the corresponding computing time was
from 330 to 480 s on the IBM 360/70 computer.
To expedite plotting of the results, an auxiliary computer program
was developed to locate points lying on specified isotherms and
streamlines by linear interpolation of the computed values at the grid
points. All graphs were performed using a CALCOMP 563 automatic
plotter.
In order to check the validity of the present numerical method,
comparisons have been made with other existing solutions. An ex- KT* KT* T KT1 10"
cellent agreement was observed, for low Rayleigh numbers, with the Fig. 2 Transient distributions for d and <j>, with results from pure conduction
analytical Poot's solution [28]. For higher Rayleigh numbers, the and perfect mixing for T, = 8°C, Ta = 0°C, Bi = 2, E = 2, gb3/a2 = 5.42 X
numerical results from Wilkes [29], for the case of heat transfer into 10 8 andPr = 11.6
a rectangular cavity, were integrally reproduced. Another means of
checking that the present method was numerically conservative with effect on the temperature field. Therefore the top and bottom corners
sufficient accuracy was to evaluate the sum of the wall heat fluxes are approximately at the same temperature, this latter being naturally
from equation (11). Simpson's rule was used for numerical integration a minimum. However, as time progresses, the convective motion
and a three point finite difference approximation for dd/dr. The progressively stratifies the core region, as shown in Fig. 3(6). It is noted
agreement between the results obtained by the two alternative rela- that the resulting isotherm configuration, closely spaced near the top
tions given in equation (11) was found to be within 1 to 2 percent. wall but sparely spaced near the bottom one, produces respectively
large and poor heat transfer on those two boundaries.
The inversion process starts with the appearance of a 4°C isotherm
4 Results and Discussion in the temperature field, as shown in Fig. 3(6). As a result, the density
In view of the number of parameters involved in the present of water decreases in the layer of fluid near the bottom of the cavity,
problem, namely Bi, Pr, Ti,T„,E and gb3/a2, numerical solutions leading to a situation where the flow field is potentially unstable be-
are presented only for typical cases. Previous investigations on cause of the top heavy situation. This particular density field, related
transient natural convection in enclosed rectangular cavities with to the occurrence of the 4°C isotherm, gives rise to an additional pair
maximum density effects are concerned with walls having uniform of vortices rotating opposite to the existing ones. The resulting flow
temperature [12,16]. However it is of interest to extend these studies pattern is depicted in Fig. 3(c). With the progression of the inversion
to the case of a convective boundary condition where thermal insu- process, the lower vortex gradually increases its strength and displaces
lation effect, for example, can be considered. This type of boundary progressively the original vortex, as shown on sketches {c,d) and (e)
condition gives rise to nonuniform wall temperature. of Fig. 3. The combined vortex action has two main effects. Firstly,
4.1 The Inversion Effect. Consideration will be given first to it perturbs the isotherm pattern by carrying the cooler water from the
the case of a square cavity (E = 2) with initial water temperature T; side wall directly into the core region. Secondly, it improves the bot-
= 8°C, ambient temperature T„ = 0°C,gbVa2 = 5.42 X 108, (i.e. Gr tom heat transfer 4>B by transporting the warmer core water in the
= 6.72 X 104) and Bi = 2. The Prandtl number Pr is taken to be 11.6, lower portion of the cavity. Thus when T > 0.3, 4>B becomes even larger
corresponding to the average temperature 4°C. The calculated mean than 4>T- The inversion process comes to an end when the original
temperature 8, average wall temperature ffw and the heat transfer 4>x, clockwise vortex has been completely overtaken by the vortex origi-
<I>B and 4>L related to the top, lower and lateral boundaries respectively nating from below. In Fig. 3(/), i.e. at r = 0.375, the original clockwise
are presented in Fig. 2. The results for the two limit cases of pure circulation is completely reversed. The new motion brings gradually
conduction and perfect mixing are also plotted on this graph for the warm core in the lower region of the cavity. Consequently the
comparison. stratification of water is now characterized by weak gradients at the
The contours for time dependent stream function and isotherms, top and strong gradients at the bottom. Results obtained for higher
corresponding to different times r, are represented in Fig. 3, for flow values of T (not presented here) show that the momentum of the eddy
visualization studies. The maximum and minimum values of the motion is slowly dissipated by the opposing viscous forces and that
stream function ST/ and the contour intervals ASP for each graph are the fluid motion vanishes asymptotically with time.
given for reference purpose. The 4°C isotherm, when it exists, is shown The above description of the cooling process contains many aspects
as a heavy dashed line on the isotherm and streamline plots. The that are common to the other cases discussed in the present article.
symbols A, B, C, etc., shown in Fig. 2, correspond to the sequence of Thus the next sections will be limited to the description of the par-
the transient streamline pattern and isotherms in Fig. 3. ticular features characterizing those cases.
At the very beginning, the cooling process is characterized by a pure 4.2 The Biot Number effect. In this section, the effect of a
conduction heat transfer. The departure from the initial uniform relatively large Biot number (Bi = 20.) on the heat transfer is pre-
temperature T; occurs near the walls, giving rise to an important sented. Except for the Biot number, the other governing parameters
temperature gradient in that region, A resulting motion of the remain the same as in the preceeding section. The computed tem-
boundary layer type is thus set up near the vertical walls. This motion perature and heat transfer curves are shown in Fig. 4. In order to
is directed downward since it is associated with a horizontal density understand the effect of the Biot number combined to the maximum
gradient having the heavier fluid near the wall. As a consequence of density effect, it is worth to consider first the simpler problem of pure
the symmetry and the continuity, the resulting fluid motion inside conduction. In this case, an increase of the Biot number enhances the
the whole cavity consists of two counterrotating vortices. Figure 3(a) heat transfer (/>* as it can be seen from Figs. 2 and 4 where the tem-
only shows the right clockwise vortex. The isotherms represented on perature curves 6C*,6* and 6W* decrease faster toward zero with time.
this figure indicate that the convection has not yet exerted a sensible When convection is allowed, in the absence of a maximum density

530 / VOL. 103, AUGUST 1981 Transactions of the ASME

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 Fig. 3 Transient streamline pattern and isotherms for T, = 8°C, T„ = 0°C, Bi = 2, E = 2, gb3la2 = 5.42 X 10° and Pr = 11.6 at various time T . The centerline
and walls have value ¥ = 0. The values of T, * m a „ ¥ m l n and A * are (a)r = 0.025, * m a x = 9.9, AV = 0.99; (b) r = 0.075, t m , = 9.5, A ^ = 0.95; (c)
T = 0.175, * m a x = 4.6, ¥ m l n = - 0 . 2 , A * = 0.48; (d) T = 0.225, ¥ m a x = 3.1, ¥ m l n = - 0 . 7 , A * = 0.39; (e) T = 0.275, ¥ m a x = 1.2, * m , n = - 2 . 1 , A * =
0.33; ( 0 T = 0.375, ¥ „ „ „ = - 5 . 5 , A * = 0.55.

effect, the fluid motion taking place inside the cavity enhances </>T and
attenuates <J>B- An increase of the Biot number promotes the con-
vection, thus amplifying its effect on <j>r and <j>B. The presence of a
maximum density introduces an additional complexity. In fact an
increase of the Biot number has two opposing effects: one is the initial
circulation inside the cavity which is strengthened and the other is
the inversion process which occurs earlier and acts against the initial
circulation. There are situations where the inversion process takes
place so quickly that it annihilates almost completely the initial cir-
culation.
It is noted that a Biot number equal to infinity corresponds to the
case of non-insulated walls. The situation is equivalent to abruptly
lowering T to T „ at the initiation of the cooling process. This type of
situation has already been discussed in reference [30].
4.3 The Aspect Ratio Effect. The influence of the aspect ratio
E on the present problem has been studied for elongated cavities with
E = 1 and 4, respectively. In both cases the other parameters of the
problem were kept equal to those of case 4.1
The results obtained (see reference [31]) with both cavities have
shown that the general features of the density inversion remain Fig. 4 Transient distribution for 0 and <j>, with results from pure conduction
qualitatively similar to those of the square cavity. However, for the and perfect mixing, for T, = 8°C, r „ = 0°C, Bi = 20, E = 2, gb3/a2 = 5.42
X 10 8 andPr = 11.6
vertically elongated cavity (E = 4), it was observed from numerical
results that, when E is increased, the inversion process requires a E > 1 because of the inherent relative importance of the unstable layer
longer period of time to be a accomplished. The growing and spreading associated with this type of configuration [9]. A vertically elongated
of the counter clockwise vortex in the entire cavity was found to be cavity is more stable since the instability zones related to the vertical
a slower process than in the case of a square cavity. The results ob- density gradient along the horizontal walls are less important than
tained with the horizontally elongated cavity (E — 1) have shown that the driving forces generated by the horizontal density gradient near
the convective motion inside the cavity was considerably reduced as the vertical wall.
compared to the case with E = 4. This is due to the fact that in both 4.4 The Physical Parameter Effect. The effect of an increase
cases the parameter value g b 3 / a 2 was kept constant and that a de- of the physical parameter gba/a2 on the present problem may be
crease of E implies a reduction of the overall area of the cavity and observed by comparing the sequence of Fig. 5 to that of Fig. 3. It is seen
thus a proportional decrease of the convection. It has to be noted that from Fig. 5(a) that the flow starts with the usual pair of counterro-
a cavity with E < 2 is more difficult to treat numerically than one with tating vortices. However at time T = 0.035, it is observed from Fig. 5(6)

Journal of Heat Transfer AUGUST 1981, VOL. 103 / 531

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 Fig. 5 Transient streamline pattern and isotherms for 7} = 8°C, T„ = 0°C, Bi = 2, E = 2, gb3/a2 = 4.3 X 109 and Pr = 11.6 at various time T. The centerline
and walls have value ^ = 0. The values of T, ^ m a x , * » i „ and A ^ are (a) T = 0.025, ^ m a x = 30.1, At = 3.01; (6) T = 0.035, ^ m a , = 22.9, ^ m i „ = - 0 . 7 ,
A'S' = 2.36; (c) T = 0.150, * „ , a » = 8.4, ^ m , „ = - 4 . 4 , A t y = 1.28; (d) T = 0.250, * „ , „ = - 1 1 . 2 , A ¥ = 1.12

that an additional pair of secondary vortices symmetrically located similar to that studied by Seki and it is believed that a two-dimen-
on the axis of symmetry occurs near the top boundary. The appear- sional approach describes adequately the flow. However, the second
ance of these secondary vortices is directly related to the unstable mode of convection is characterized by the presence of additional rolls
situation resulting from the particular temperature field prevailing near the top boundary, those vortices resulting from the vertical
near the top wall. As the cooling progresses, the secondary vortex pair density gradient prevailing in this region. This situation is similar to
increases its strength. At time T = 0.15, the inversion process starts the case of a cavity heated from below for which a rich variety of flow
with the occurence of the lower corner vortex and develops as already structures have been observed in the past, as discussed for instance
described in Section 4.1. At time r = 0.25 the lower vortex has com- by Lindhorst, et al. [32]. It was found that, depending on the aspect
pletely overtaken the two original vortices and the circulation inside ratio of the cavity and the Rayleigh number, the resulting flow pattern
the cavity is completely reversed. Results obtained at higher values could be stationary or nonstationary. For stationary flow, three di-
of T show that themotion vanishes asymptotically with time. mensional torus like roll or rolls parallel to the shorter side of the
The existence of the additional vortex pair of Fig. 5(6) must be cavity [33] could be observed. Although similar, the present situation
dissociated from the inversion process since it appears before the is complicated by the fact that the secondary rolls result from the
occurrence of the 4°C isotherm in the temperature field. Moreover interaction between an unstable layer of fluid and an already estab-
the existence of this additional vortex is not a result of the nonlinearity lished flow field generated by the vertical walls. It is thus believed that
effects since it has also been observed in the present investigation for in the physically realized flow three-dimensional motion is likely to
situations in which the relationship between the water density and occur when those rolls are present. For such a situation, the present
its temperature was linear. For this case it was found that for relatively numerical analysis will not predict adequately the flow field in this
low Rayleigh numbers, i.e. for a low intensity of the convective motion, region.
the flow inside the cavity developed into a standard counterrotating 4.5 The Case with Supercooling. The sequence of Fig. 6 il-
vortex pair. However, when the Rayleigh number was sufficiently lustrates a case in which the convective motion is relatively important
high, the previous flow pattern was modified by the superposition of prior (Gr = 1.17 X 106) and after the inversion process. Shortly after
the two secondary vortices. Therefore, it is not suprising that for the the initiation of the cooling process a secondary motion occurs, as seen
present problem the second mode of convection does occur prior to on Fig. 6(a), in a way comparable to the case described in Section 4.4.
the inversion process since the convective motion has been enhanced At time T = 0.05, the inversion process gives rise to the usual inversion
by the increase of the physical parameter, g6 3 /a 2 , as compared with vortex near the lower corner of the cavity (Fig. 6(6)). As the cooling
the situation involved in Section 4.1. progresses, the relatively intense convective motion interacts with
The effect of density inversion on steady free convection of cold the unstable layer of fluid located between the 4°C isotherm and the
water contained in a rectangular cavity with isothermal vertical walls lower boundary. As a result a new pair of secondary vortices appears
at different temperatures has been studied numerically and experi- in the lower central portion of the cavity (see Fig. 6(c)). This addi-
mentally by Seki, et al. [17]. In that analysis a two-dimensional ap- tional vortex pair was not present in the case described in Section 4.4
proach was assumed and verified experimentally for cavities with due to the fact that the convective motion was not sufficiently strong
aspect ratio E = 2 to 10. during the inversion process. It is interesting to note that this new
When the first mode of convection is involved in the present vortex is connected to the original main one and that both rotate in
problem, the mechanism giving rise to the inversion process is quite the same clockwise direction. However, at time T = 0.10, Fig. 6(d)

532 / VOL. 103, AUGUST 1981 Transactions of the ASME

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 Fig. 6 Transient streamline pattern and isotherms for T, = 10°C, T"„ = - 1 0 ° C , Bi = 2, E = 2, gb3/a2 = 4.3 X 109 and Pr = 13.7 at various time T . The
centerline and walls have value ^ = 0. The values of T, * „ . , *min and A ^ are (a)r = 0.025, * m a x = 32.1, • „ , , „ = - 1 2 . 0 , A ¥ - 4.41; (b) T = 0.050,
^ m a x = 22.1, * m l n = -12.3, A* = 3.44; ( c ) T = 0.085, * m a x = 14.2, ^ m i n = -15.9 A * = 3.11; (d) T = 0.100, * m a l I = 5.1 ,.*„,,„ = - 2 2 . 1 , A * = 2.72;
(e) T = 0.125, ^ m a x = 7.6, ^ m | n = -18.9, A * = 2.65; (/) T = 1.300, ^ m a x = 1.5, ¥ m i n = - 6 . 4 A * = 0.79

shows that this connection is broken by the growth of the inversion investigation is not possible. Nevertheless, a few qualitative com-
vortex. Thus the upper secondary vortex connects now with the in- parisons can be made in particular with Gilpin's results. For instance
version vortex, both rotating counter-clockwise. A very small eddy the transient behavior of the inversion process, as described in Section
in the upper right corner of the cavity may also be observed on the 4.1 of the present paper, is quite similar to that observed experi-
same figure. With progression of cooling the original clockwise vortex mentally by Gilpin [16] in a pipe. In particular detailed observations
gradually fades away due to the continuous growing of the inversion of the flow pattern during the transition show that for relatively low
vortex. Thus at time T = 0.125 only two vortices are present inside the intensity of convection, the inversion begins with a reversal of the flow
cavity. The flow pattern of Fig. 6(e) is now opposite to that of Fig. 6(a) direction in the boundary layer near the bottom of the cylinder. This
and the inversion process is over. The remaining secondary vortex in reversal spreads to the top of the cylinder and then begins the reverse
the bottom of the cavity is similar in character to the initial secondary flow in the core. It is interesting to note that the secondary convection
vortex. In fact its presence is maintained by the combined action of rolls predicted by the present theory have not been observed nu-
the existing unstable layer near the bottom boundary and the intense merically or experimentally in the case of a horizontal pipe. The for-
convective motion which follows the inversion. Fig. 6(/) indicates that mation of a secondary vortex over a horizontal circular ice cylinder
the time required for the secondary motion to disappear is very long. located in quiescent water of constant temperature has however been
Thus at time r = 1.3, the secondary vortex is still present, although reported by Saitoh, et al. [35].
greatly reduced in strength, as indicated by the value of ^min- In fact Experiments concerning the supercooling of water inside a circular
a value T » 2 is required for this secondary vortex to disappear com- pipe have been performed by Gilpin [3]. Results were obtained for
pletely. Biot number of 0.2 and 15, respectively. It was observed that in both
4.6 Comparison with Existing Experimental Studies. The cases, following the inversion process, dentritic ice started to form at
cooling of a mass of water through its maximum density at 4°C has the top of the pipe, due to the presence of cold water at a temperature
been studied experimentally in the past for different flow geometries of approximately —4°C in this region. However, the overall growth
and boundary conditions. For instance, Seki, et al. have investigated of ice inside the pipe was considerably affected by the Biot number.
the effect of density inversion on the steady natural convection of a For the higher Biot number the rapid cooling rate produced large
mass of water contained between two horizontal concentric cylinders temperature differences in the water and dentritic ice only grew ad-
maintained at different constant temperatures [34] and in a confined jacent to the pipe walls. For the lower Biot number, the cooling process
rectangular vessel in which different temperatures were imposed on was more uniform inside the cavity, and, as a result, dentritic ice
the opposing vertical walls [17]. Furthermore the transient natural blocked the entire section of the pipe. Numerical results obtained in
convection heat transfer of a horizontal cylinder of water cooled the present study, but not presented here, concerning the supercooling
through its maximum density has been considered by Gilpin for two of a rectangular cavity at various Biot numbers were qualitatively in
types of thermal boundary conditions, namely a convective cooling agreement with Gilpin's results. However, the present case is char-
[3] and a constant cooling rate [35]. Due to the differences in geome- acterized by the fact that the corners of the cavity cool down more
tries or boundary conditions a direct quantitative comparison between rapidly than the rest of the boundary. It is thus expected that ice
the results of those experimental studies and the present numerical formation will generally start at corners.

Journal of Heat Transfer AUGUST 1981, VOL. 103 / 533

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 5 Conclusions 11 Sugawara, M., Fukusako, S., and Seki, N., "Experimental Studies on
the Melting of a Horizontal Ice Layer," Bulletin of the Japanese Society of
A n u m e r i c a l s o l u t i o n is o b t a i n e d for t h e t r a n s i e n t n a t u r a l convec-
Mechanical Engineering, Vol. 18, No. 121,1975, pp. 714-721.
t i o n of w a t e r , enclosed in r e c t a n g u l a r cavities w i t h a convective 12 Cheng, K. C , and Takeuchi, M., "Transient Natural Convection of Water
b o u n d a r y condition, a n d cooled t h r o u g h t h e m a x i m u m d e n s i t y p o i n t in a Horizontal Pipe with Constant Cooling Rate through 4°C," ASME JOUR-
a t 4 ° C . T h e n u m e r i c a l solution, b a s e d on a n implicit finite difference NAL OF HEAT TRANSFER, Vol. 98,1976, pp. 581-587.
13 Cheng, K. C , Takeuchi, M., and Gilpin, R. R., "Transient Natural
m e t h o d , yields d e t a i l e d s t r e a m l i n e p a t t e r n s a n d i s o t h e r m s r e a d i l y
Convection in Horizontal Water Pipes with Maximum Density Effect and
for flow visualization. T h e results o b t a i n e d in t h e p r e s e n t s t u d y m a y Supercooling," Numerical Heat Transfer, Vol. 1,1978, pp. 101-115.
b e s u m m a r i z e d as follows: 14 Gilpin, R. R., "The Effects of Dentritic Ice Formation in Water Pipes,"
1 C o n v e c t i o n in t h e a b s e n c e of m a x i m u m d e n s i t y effect is e n - International Journal of Heat and Mass Transfer, Vol. 20, 1977, pp. 693-
699.
h a n c e d b y a n increase of t h e p a r a m e t e r s gb3/a2 a n d Bi. A n a d d i t i o n a l
15 Robillard, L. and Vasseur, P., "Effet du maximum de densite sur la
v o r t e x pair was o b s e r v e d t o a p p e a r n e a r t h e u p p e r b o u n d a r y for suf- convection libre dans une cavite fermee," Canadian Journal of Civil Engi-
ficiently high convective m o t i o n . neering, Vol. 6, No. 4,1979, pp. 481-493.
2 T h e effect of m a x i m u m d e n s i t y c o n s i s t s in slowing d o w n t h e 16 Gilpin, R. R., "Cooling of a Horizontal Cylinder of Water through its
Maximum Density Point at 4°C," International Journal of Heat and Mass
initial circulation inside t h e cavity a n d s u b s e q u e n t y in r e v e r s i n g it.
Transfer, Vol. 18,1975, pp. 1307-1315.
T h e r e s u l t i n g h e a t t r a n s f e r is t h u s r e d u c e d in c o m p a r i s o n t o a s t a n - 17 Seki, N., Fukusako, S. and Inaba, H., "Free Convective Heat Transfer
d a r d situation w i t h o u t m a x i m u m density effect. A consequence of t h e with Density Inversion in a Confined Rectangular Vessel," Warme-und Stof-
m a x i m u m d e n s i t y is t h e s w i t c h of t h e m a x i m u m h e a t t r a n s f e r from fabertragung, Vol. 11,1978, pp. 145-146.
t h e t o p wall t o t h e b o t t o m o n e d u r i n g t h e cooling p r o c e s s . 18 Dorsey, N. E., "The Freezing of Supercooled Water," Transactions of
the American Society of Philosophical Society, Vol. 38,1948, pp. 248-325.
3 T h e t e m p e r a t u r e in t h e w a t e r can d e p a r t significantly from 19 Newell, M. E. and Schmidt, J., "Heat Transfer by Laminar Natural
those t h a t would be predicted on t h e basis of simplifying a s s u m p t i o n s Convection within Rectangular Enclosures," ASME JOURNAL OF HEAT
s u c h as " p u r e c o n d u c t i o n " or " p e r f e c t m i x i n g " . T h i s r e s u l t i m p l i e s T R A N S F E R , Vol. 92,1970, pp. 159-168.
t h a t t h e p r e d i c t i o n of t h e c h a r a c t e r i s t i c s of t h e freezing of a m a s s of 20 Gray, D. D., and Giorgini, A., "The Validity of the Bousinesq Approxi-
mation for Liquids and Gases," International Journal of Heat and Mass
water will be in error if either of those simplifying a s s u m p t i o n s is used. Transfer, Vol. 19,1976, pp. 545-551.
T h e s e errors will b e p a r t i c u l a r l y i m p o r t a n t a t large B i o t n u m b e r s . 21 Kreith, F., Principles of Heat Transfer, Intext Educational Publishers,
4 T h e r e s u l t s o b t a i n e d in t h i s s t u d y for s u p e r c o o l i n g c o n d i t i o n s New York, 1973.
22 Fujii, T., "Fundamentals of Free Convection Heat Transfer," Progresses
involve a relatively h i g h c o n v e c t i v e m o t i o n w i t h s e c o n d a r y v o r t e x
in Heat Transfer Engineering, Vol. 3,1974, pp. 66-67.
pairs prior a n d after t h e inversion process, giving rise t o complex flow 23 Weast, R. C , Handbook of Chemistry and Physics, 52nd ed., Chemical
patterns. Rubber Co., Cleveland, Ohio 1972, p. F5.
24 Roache, P., Computational Fluid Dynamics, Hermosa Publishers,
Acknowledgment 1976.
25 Woods, L. C , "A Note on the Numerical Solution of Fourth Order Dif-
T h i s r e s e a r c h was s u p p o r t e d b y t h e N a t i o n a l R e s e a r c h Council of ferential Equations," Aeronautical Quarterly, Vol. 5, Part 3,1954, p. 176.
C a n a d a t h r o u g h g r a n t s N R C A-4197 a n d N R C A-9201, for w h i c h t h e 26 Briley, W. R., "A Numerical Study of Laminar Separation Bubbles Using
a u t h o r s a r e grateful. the Navier-Stokes Equations," Report J110614-1, United Aircraft Research
Laboratories, East Hartford, Conn, 1970.
27 Rubel, A. and Landis, F., "Numerical Study of Natural Convection in
References a Vertical Rectangular Enclosure," High Speed Computational Fluid Dy-
1 Farhadieh, R., and Tankin, R. S., "A Study of the Freezing of Sea Water," namics, Vol. II, 1969, pp. 208-213.
Journal of Fluid Mechanics, Vol. 71, part 2,1975, pp. 293-304. 28 Poots, G., "Heat Transfer by Laminar Free Convection in Enclosed Plane
2 Tankin, R. S., and Farhadieh, R., "Effects of Thermal Convection Gas Layer," Quarterly Journal of Mechanics and Applied Mathematics., Vol.
Currents on Formation of Ice," International Journal of Heat and Mass 11,1958, pp. 257-273.
Transfer, Vol. 14,1971, pp. 953-961. 29 Wilkes, J. D., "The Finite difference Computation of Natural Convection
3 Gilpin, R. R., "The Effect of Cooling Rate on the Formation of Dentritic in an Enclosed Rectangular Cavity," Ph.D. thesis, University of Michigan,
Ice in a Pipe with no Main Flow," ASME JOURNAL OF HEAT TRANSFER, Vol. 1963.
99", 1977, pp. 419-424. 30 Vasseur, P. and Robillard, L., "Transient Natural Convection Heat
4 Goren, S. L., "On Ice Convection in Water at 4°C," Chemical Engi- Transfer in a Mass of Water Cooled through 4°C," International Journal of
neering Science, Vol. 21,1966, pp. 515-518. Heat and Mass Transfer, Vol. 23, no. 9,1980, pp. 1195-1205.
5 Schechter, R. S., and Isbin, H. S., "Natural Convection Heat Transfer 31 Robillard, L. and Vasseur, P., "Transient Natural Convection Heat
in Regions of Maximum Density," American Institute of Chemical Engineering Transfer of Water with Maximum Density Effect and Supercooling," ASME/
Journal, Vol. 4, No. 1,1958, pp. 81-89. AICHE National Heat Transfer Conference, Orlando, Fla, July 27-30,1980,
6 Schechter, R. S., "Natural Convection Heat Transfer in Regions of Paper No. 80-HT-74.
Maximum Density," Ph.D. Thesis, University of Minnesota, 1956. 32 Linthorst, S. J. M., Schinkel, W. M. M. and Hoogendoorn, G. J., "Flow
7 Vanier, C. R., and Tien, C , "Effect of Maximum Density and Melting Structure with Natural Convection in Inclined Air-Filled Enclosures,"
on Natural Convection Heat Transfer from a Vertical Plate," Chemical Engi- A.I.Ch.E./ASME National Heat Transfer Conference, Orlando, Fla, July 27-30,
neering Progress Symposium Series, Vol. 64, No. 82,1968, pp. 240-254. HTD- Vol. 8,1980, pp. 39-46.
8 Vasseur, P., and Robillard, L., "Natural Convection of Water near the 33 Stork, K. and Muller, U., "Convection in Boxes: Experiments," Journal
Freezing Point," IARH Symposium on Ice Problems, Luela, Sweden, 1978, pp. Fluid Mechanics, Vol. 54, Part 4,1972, pp. 599-611.
257-270. 34 Seki, N., Fukusako, S. and Nakaoka, M., "Experimental Study on
9 Merker, G. P., Waas, P., and Grigul, U., "Onset of Convection in a Natural Convection Heat Transfer with Density Inversion of Water Between
Horizontal Water Layer with Maximum Density Effects," International Two Horizontal Concentric Cylinders," ASME JOURNAL OF HEAT TRANSFER,
Journal of Heat and Mass Transfer, Vol. 22,1979, pp. 505-515. Vol. 97,1975, pp. 556-561.
10 Cheng, K. C , and Wu, R. S., "Maximum Density Effects on Thermal 35 Saitoh, T. and Hirose, T., "Thermal Instability of Natural Convection
Instability of Horizontal Laminar Boundary Layers," Applied Sciences Re- flow over a Horizontal Ice Cylinder Encompassing a Maximum Density Point,"
search, Vol. 31,1976, pp. 465-479. ASME J O U R N A L O F H E A T T R A N S F E R , Vol. 102,1980, pp. 261-267.

534 / VOL. 103, AUGUST 1981 Transactions of the ASME

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 Flow Structure with Natural
S. J. M. linthorsi
Research Scientist.
Convection in Inclined Air-Filled
W. M. M. Schinkel 1
Research Scientist. Enclosures
C. J. Hoogendoom Flow visualization observations are described for natural convection flow in rectangular
Professor. inclined enclosures. Observations are made in air-filled enclosures of small and moderate
aspect ratio (0.25 < Ax < 7), angles of inclination from 0 to 90 deg and Rayleigh number
University of Technology, between 5 X 10a and 2.5 X 10s. For the range of parameters considered we determined the
Delfl, Holland transition from stationary to nonstationary flow and the transition from two-dimensional
to three-dimensional flow. Also, the different stationary flow structures are described.
In addition measurements of the velocity profile of the stationary flow have been per-
formed by means of a laser doppler anemometer. Special attention has been given to the
side wall boundary condition of the enclosure.

Introduction It is generally known that for small Ra numbers there is an uni-

For several years increasing attention has been paid to the problem cellular two-dimensional motion, commonly referred to as the con-
of natural convection in inclined enclosures. Recently an excellent ductive base flow. The heat transfer in this situation is mainly due
review was given by Catton [1]. Most investigations deal with the to conduction. With increasing Rayleigh number this base flow be-
dependence of heat transfer on the relevant parameters. Relatively comes unstable. It appears that two types of instabilities can be dis-
little attention has been given to the flow behavior, especially for low tinguished. For angles of inclination smaller than a certain transition
(Ax =S 1) and moderate (1 < Ax < 7) aspect ratios. angle <j>', the instability is mainly buoyancy driven and leads to lon-
However, knowledge of flow behavior is important for a number gitudinal rolls superposed on the base flow. Longitudinal rolls have
of reasons. It has been observed [2] that a transition from one flow their axes upslope and the rotation direction of a longitudinal roll is
pattern into another causes a discontinuity in the slope of the Nusselt opposite to the direction of rotation of its direct neighbors. For angles
number relations. Moreover, it is necessary for calculation purposes of inclination between (j> = <j>' and qi = 90 deg the instability is mainly
to know in what range of a parameter the flow is stationary and two- hydrodynamic and leads to transverse rolls (small and moderate
dimensional or three-dimensional. The transition to nonstationary Prandtl number) or travelling waves (high Prandtl number). A further
and to turbulent flow is important from the more fundamental point increase of the Ra number causes a transition to nonstationary
of view. flow.
The objective of this experimental investigation is the determina- At moderate and small aspect ratios there are only a few investi-
tion of natural convection flow characteristics, including secondary gations of the flow structure. Ozoe, et al. [9, 10, 11] investigated the
motion, the transition from two-dimensional to three-dimensional flow and heat transfer both theoretically and experimentally. The
flow and the transition from stationary to nonstationary flow in air- aspect ratio varied between 1-15.5 and Ra between 3 X 10 3 and 106.
filled enclosures of small and moderate aspect ratio. Two types of In most experiments large Prandtl number fluids have been used. For
experiments have been done. In order to investigate the flow quali- Ax > 8 air was the working fluid. For Ax = 1 Ozoe, et al. obtained a
tatively flow visualization experiments have been performed. Sec- unicellular two-dimensional motion for 10 deg < 0 ^ 90 deg. At 0 =
ondly velocity measurements with a laser doppler anemometer (Ida) 10 deg a transition to a complex three-dimensional stationary flow
have been carried out in order to obtain quantitative results. has been observed. Also the heat transfer results showed this transi-
In our experiments, the Rayleigh number (Ra), based on the plate tion: A minimum value of the Nusselt number occurred for 4> = 10 deg.
distance D, varied from 5 X 10 3 to 2.5 X 106 and the angle of inclina- At larger aspect ratio Ax a similar behavior has been observed.
tion 4> from 0 to 90 deg. The value of the height to width aspect ratio However the angle of transition (<//) increases with increasing aspect
Ax was between 0.25 and 7, the depth to width aspect ratio Az was, ratio.
dependent on the plate distance D, either 5 or 10. For notation see Fig.
1. Experimental Setup
Both the flow visualization experiments and the laser doppler an-
Flow Regimes emometer measurements are performed with the same model (see Fig.
Most studies of the flow behavior have been done in the horizontal 2). The model consisted of one copper plate and one aluminium plate
2
(4> = 0 deg, heated from below) orientation and the vertical orientation (0.20 X 0.30 m ) for the isothermal hot wall and isothermal cold wall,
(4> = 90 deg). For notation see Fig. 1 Krishnamurti did flow visual- respectively. The distance between the walls could be varied by in-
ization experiments for large aspect ratio and $ = 0 deg [3,4,5] Stork
and Muller [6] considered the aspect ratio dependence of the flow for
Az < 6 and Az < 6 for the horizontal orientation.
In the vertical orientation (tj> = 90 deg) flow visualization has been
done by Elder [7] for aspect ratios of 1 to 60, while recently Morrison
and Tran [8] performed velocity' measurements of the natural con-
vection flow for Ax = Az = 5 at Ra = 5 X 104.
For the inclined orientation some flow visualization experiments
are done by Ozoe [9].
1
Present address: Shell/Amsterdam.
Contributed by the Heat Transfer Division and presented at the 19th
AlChE/ASME National Heat Transfer Conference, July 27-30,1980, Orlando,
Pla. Revised manuscript received by the Heat Transfer Division October 17,
1980. Fig. 1 A schematic view of the enclosure

Journal of Heat Transfer AUGUST 1981, VOL. 103 / 535

Copyright © 1981 by ASME
Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 a b c d

=
Flg.3 Flow structure In the vertical orlentallon: (a) Ra 3.6 X 10·, Ax ::
=
7; (b) Ra = 2.0 X 105 , Ax = 7; (e) Ra 2.5 X 10 5 , Ax= =
7; (d) Ra 2.5 X
10 5 , Ax = 4

He explains that the secondary motion in the side wall regions is

caused by the stratification of the flow in this region. The secondary
Fig. 2 The experimental model
motion in the center is buoyancy driven as explained by Mallinson
serting two different covers, one for a plate distance of 0.02 m and one and de Vahl Davis [14J. The tertiary motion is a result of the inter-
for D = 0.04 m. So the depth to width aspect ratio, A., was either 10 action between the secondary motion in the side wall region and in
or 5. All the sidewalls were made of perspex to make flow visualization the center.
and Ida measurements possible. The temperature of the isothermal At Ra = 2.5 X 105 the aspect ratio Ax has been decreased. It ap-
walls could be controlled by means of two thermostatbaths. The peared that the secondary and tertiary motions became more pro-
temperature difference of the isothermal walls was varied up to 50 K. nounced (Fig. 3(d)) with decreasing aspect ratio up to Ax = 2. For
Both walls were isothermal to within 2 percent of the temperature Ax < 2 the tertiary motion disappeared and for Ax = 1 only two sec-
difference at!::..T = 40 K and within 8 percent for!::..T = 5 K. To vary ondary rolls remain. For Ax < 1 no secondary motion was ob-
Ax, perspex walls could be placed at different heights in the enclosure. served.
For Ax < 1 a slit structure, made of Lexan, was inserted in the cavity. For Ax <;; 7 the flow remained stationary for all Ra numbers con-
At one of the shorter side walls small inlets were placed to inject smoke sidered. It appeared that for Ax <;; 7 transverse rolls, as predicted and
to visualize the flow pattern. observed for large aspect ratios (Korpela [15]) never occured. This
To illuminate ayz or xy-plane of the enclosure we created a parallel fact is in agreement with the stability theory of Bergholz [16J. In his
lightbeam by means of an optical setup and a Hg source of 500 W analysis he took the effect of stratification in the core region into ac-
power. The model was placed on an apparatus to vary the angle of count and found that the stratification causes an increase of the
inclination with increments of 10 deg. The exposure time for making critical Rayleigh number for the onset of transverse rolls. Our results
photographs varied between l/S to 2 s for a film of 400 ASA (27 DIN). confirmed by numerical results of Schinkel [13], show that the region
The camera we used was an Asahi Pentex ES II with a SMC Takumar of transverse rolls is never entered.
f 1.8/55 mm lens. Close-up and wide-angle lenses were added. Horizontal Orientation. Our experiments in the horizontal or-
The Ida was used in the reference beam mode and only velocity ientation (</> = 0 deg) are generally in good agreement with other in-
profiles of stationary flows could be measured. We used a 5 mW vestigations.
He-Ne laser. A grating divides the laser beam into an intense oth order For Ax = 0.5 and Ax = 0.25 the flow appeared to be stationary for
scattering beam and a weak first order reference beam. Instead of all Rayleigh numbers considered. For Ax ~ 1 the flow becomes non-
smoke, we used particles of aluminium oxide with a mean diameter stationary when the Rayleigh number exceeds a critical value. The
of 1 /-Lm for the Ida measurements. The light, scattered from a particle dependence of the critical Rayleigh number on Ax has been given in
passing the measuring volume, interferes with the first order beam Fig. 4. With increasing Ax the critical Rayleigh number decreases,
on the photodiode. After passing a frequency filter, an amplifier and which is expected according to the theory of Clever and Busse [17].
a Schmitt trigger, eight periods of the signal of the photodiode are For Ax <;; 1 we observed longitudinal rolls with the rotation axis
counted which gives the value of the velocity component of the flow. parallel to the x -axis. For A. = 5, four rolls were observed at Ax = 1
Details of the Ida experimental set-up can be found in Schinkel and Ra = 5 X 104 (Fig 5(a)); for Ax = 0.5 we observed six rolls. For Ax
[12]. = 0.25 situations of both six and eight rolls were obtained (see Fig. 5( b)
and (c )). The fact that the number of rolls increases with decreasin/1
Flow Visualization Results aspect ratio Ax is in agreement with Stork [6]. However, the number
Vertical Orientation. For Ax = 7 and A. = 5 unicellular two- of rolls is not in agreement with his results. Also the flow behavior for
dimensional motion was still observed at Ra = 3.6 X 104 (see Fig. 3(a)). Ax > 1 that we observed in our experiments is not according his re-
At this Rayleigh number the flow is in the so-called transition regime. sults. We obtained for Ax > 1 a torus like roll with fluid moving down
With increasing Rayleigh number a gradual transition to the boundary at the four side walls and moving up in the center of the enclosure.
layer regime has been observed. The first secondary motion appeared This flow pattern has also been obtained by Krishnamurti [3,4,5] who
at Ra = 1.5 X 105 (see Fig. 3(b) for clarity a photo is shown for Ra = mentioned that curved rolls are the rule rather than the exception.
2 X 10 5) near the side wall regions. Secondary motion has the same This particular flow behavior can be explained by side wall effects.
rotation direction as the base flow. Further increase of Ra leads to If the mean temperature T m (= (Th + T c )/2) of the fluid is not equal
additional secondary motion in the center of the enclosure. At Ra = to the ambient temperature a net heat transfer will occur through the
2.5 X 10 5 a weak tertiary motion located between the secondary mo- side walls to the surroundings. In our case the mean temperature T m
tion, is observed (see Fig. 3(c)). The tertiary motion rotates in a di- is greater than the ambient temperature T a and the net heat transfer
rection opposite to the base flow. The secondary and tertiary motions is positive. This effect forces the fluid to move downwards at the side
have also been obtained in numerical calculations of Schinkel [12, 13]. walls.

536 / VOL. 103, AUGUST 1981 Transactions of the ASME

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 Ra

(al
non- stationary flow

stationary flow

o 2 4 6 8 Ax
Fig. 4 The Rae for transition from stationary to nonstationary flow for the
horizontal orientation (cP = 0 deg) dependent on Ax

For Ax ~ 1 this effect results always in an even number oflongitu-

dinal rolls while for Ax > 1 the torus shaped roll appears.
For Ax = 1 and A z = 10 we observed as expected ten equal sized TC_·~""'''''''''~~'''-=-1fl1I'!'
yl •
Th-~"'''''IIIIIiII:l_''
longitudinal rolls (Fig. 5(d».
The Inclined Situation. If we decrease the angle of inclination -r-
(d)
from <p = 90 deg, we find a similar change in the flow pattern for aspect
ratios 1 ~ Ax ~ 7. Flg.5 Longitudinal rolls in the horizonlal orientation (<I> = 0 deg): (a) Ax =
With decreasing angle, the tertiary and secondary motion tends to
1, Ax =5; (b) A. = 0.25, Ax = 5; (e) Ax =0.25, Ax = 5; (d) Ax = 1, Ax =
10
disappear before the flow becomes nonstationary or three-dimen-
sional.
An example of the change in the flow pattern with <p can be seen in
Fig. 6 for Ax = 1. By decreasing <p and Ra, we obtained the transition
curves for each aspect ratio for stationary to nonstationary flow and
for two-dimensional to three-dimensional flow. In Fig. 7 the transition
curves are given for each aspect ratio. With decreasing aspect ratio
the flow remains stationary for higher Ra number and smaller <p. Even
for Ax = 0.5 and 0.25 only stationary flow was observed for the range
of parameters considered. The transition from stationary two-di-
mensional to stationary three-dimensional flow occurs with increasing
Ax for increasing <p for Ax ~ 1. This is in agreement with experiments (a) cP =90 deg, Ra = 2.10 5 (b) cP = 70 deg, Ra = 2.10 5
of Ozoe [10]. For Ax < 1 the transition occurs for increasing <p with
decreasing Ax (for example <p' = 40 deg for Ax = 0.5 and <p' = 60 deg
for Ax = 0.25 at Ra = 2 X 105). The angle oftransition for transition
to three-dimensional flow is only slightly dependent on Ra.

/;y,;-:-=~
. . 'O~"'1'"
For a plate distance of 0.02 m no significant change for the two-
dimensional flow has been observed. This indicates that for D = 0.04
m the depth aspect ratio (A z = 5) is yet great enough to insure two-
dimensionality of the base flow at the midplane of the enclosure. But, :". . .'
as already explained for the horizontal orientation, it effects the
number of longitudinal rolls. This number is also influenced by the . ''''~~. /.
heat transport through the sidewalls as explained before.
At Ra = 2 X 10 5 , <p = 60 deg and Ax = 1 we observed a particular
flow pattern, i.e., two different orientations of the flow in different
cross sections (see Fig. 6): one more or less horizontal and one vertical. (e) </J = 60 deg, Ra = 2.10 5 (d) cP = 60 deg, Ra = 2.10 5
In the z -direction there are 14 recirculating cells, seven for each or- Fig. 6 Observed flow for Ax = 1, Ax = 5, for different angle ollncllnallon
ientation. For a smaller value of Ra, only the unicellular base flow
remains. develops from a creeping flow at low Ra number to the boundary layer
regime. With increasing Rayleigh number, the maximum velocity
Results of Laser Doppler Measurements increases and the flow is concentrated near the hot and cold walls
Velocity profiles wer.e measured for stationary flows. In Fig. 8 the while the center of the enclosure is relatively tranquil. This devel-
U-\'elc,,,itv profile
(velocity in the x -direction) dependence on y is given opment is in agreement with the theory described by Batchelor
the midplane of the enclosure (x = 0.5 Ax, z = 0.5 A z ) for Ax = 7 [18].
different Rayleigh numbers. One can clearly see that the flow For Ra = 2.2 X 105 u-velocity profiles are measured for different

JOIJrnal of Heat Transfer AUGUST 1981, VOL. 103 / 537

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 i i i 1
1/ ' I | ' 1
2 1 / | -
nor -stationary
Ra '/ /
Ax=1/

103 — 1 2/ 1
4/ 7 _

I
I
5- I
I
-
' 1 /
I
I u(m/s)
' /

104
2

-
-
1
!'.
*
/
-

/ stationary
i i —^r i
5 -
<
^T .5 .25 1
'
4
1
1
7
i „ i i 1 i ! , 1
20 40 60 0 80
Fig. 7 Transition curves lor different Ax dependent on Ra and 4>~- — sta-
tionary to nonstationary flow; - - - two-dimensional to three-dimensional
flow
Fig. 8 u—velocity profiles for different Ra numbers for Ax = 7,<p = 90 deg:
( a ) Ra = 3.6.10"; (b) Ra = 6.8.10"; (c) Ra = 1.3.10s; (d)Ra = 1.8.105; (e)
Ra = 2.2.105; (/) Ra = 2.S.105
aspect ratio Ax = 1, 2, 3,4 and 5 (see Fig. 9). With increasing Ax the
maximum velocity and the thickness of the boundary layer increases.
In Fig. 10, the measured velocity profiles are given for Ax = 1 for
different angles of inclination. It appears that with decreasing angle
of inclination the maximum velocity remains approximately the same
when $ > 60 deg. Moreover, the velocities in the core region increase
the decreasing angle of inclination.
To determine the influence of the heat transfer through the side
walls on the flow, we measured at different depths in the enclosure
the velocity profile (Fig. 11). The temperature Ta was smaller than
the mean temperature Tm during the measurements. So we expected
a downward directed flow near the side walls. In Fig. 11 this downward
directed flow can be seen. Also we notice that for z/L = 0.3 to 0.5 the
velocity profiles are equal which indicates that here the influence of
the side walls is negligible. However, the velocity profiles at the cold
and the hot wall are not fully symmetrical. Because of the net down-
ward directed flow at the side walls the upward flow in the center of
the enclosure should compensate this effect. Therefore, the net up-
ward flow is positive. Measurements with insulated side walls (0.03
m polystyre) give a much more symmetrical velocity profile (see Fig.
12), which is in good agreement with measurements of Morrison and
Tran [8]. This side wall boundary effect is less for smaller aspect ratio
Ax, indicating that the heat transfer effect decreases with decreasing
aspect ratio.
Fig. 9 u—velocity profiles for different Ax, (j> = 90 deg: 9 Ax = 1; O Ax =
Conclusions 2;TJ»=3;DJ, = 4;AA,=5
Flow visualization observations for the vertical orientation for small
Ro,l310S
and moderate aspect ratio 0.25 =S Ax ^ 7 and 5 X 103 « Ra « 2.5 X 10B
Ax ' u (mi's)
did not show transverse rolls. Instead for high Rayleigh number (Ra x z = .5
> 1.4 X 105) and Ax>\ secondary and tertiary motion occurred. .04
In the horizontal situation we observed for Ax < 1 longitudinal rolls. - If .- 90°
70°
The number of rolls is influenced by the side wall boundary condition. ^ , ' '// \
For Ax > 1 we obtained a torus-like flow pattern also due to the side 40°
wall boundary condition of the enclosure. - . 30°
/ - ' .••"""' y ' 7
/• s' */ \i-
For all angles of inclination we determined the transition curves
for the transition of stationary flow to nonstationary flow and for .,-•'' •1
1
transition of two-dimensional flow to three-dimensional flow. We ^Z '
s
/ \
observed only stationary flow for Ax < 1 at the parameter range ....•'*;-•- I 1
5
considered. For Ax > 1 the onset of nonstationary flow occured for % 4 .3 .2 1 0
the three-dimensional flow at a smaller value of the Rayleigh number Fig. 10 u—velocity profiles for different <j>, Ax = 1

538 / VOL 103, AUGUST 1981 Transactions of the ASME

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 u(m/s)

.3 .2 .1
y/D
Fig. 12 u—velocity profiles with different side wall boundary condition: ®
noninsolated; O insolated

6 Stork, K., Muller, U., "Convection in Boxes: Experiments," Journal of

Fig. 11 u—velocity profiles at different depths in the enclosures: 9 z/L = Fluid Mechanics, Vol. 54, Part 4,1972, pp. 599-611.
0.5; M z/L = 0.4; • z/L = 0.3; V z/L = 0.2; A z/L = 0.1, O z/L = 0.05 7 Elder, J. W., "Laminar Free Convection in a Vertical Slot," Journal of
Fluid Mechanics, Vol. 23, Part 1,1965, pp. 77-98.
8 Morrison, G. I., Tran, V. Q., "Laminar Flow in Vertical Free Convective
than for the two-dimensional flow. The critical Rayleigh number for Cavities," International Journal of Heat and Mass Transfer, Vol. 21, 1978,
onset to nonstationary flow increased with decreasing Ax. pp. 203-213.
From the Ida measurements it can be concluded that the value of 9 Ozoe, H., Sayama, H., Churchill, S., "Natural Convection in a Long In-
the depth to width aspect ratio (Az = 5) is great enough to insure clined Rectangular Box Heated from Below," International Journal of Heat
and Mass Transfer, Vol. 20,1977, pp. 123-129.
two-dimensional flow at the midplane. For increasing Ra numbers 10 Ozoe, H., Sayama, H., Churchill, S., "Natural Convection in an Inclined
the flow develops to a boundary layer flow. For decreasing <j> the flow Rectangular Channel at Various Aspect-Ratios: Experimental Measurements,"
loses the boundary layer behavior and the velocities in the core region International Journal of Heat and Mass Transfer, Vol. 18,1975, pp. 1425-
increase. 1431.
11 Ozoe, H., Sayama, H., Churchill, S., "Natural Circulation in an Inclined
Rectangular Channel Heated on one Side and Cooled on the Opposing Side,"
Acknowledgment International Journal of Heat and Mass Transfer, Vol. 17, 1974, pp. 1209-
This work was supported by the Dutch Foundation for Pure Sci- 1217.
entific Research (ZWO). 12 Schinkel, W. M. M., "Natural Convection in Inclined Air-Filled Enclo-
sures," Ph.D.-thesis, Delft University of Technology, 1980.
13 Schinkel, W. M. M., Linthorst, S. J. M., Hoogendoorn, C. J., "The
References Stratification in Natural Convection in Vertical Enclosures," Proceedings 19th
1 Catton, I., "Natural Convection in Enclosures," Proceedings Sixth In- National Heat Transfer Conference, session Natural Convection in Enclosures,
ternational Heat Transfer Conference, Toronto, Canada, Aug. 7-11, Vol. 6, Orlando, Fla, July 1980, pp. 31-38.
1978, pp. 13-31. 14 Mallinson, G. D., De Vahl Davis, G., "Three-Dimensional Natural
2 Arnold, J. N., Catton, 1., Edwards, D. K., "Experimental Investigation Convection in a Box: a Numerical Study," Journal of Fluid Mechanics, Vol.
of Natural Convection in Inclined Rectangular Regions of Different Aspect- 83, Part 1,1977, pp. 1-31.
Ratios," ASME JOURNAL OF HEAT TRANSFER, Vol. 98, No. l, 1976, pp. 67- 15 Korpela, S. A., Gozum, D., Baxi, C. B., "On the Stability of the Con-
71. duction Regime of Natural Convection in a Vertical Slot," International
3 Krishnamurti, R., "On the Transition to Turbulent Convection. Part Journal of Heat and Mass Transfer, Vol. 16,1973, pp. 1683-1690.
1. The transition from Two- to Three-Dimensional Flow," Journal of Fluid 16 Bergholz, R. F., "Instability of Steady Natural Convection in a Vertical
Mechanics, Vol. 42, Part 2,1970, pp. 295-307. Fluid Layer," Journal Fluid Mechanics, Vol. 84,1978, pp. 743-768.
4 Krishnamurti, R., "On the Transition to Turbulent Convection. Part 17 Clever, R. M., Busse, F. H„ "Instabilities of Longitudinal Convection
2. The Transition to Time-Dependent Flow," Journal of Fluid Mechanics, Vol. Rolls in an Inclined Layer," Journal of Fluid Mechanics, Vol. 81, Part 1,1977,
42, Part 2,1970, pp. 309-320. pp. 107-127.
5 Krishnamurti, R., "Some Further Studies on the Transition to Turbulent 18. Batchelor, G. K., "Heat Transfer by Free Convection Across a Closed
Convection," Journal of Fluid Mechanics, Vol. 60, Part 2, 1973, pp. 285- Cavity Between Vertical Boundaries at Different Temperatures," Quarterly
303. Journal of Applied Mathematics, Vol. 12,1954, pp. 209-233.

Journal of Heat Transfer AUGUST 1981, VOL. 103 / 539

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 W. D. Bennon 1
Mixed Confection Heat Transfer
F. P. Incropera
Heat Transfer Laboratory,
School of Mechanical Engineering,
from Horizontal Cylinders in the
Purdue University,
West Lafayette, IN 47907
Crossflow of a Finite Water Layer
Heat transfer measurements and flow visualization studies have been performed for a sin-
gle horizontal cylinder, a single row of six horizontal cylinders, and two rows of 13 hori-
zontal cylinders submerged in an open channel flow. Regimes associated with free convec-
tion, combined convection, and forced convection have been delineated, and the effects
of finite boundaries have been resolved. For a restricted set of space ratios, correlations
of heat transfer data have been developed for the combined convection regime.

Introduction
Although frequently neglected, buoyancy effects can strongly in- itored using a digital wattmeter calibrated to an accuracy of ±1W over
fluence convection heat transfer from circular cylinders under forced the range 0 to 300 W.
flow conditions. The effects are particularly significant for low ve- As shown in the insert of Fig. 1, four rectangular grooves are milled
locity, high heat flux conditions and may influence the performance in each cylinder surface at 90 deg intervals, extending from one end
of systems involving waste heat utilization, nuclear waste disposal, to the midplane. Thermocouple leads are inserted in the grooves, with
reactor cooling, and sensible energy storage. junctions attached to the surface at the midplane. The grooves are
Several experimental studies have been performed to develop filled with a mixture of epoxy and finely ground copper filings, and
general heat transfer correlations for mixed convection conditions. the cylinder is polished with an abrasive cloth to provide a smooth
Zijnen [1] was partially successful in using a vectorial addition of surface condition.. The cylinder surface temperature is taken to be the
natural and forced convection results to correlate data obtained for arithmetic mean of the four readings.
air in crossflow over horizontal cylinders. There is little physical basis Inlet (upstream) and outlet (downstream) water temperatures are
for such an approach, however, and more emphasis has been placed measured using two vertical arrays of five thermocouples, each
on the use of an equivalent Reynolds number to correlate results for mounted directly in the channel sidewall and extending 25.4 mm from
a variety of fluids and flow conditions [2-4]. Other efforts [5,6] have the interior channel surface. The temperatures are evaluated by
concentrated on identifying regions associated with natural, forced, taking the arithmetic mean of the five values for an array. All ther-
and mixed convection. The regions have been delineated in terms of mocouple data are recorded and stored using an Esterline Angus
a Grashof number-Reynolds number ratio of the form Gr/Re", where Model PD-2064 programmable data acquisition system with a Ken-
n is typically in the range from 2 to 3, and some success has been nedy Model 1610 magnetic tape recorder. An uncertainty of ±0.5° C
achieved in using this ratio to correlate heat transfer data. In addition, is assigned to each thermocouple measurement.
several theoretical studies have been performed [7-9] and have been A closed flow system, which permitted variation of the water layer
useful in delineating regimes for which mixed convection effects are height and the flow velocity, was used for all of the experiments, with
significant. tests performed under darkened laboratory conditions. The ambient
Despite the foregoing efforts there remains a need for heat transfer air and inlet water temperatures were approximately 21°C, with
correlations which apply over a wide range of mixed convection con- temperature differences between the cylinder surface and the water
ditions and for which the effects of finite boundaries have been de- ranging up to 21 °C.
termined. Moreover, all studies performed to date have been con- The upstream velocity field in the channel was measured by using
cerned with a single cylinder, when, in fact, most applications involve the hydrogen bubble technique. A 50 mm long, 0.025 mm dia platinum
an array of cylinders. Accordingly, the primary objective of this study wire was attached to a support frame which could be traversed ver-
has been to measure and to correlate heat transfer parameters under tically and in the flow direction. The wire was electronically pulsed,
combined convection conditions for a single cylinder and for hori- and the local velocity was inferred from photographs of the bubble
zontal cylinder arrays. The experiments have been performed in a traces. From repetitive measurements made for velocities in the range
water channel for which the water layer height and cylinder placement of interest (0 to 65 mm/s), the uncertainty associated with this pro-
may be varied, and measurements have been performed for the pa-
rameter range 0 < ReD S 2300 and 0.04 < (Gra/Res 2 ) < ». The ef-
fects of finite boundaries have been determined, and the data corre-
sponding to the combined convection regimes have been corre-
lated.
^Groove Width =0.78 mm
Heated Copper ? groove Depth =0.91 mm
Experimental Procedures Cylinder

The experiments were conducted in an open water channel which Air-Water

re Interface
is 2.36 m long and constructed of 12.7 mm plexiglass. The channel has D = 25.4 mm
a fixed width of 280 mm, and up to 13 horizontal copper cylinders may S^3101.6 mm
Water
Flow S T =50.8mm
be inserted according to the arrangement of Fig. 1. The first cylinder H = 38.1 mm
of the array is located 1 m from the channel entrance and each cylinder
is drilled along its axis to permit insertion of an electrical heater.
Power to each heater is regulated by a voltage controller and is mon- © © O OO
T-0
1
Presently with Alcoa Technical Center, Alcoa, PA.
+0OOOOOO
Contributed by the Heat Transfer Division for publication in the JOURNAL { / / / / / / / / / / / / / / / / / I
OF HEAT TRANSFER. Manuscript received by the Heat Transfer Division
October 17,1980. Fig. 1 Longitudinal section of water channel with 13 cylinder array

540 / VOL. 103, AUGUST 1981 Transactions of the ASME

Copyright © 1981 by ASME
Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 cedure was estimated to be ±10 percent. The vertical distribution of 50
Hlmtn) Hc(mm)
the velocity was characterized by a central core of uniform velocity A 75 38.1
and upper and lower regions, approximately 25 mm thick, in which o 100 38.1
D 240 38.1
the velocity decreases with increasing distance from the core. Typi- °40 - m 240 88.9
cally, however, conditions were dominated by the central core, whose
velocity, Uo, was used to calculate the Reynolds number.
Observation of thermal plume development from the heated cyl- 30 - QpjQ—— D ~~
inders was achieved by using the shadowgraph technique. Collimated
light from a 100 W mercury arc lamp was focused and passed through
the channel sidewalls, and the transmitted light was cast on a thin 20 -
sheet of frosted glass. Shadows appearing as a result of density vari-
ations were used to infer qualitative features of the plume develop- ^Nu0=O.525Ra£
ment. 1 1 1 i
10
The average heat transfer coefficient for an N cylinder array may 10
RAYLEIGH NUMBER,RaD*l0=
be obtained from an expression of the form
Fig. 2 Single cylinder Nusselt number as a function of Rayleigh number for
_ 1 N _ 1 N Pj pure free convection
(1)
Tm,i)
and G r c / R e o 2 were determined to be approximately 10 and 20 per-
where hj, Pj and T c j are the average convection coefficient, the power cent, respectively, for the single cylinder experiments. The corre-
dissipation, and the average surface temperature associated with a sponding uncertainty in NUD depends strongly on UQ and Pj, varying
single cylinder. A Nusselt number for the array may then be expressed from a high of 20 percent for large values of Uo and small values of Pj
as NUD = hD/k, and, for the conditions of the study, may be presumed to a low of 4 percent for small values of £7o and large values of Pj.
to have a functional dependence of the form These high and low values increase to approximately 40 and 10 per-
cent, respectively, for the six cylinder array and to approximately 60
Nu D = / ( R e D , Gr fl , H/D, HJD, SJD, ST/D) (2)
and 15 percent, respectively, for the 13 cylinder array.
The Reynolds and Grashof numbers account for flow field effects
related to inertia and buoyancy forces, respectively, while the space Experimental Results: Single Cylinder
ratios account for cylinder spacing and finite medium effects. All Results obtained for Rerj = 0 (Gro/ReD 2 = °°) correspond to free
water properties appearing in these parameters were evaluated at a convection and are plotted as a function of Rayleigh number in Fig.
mean film temperature defined as Tf = [Tc + (Twj + Twfi)/2]/2. 2. The data are compared with the widely used correlation of McAd-
Three sets of experiments have been performed, corresponding to ams [11] for an infinite medium, and the excellent agreement (to
a single cylinder, a horizontal array of six cylinders in a single row, and within ±15 percent) suggests that the finite geometry of the experi-
a horizontal array of thirteen cylinders in two rows. For the single ments has little effect on the results. There is no consistent variation
cylinder, data have been obtained over the parameter range 0 < Rec of the data with the water layer height, H, and although the Nusselt
< 2300 and 0.025 < G r D / R e D 2 < «>, as well as for four different con- number appears to increase slightly with decreasing distance, Hc,
ditions related to cylinder placement and water layer height (H = 240 between the cylinder and the channel bottom, the trend is weak and
mm, Hc = 38.1 mm; H = 100 mm, Hc = 38.1 mm; H = 75 mm, Hc = well within experimental uncertainties. Although free convection heat
38.1 mm; H = 240 mm, Hc = 88.9 mm). For the six cylinder horizontal transfer from a cylinder in a finite medium is known to be enhanced
array, the parameter ranges were 0 < Reo ^ 2000 and 0.040 < by the establishment of recirculating flows [12, 13], such flows require
Grc/Rec 2 < « ( with three different combinations of cylinder place- the presence of a finite width, as well as a finite height, and are hence
ment and water layer height (H = 240 mm, Hc = 38.1 mm; H = 100 absent from the present experiments.
mm, Hc = 38.1 mm; H = 240 mm, Hc = 88.9 mm); while for the 13 The effect of Reynolds number is illustrated in Fig. 3, where rep-
cylinder array, the ranges were 140 < Reo < 2050 and 0.026 < resentative data are plotted for H = 100 mm and Hc = 38.1 mm. For
G r D / R e D 2 ^ 53, with H = 240 mm and H = 130 mm. The cylinder small values of the Reynolds number, ReD 5 200, the Nusselt number
diameter was fixed at D = 25.4 mm for all experiments, while the is independent of Reo, depending entirely on the heater power input
longitudinal pitch was fixed at Si = 101.6 mm for both the six and and hence on the Rayleigh number. Accordingly, free convection
13 cylinder arrays and the transverse pitch was fixed at ST = 50.8 mm correlations may be applied without error. With increasing Reynolds
for the 13 cylinder array. number, however, there is a gradual reduction in the dependence of
An experimental uncertainty analysis was performed according to NUD on RaD and an increasing dependence on ReD- The same trends
established procedures [10], and uncertainties in the parameters Rec were obtained for each of the other cylinder and water layer height

..Nomenclature-
A = cylinder surface area, m 2 Tc = average surface temperature for an
C; = constants in the heat transfer correla- N = number of cylinders in the array (1,6 or
tion _13) array of cylinders, £ Tcj\ IN, K
D = cylinder diameter, m NUD = average Nusselt number for the array
g = acceleration due to gravity, m/s 2 of cylinders, hD/k Tcj = average surface temperature for a
Gr D = Grashof number, g/3(Tc - TWti)D3/ NUD,/ = average Nusselt number based on single cylinder, K
i/ 2 forced convection correlation Tf = film temperature, K
H = water layer height, m Pj = heater power dissipation, W Tw,i - water inlet (upstream) temperature,
Hc = distance from cylinder centerline to PL,PT ~ dimensionless longitudinal and K
channel bottom, m transverse pitch (P = S/D) Tw,o — water outlet (downstream) tempera-
h = average heat transfer coefficient for the Pr = Prandtl number ture, K
_ array of cylinders, W/m 2 -K Rao = Rayleigh number, GroPr f/o = uniform core velocity associated with
hj = average heat transfer coefficient for a Reo = Reynolds number, UoD/v channel flow, mm/s
single cylinder, W/m 2 -K Si,ST = longitudinal and transverse pitch, /3 = thermal expansion coefficient, K~x
k = thermal conductivity, W/m-K v = kinematic viscosity, m 2 /s

Journal of H@at Transfer AUGUST 1981, VOL. 103 / 541

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 100 u.ou Q 52 a3
o P = |0OW NuD|f=(035-0.56Re D )Pr °
o P = I50W
A P=Z00W Q0.40 °a a H=240mm,Hc=38.lmm
80 O P=250W 0 H=240mm,Hc=88.9mm
V P=300W li P=300W
DQ^V - \ 0.30
60 ID

'o °-20 d?
OD n D
40 H=l00mm \z oo Q
H =38.1mm - 0.10 ° D o n
° o
1 1 1 1
400 800 1200 1600 2000 2400
REYNOLDS NUMBER,Re~
400 800 1200 1600 2000 2400 Fig. 4 Variation with Reynolds number of the difference between experi-
REYNOLDS NUMBER,Re„ mental single cylinder Nusselt numbers and predictions from a standard forced
Fig. 3 Single cylinder Nusselt number as a function of Reynolds number for convection correlation [11,14]
different heater power setting

combinations. Although the data revealed a slight increase in NUD I00

with decreasing H for large Reo, differences were within the experi- P(W) =
mental uncertainty and space effects could again be neglected. 50 I0OI5O2OO25O350
80
Figure 4 reveals the error associated with using a forced convection
correlation to predict the mixed convection results of this study. Heat
' a I o I AI v / o /O FORCED CONVECTION

transfer data obtained for P = 300 W and two of the geometrical

conditions are plotted as (NUD — N U D / ) / N U D , where NUD,/ is ob-
tained from a standard forced convection correlation of the form [11,
14]

NUD,/ = (0.35 + 0.56Re D °' 62 )Pr 0 - 30 (3)

The discrepancy is large for Reo = 200, where buoyancy effects en-
hance heat transfer by approximately 50 percent, and decreases with
increasing Reo- Even for Rec a* 2000, however, the data remain
consistently underpredicted by the correlation. Fig. 5 Single cylinder heat transfer data corresponding to tf = 240 mm and
The ratio G r c / R e o 2 may be used to delineate the data according Hc = 38.1 mm, with mixed convection boundaries delineated by the criteria
to zones. In Fig. 5 representative raw data are plotted in terms of flow of Fand and Keswani [6]
velocity and temperature difference, and the four zone boundaries
determined by Fand and Keswani [6] are superposed on the plot. For of typical forced convection correlations and the third term accounts
a prescribed power, P, the value of (Tc — Twj) provides a measure of for free convection effects. The constants Ci to Cg were evaluated from
the average cylinder heat transfer coefficient through equation (1). a nonlinear regression analysis of the data, and the resulting corre-
For a fixed P and values of Uo £ 10 mm/s, the value of (Tc — TWii) is lation is of the form
independent of UQ. Hence NUD is independent of Ren, and the range
Nu f l = [4.11 + 0.20 Re/, 0 - 66 + 0.71(Gr/,/ReD 2 ) 0 - B2 ]Pr 1/3 (5)
of Gro/Reo 2 > 40 is one of free convection. For 2 < Gro/Reo 2 < 40,
forced convection effects become discernable, and for a prescribed All of the data which are in the range 175 < R e B < 2300 and 0.025 <
heater power, the trend is generally one of decreasing (T c — Twj), and Grrj/Rer>2 < 48 are correlated by this expression to within ±15 per-
hence increasing h, with increasing Ug. This trend is also character- cent, irrespective of the values of H and Hc, and 93 percent of the data
istic of the region Gro/ReD 2 < 2, although the effect becomes less are correlated to within ±10 percent. For Reo < 175, the data are
pronounced for G r o / R e o 2 < 0.5, which is a region associated with correlated to within ±15 percent by the free convection correlation
domination by forced convection effects [6]. of McAdams [11].
Transition from free to forced convection is also characterized by The foregoing results may be compared with those of Fand and
significant changes in flow conditions. For free convections at low to Keswani [6], who divided their data into four regions according to the
moderate values of Gr/), it is known [15,16] that a laminar boundary value of Gro/Rerj 2 . For Grc/Re/) 2 < 0.5, heat transfer is primarily
layer originates at the lower stagnation point, develops around the by forced convection and their data were correlated by an expression
cylinder, and merges at the upper stagnation point without separation. of the form
A plume is formed above the cylinder and rises vertically until it is
deflected by the air-water interface. This behavior was also observed N u D = (0.255 + 0.699 Re D °- 5 )Pr a 2 9 (6)
with the shadowgraph, along with changes which occur due to su- All comparable data obtained in this study are correlated to within
perposition of a horizontal, forced flow. In particular, with decreasing ±20 percent by this expression, although 95 percent of the data exceed
GrrVReD2, the point of boundary layer merger and plume detachment the predictions. Natural convection effects begin to become significant
is shifted clockwise. For values of Gro/Reo 2 » 40, this shift is slight, for 0.5 < G r n / R e o 2 < 2.0 and are comparable to those of forced con-
as the cylinder boundary layer development remains dominated by vection for 2.0 < Gr/j/Rec 2 < 40. Although Fand and Keswani pro-
buoyancy forces. With decreasing Grrj/Reo 2 , inertia forces become pose the following correlation for the first of these regimes,
more prominent, as the angle associated with plume detachment in-
creases, and eventually (Gro/Ren 2 5 0.5) inertia forces dominate over Nu D = [0.255 + 0.699Re/, 0B + 0.188(Gr D /Re B 2 )Gr D 0 - 2B ]Pr 0 - 29
those associated with buoyancy. Boundary layer separation is then (7)
observed to occur, as a wake develops downstream of the cylinder.
In view of the foregoing results and the fact that the data are only no correlation is presented for the second regime. The data of this
weakly dependent on H and Hc, it is reasonable to seek a heat transfer study are within ±17 percent of predictions based on equation (7),
correlation of the form although 94 percent of the data exceed the predictions. For Grrj/Reo 2
> 40, free convection effects dominate and Fand and Keswani rec-
Nurj = [Ci + C 2 Re D c 3 + C 4 (Gr D /Re D 2 ) c 5]Pr 1/3 (4) ommend an expression of the form

where the first two terms on the right-hand side resemble the form Nu D = 0.406(1.7GrD + Re B 2 ) 1 / 4 Pr 1 /< (8)

542 / VOL. 103, AUGUST 1981 Transactions of the ASME

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 The limited data obtained in this study are correlated to within ±9 Air-Woter
w Interface
percent by this expression. Accordingly, the results of this study are
consistent with those of Fand and Keswani, although equation (5) is O
CYL 1
O
CYL 2
o o O
CYL 5
©Jr
CYL 6 "c
more versatile through its applicability to a wider range of Gro/ CYL 3 CYL 4

Re D 2 . a.
(Grn/Re? 1-0.20,

Experimental Results: Six Cylinder Array

The shadowgraph was also used to observe hydrodynamic effects
associated with the six cylinder array. For free convection (Oro/Rea 2
£30
= <»), a laminar, vertical plume originates from each cylinder and rises
to the air-water interface, where it is deflected to form a thin surface
layer of warm water. Although plume oscillation and recirculation 2 3 4 5
effects have been observed in similar experiments [16], no such effects CYLINDER NUMBER
could be discerned in this study. With the superposition of small to Fig. 6 Distribution of Nusselt numbers for the six cylinder array (H = 240
moderate forced flows, (0.5 < Oro/Reo 2 ), the point at which the mm, Hc = 38.1 mm, P = 300 W)
plumes detach from the cylinders is shifted clockwise and the plumes
IOO
are cantered at an angle which increases with decreasing Gro/ReD 2 - O H=240mm, H.=38.lmm
Plumes generated by each of the six cylinders do not interact with the O H=IOOmm, H c =38.lmm
a. e H=240mm, H = 88.9mm
other cylinders, but mix in the free stream (above the array) as they g 80
ascend to the air-water interface. For Gro/Reo 2 ^ 0.5, however,
plume cylinder interactions begin to occur until each cylinder
downstream of the lead cylinder is submerged in a turbulent flow 60
originating from the wakes of the preceding cylinders. orfP3
ID MM
The influence of the foregoing hydrodynamic effects on the cylinder jf 40
Nusselt number is shown in the representative results of Fig. 6. For
(Gro/Rec 2 ) £ 0.5, there is little difference between cylinder Nusselt , 20'
numbers, and it may be concluded that the cylinders act as isolated
heat sources for which mutual thermal and hydrodynamic interactions
are negligible. For (Gri>/Re.D2) = 0.20, however, the Nusselt number ** "O 500 I000 1500 2D00
of the second cylinder in the array is approximately 10 percent larger REYNOLDS NUMBER, ReD
than that of the lead cylinder, which may be attributed to flow sepa- Fig. 7 Effects of Reynolds number, water layer height, and array placement
ration from the lead cylinder. The gradual decline in the Nusselt on the average Nusselt number for the six cylinder array
number from cylinders three to six can be attributed to a progressive
warming of the water, which causes the cylinder surface temperature Experimental Results: Thirteen Cylinder Array
to rise for a fixed heat dissipation. Since the upstream water tem- Hydrodynamic effects associated with the thirteen cylinder array
perature is also fixed, the Nusselt number must decline. have also been observed shadowgraphically. As for the single cylinder
Effects on the average array Nusselt number of the Reynolds and the six cylinder array, free convection (Grrj/Rez) 2 = °°) is char-
number, water layer height, and the separation of the array from the acterized by laminar plumes which rise to the air-water interface and
channel bottom are shown in Fig. 7. For ReD < 300, the Nusselt form a thin surface layer of warm water. Hence each cylinder acts as
number is only weakly dependent on Rec and is independent of H an isolated heat source. For small to moderate forced flows (0.5 S
and Hc. Variations in the data are due entirely to different values of G r B / R e D 2 < 40), the rear stagnation points are shifted clockwise and
the power input. Moreover, the results agree well with those obtained the plumes leave the cylinders cantered at angles which increase with
for the single cylinder (Fig. 3), supporting the supposition that each decreasing Gro/ReD 2 . Plumes generated by the lower row of cylinders
cylinder acts as an isolated heat source and is unaffected by the interact with the upper row, and above the array the plumes mix and
presence of other cylinders. For Reu 5; 300, however, the Nusselt are carried downstream as they ascend to the air-water interface. For
number increases with increasing Reynolds number, and for Rerj S; G r D / R e o 2 5 0.5, each cylinder downstream of the leading cylinders
700 the results are influenced by H and Hc. In particular, there is a is submerged in a turbulent flow originating from the wakes of the
small, but discernable increase in NUD with increasing Hc, although preceding cylinders.
differences are within experimental uncertainties. There is a larger
The influence of the foregoing hydrodynamic effects on the cylinder
increase in NUD with decreasing H, and differences are most pro-
Nusselt number is shown by the representative results of Fig. 8. For
nounced for the range 700 < Reo 5 1500. This increase in Nusselt
(Gro/Reo 2 ) = 40.4, Nusselt numbers associated with cylinders in a
number with decreasing water layer height may be attributed to re-
given row are approximately the same, although the Nusselt numbers
circulation effects. With decreasing H, the space available for free-
associated with the upper row are approximately 25 percent less than
stream plume mixing is reduced and the probability of recirculating
those of the lower row. This result is due to the fact that, although free
plumes interacting with downstream cylinders increases. With in-
convection effects are dominant, the forced flow is sufficient to deflect
creasing Reynolds numbers above 1500, however, all plumes are swept
plumes from the lower row into cylinders of the upper row. Accord-
horizontally in the direction of the forced flow and the probability of
ingly, surface temperatures associated with cylinders of the upper row
freestream mixing and recirculation from above the array is reduced.
are larger than those of cylinders in the lower row, and the Nusselt
Hence the effect of H on NUD should diminish.
number must be correspondingly smaller. Similar results have been
If the effects of H and Hc on the heat transfer data are neglected, observed by Eckert and Soehngen [17] for free convection from heated
a correlation of the form prescribed by equation (4) may be attempted cylinders in a vertical array. With decreasing (Gro/Reo 2) differences
for the average array Nusselt number. Obtaining the constants C\ to between results for the top and bottom rows diminish, as the effects
C 6 from a nonlinear regression analysis of the data for Reo > 140, the of the forced flow become more significant. For (Gru/ReD 2 ) = 0.18,
following expression was found to correlate all results to within ±26 forced flow effects become dominant and the results are characterized
and 94 percent of the data to within ±15 percent by increasing Nusselt number for the first few cylinders and a gradual
decline for the cylinders which follow. The increase for the leading
NUD = [7.34 + 0.019Re D 0 - 94 + 0.92(Gr D /Re D 2 )°- 47 ]Pr 1 <' 3 (9) cylinders may be attributed to flow separation effects, while the
For ReD < 140, the data are correlated to within ±15 percent by the subsequent decline may be attributed to increases in the cylinder
free convection correlation of McAdams [11]. surface temperature due to the rising water temperature.

Journal of Heat Transfer AUGUST 1981, VOL. 103 / 543

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 IOO
Air-Wot er a,® P^ICOW
V Interface o
O,® P^ISDW
Flow O O O O 0 O T \£ & A P( =200W
^ C Y L 2 ^ . C Y L 4 ^ C Y t 6 ^ C T T - 8 ^ . C T L 1 0 ^ . CTL 1 2 ^ H V , ^ P »250W

uj
CYLI
" "CYL3
' «LS CYL7
H = l30,»m CYL9
/ CYLI1
" " CYU13 1 s
z
§60 - m&
ti 60 8 *
* (G^/Rl^l'O.ia/
LU
b
CO (GrD/ft£)«l.7k
I _ ce»
5 40
z
8 § ^
Q P UJ 20 OPEN SYMBOLS: H = 240mm
Z SOLID SYMBOLS: H» 130mm
_l
l5 1 i i i i i i i i i i 1
° 1 2 3 T " 1 T T ~ 8 9 10 II 12 13 1 1 _ ._l 1
CYLINDER NUMBER 500 1000 1500 2000
REYNOLDS NUMBER, ReD
Fig. 8 Distribution of Nusselt numbers for the 13 cylinder array (H = 130
mm, P = 250 W) Fig. 9 Effects of Reynolds number, cylinder power dissipation, and water
layer height on the average Nusselt number for the 13 cylinder array
The effects of Reynolds number, cylinder power dissipation, and be approximated as an isolated heat source characterized by the same
water layer height on the average array Nusselt number are shown Nusselt number. For Grn/Reo 2 5 0.5, however, interactions between
in Fig. 9. For H ~ 240 mm, the Reynolds number dependence is typ- cylinder flow fields induce a small increase in the Nusselt number of
ical of that obtained for the six cylinder array and the results are only the second cylinder, which is followed by a gradual decline in the
weakly dependent on the cylinder power dissipation. The strongest Nusselt number of subsequent pylinders. For the 13 cylinder array,
dependence occurs at low values of Rep and the increase in Nun with isolated heat source behavior may not be assumed for Gru/ReD 2 §
increasing Pj is due to the importance of free convection effects. The 50. For larger values of Grn/Ren 2 associated with this range, thermal
same trends characterize the low Reynolds data, which do not appear plumes generated in the bottom row of cylinders interact with the top
in the figure, for the shallow water layer (H - 130 mm). However, at row to significantly reduce Nusselt numbers associated with this row.
larger values of Rerj, NUB begins to decrease significantly with in- With decreasing (Gr£>/ReD2) differences in the Nusselt numbers
creasing Pj for the shallow layer. This trend is attributed to the effect between rows diminish; while in the lower range of (Gvo/Reo 2 ), there
which both increasing Re/j and decreasing H have on increasing the is an increase in the Nusselt number for cylinders immediately
probability of warm water heated by the lead cylinders interacting downstream of the lead cylinders. This increase is followed by a
with the downstream cylinders. Note that, for moderate to large values gradual decline in the Nusselt number associated with subsequent
of Reo, the average array Nusselt number increases with decreasing cylinders.
water layer height, which is consistent with results obtained for the
six cylinder array. 4 Neglecting space ratio effects, the average Nusselt number for
the six cylinder array is correlated to within ±26 percent by equation
If space ratio effects are neglected, a correlation of the form pre-
(9) for 140 < R e B < 2000 and 0.04 < (Gr/j/Rerj 2) < 54. Similarly, the
scribed by equation (4) may also be obtained for the 13 cylinder array.
average Nusselt number for the thirteen cylinder array is correlated
Applying the nonlinear regression analysis to the data, it follows
to within ±23 percent by equation (10) for 140 < Ren < 2050 and
that
0.026 < G r D / R e D 2 < 53.
NUD = [9.97 + 0.039 Re fl 0 - 83 + 0.54(Gr D /Re D 2 )- ( ! M ]Pr 1 / 3 (10) Because they pertain to a restricted set of space ratios, caution
should be exercised in using the correlations of this study. For each
All of the data which are in the range 143 < Re D < 2046 and 0.026 < of the three geometries, the correlations are applicable when H/D >
(Grn/Ren 2 ) < 53 are correlated to within ±23 percent by this ex- 2.95 and Hc/D & 1.5. However, if the cylinders are placed close to the
pression and 92 percent of the data are correlated to within ±15 air-water interface (H/D < 2.95) and/or close to the channel bottom
percent. Much of this discrepancy may be attributed to the ignored (Hc/D < 1.5), thermal and hydrodynamic interactions may introduce
effects of H on Nu/j. Note that, unlike the preceding correlations, the significant error.
exponent on the (Gra/Rez) 2 ) term is negative, indicating that Although results obtained for the six cylinder array pertain to a
buoyancy acts to reduce the overall heat transfer from the array. This single longitudinal pitch (PL = 4), they should be applicable to a wider
result is due to the effect of plume-cylinder interactions within the range of Pi, so long as each cylinder in the array continues to behave
array. For Grn/Rerj 2 > 53, plume deflection is negligible and, because as an isolated heat source. It is only for small values of PL —• 1 that
of the staggered arrangement, each cylinder behaves as an isolated the results may become suspect. The same may not be said for the
heat source for which the standard free convection correlation may thirteen cylinder array, for which the results pertain to fixed longi-
be applied. tudinal and transverse pitches of PL — 4 and PT = 2, respectively.
With decreasing Pj. a n d PT, thermal and hydrodynamic interactions
Summary are enhanced; while for increasing PL and PT, the cylinders behave
Heat transfer measurements have been performed for a single more like isolated heat sources. In each case the effect on heat transfer
horizontal cylinder and for horizontal arrays of six and 13 cylinders is likely to be significant.
submerged in an open water channel flow. Major conclusions of the
study are as follows: Acknowledgments
1 For Ren 5 150 and Gro/Re,D 2 >; 50, forced convection effects The authors wish to express their appreciation to Mr. Gregory
are negligible, and in all cases the cylinders behave as isolated heat Terpay for his assistance in performing the flow visualization
sources. Heat transfer data are correlated to within ±15 percent by studies.
the standard expression for free convection in an infinite medium
[11]. References
2 For the single cylinder in the range 175 < Rec < 2300 and 0.025 1 Zijnen, B. G. Van der Hegge, "Modified Correlation Formulae for Heat
2
< Gr/j/Rez? <'48, combined convection effects must be considered Transfer by Natural and Forced Convection from Horizontal Cylinders," Ap-
and the data are correlated to within ±15 percent by equation (5). For plied Scientific Research, Vol. A6, No. 1,1956, pp. 129-140.
Reo > 2300 and Grrj/Reo 2 < 0.025, buoyancy effects may be ne- 2. Borner, H., "Ober den Warme-und Stoffubergang an Umspiilten Ein-
zelkorpern bei Oberlagerung von Freier und Erzwungener Stromung," Verins
glected, and a standard forced convection correlation, such as equation Deutscher Ingenieur Forschung.,No. 512,1965.
(3) or (6), may be used with ±20 percent accuracy. 3 Hatton, A. P., James, D. D., and Swire, H. W., "Combined Forced and
3 For Gro/Refl 2'&0.5 in the six cylinder array, each cylinder may Natural Convection with Low-Speed Air Flow over Horizontal Cylinders,"

§44 / VOL. 103, AUGUST 1981 Transactions of the ASME

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 Journal of Fluid Mechanics, Vol. 42, Part 1,1970, pp. 17-31. 11 McAdams, W. H., Heat Transmission, McGraw-Hill, 1954, pp. 175-
4 Morgan V. T., "The Overall Heat Transfer from Smooth Circular Cyl- 177.
inders," J. P. Hartnett and T. F. Irvine, Jr., eds., Advances in Heat Transfer, 12 Marsters, G. F., "Natural Convective Heat Transfer from a Horizontal
Vol. 11,1975, pp. 199-264. Cylinder in the Presence of Nearby Walls," Canadian Journal of Chemical
5 Sharma, G. K. and Sukhatme, S. P., "Combined Free and Forced Con- Engineering, Vol. 53,1975, pp. 144-149.
vection Heat Transfer from a Heated Tube to a Transverse Air Stream," ASME 13 Yaghoubi, M. A. and Incropera, F. P., "Natural Convection from a
J O U R N A L O P H E A T T R A N S F E R , Vol. 91, No. 3,1969, pp. 457-459. Heated Horizontal Cylinder in a Shallow Water Layer," Proceedings of the
6 Fand, R. M. and Keswani, K. K., "Combined Natural and Forced Con- Sixth International Heat Transfer Conference, Vol. 2, NC-15, 1978, pp.
vection Heat Transfer from Horizontal Cylinders to Water," International 269-274.
Journal of Heat and Mass Transfer, Vol. 16,1973, pp. 1175-1191. 14 Fand, R. M., "Heat Transfer by Forced Convection from a Clyinder to
7 Sparrow, E. M. and Lee, L., "Analysis of Mixed Convection about a Water in Cross Flow," International Journal of Heat and Mass Transfer, Vol.
Horizontal Cylinder," International Journal of Heat and Mass Transfer, Vol. 8,1965, pp. 995-1010.
19,1976, pp. 229-232. 15 Pera, L. and Gebhart, B., "Experimental Observations of Wake For-
8 Mucoglu, A., and Chen, T. S., "Analysis of Mixed Convection across a mation over Cylindrical Surfaces in Natural Convection Flows," International
Horizontal Cylinder with a Uniform Surface Heat Flux," ASME Paper No. Journal of Heat and Mass Transfer, Vol. 15,1972, pp. 175-177.
77-HT-45, AIChE-ASME Heat Transfer Conference, Salt Lake City, Aug. 16 Incropera, F. P. and Yaghoubi, M. A., "Buoyancy Driven Flows Origi-
15-17,1977. nating from Heated Cylinders Submerged in a Finite Water Layer," Interna-
9 Merkin, J. K., "Mixed Convection from a Horizontal Circular Cylinder," tional Journal of Heat and Mass Transfer, Vol. 23,1980, pp. 269-278.
International Journal of Heat and Mass Transfer, Vol. 20,1977, pp. 73-77. 17 Eckert, E. R. G. and Soehngen, E., "Studies on Heat Transfer in Laminar
10 Kline, S. J. and McClintock, F. A., "Describing Uncertainties in Sin- Free Convection with the Zehnder-Mach Interferometer," USAF Tech. Rep.
gle-Sample Experiments," Mechanical Engineering, Vol. 11,1953, pp. 3-8. 5747,1948.

Mandatory Excess-Page Charges for Transactions

Effective J u l y 1,1981, all Transactions p a p e r s t h a t exceed t h e s t a n d a r d l e n g t h (six pages), will b e assessed a m a n d a t o r y
p a g e c h a r g e of 125 dollars p e r page for t h o s e p a g e s e x c e e d i n g t h e six p a g e limit.

Journal of Heat Transfer AUGUST 1981, VOL 103 / 545

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 A Horizontal Flow Past a Partially
L. S. Yao
F. F. Chen
Graduate Student.
Heated, Infinite Vertical Cylinder
The free-convection boundary layer along a partially heated infinitely long vertical cylin-
Department of Mechanical and
der disturbed by a steady horizontal flow is studied. The asymptotic solution indicates
Industrial Engineering,
University of Illinois at Urbana-Champaign, that the boundary layer is mainly induced by the buoyancy force in the neighborhood of
Urbana, III. 61801 the thermal leading edge. The effect of the horizontal free stream on the boundary layer
gradually increases as one moves upward away from the thermal leading edge along the
cylinder. The boundary-layer separation does not occur in the neighborhood of the ther-
mal leading edge. However, the asymptotic solution shows that the forced convection
tends to separate the boundary layer along the aft body of the cylinder. The importance
of the effect of the forced convection is governed by the parameter tz1?2 where e is the in-
verse of the square root of the Richardson number. The asympotic solution is valid as long
as the value of ez 1 / 2 is small, and is not explicitly restricted by the Richardson number.

1 Introduction
Owing to the shortage of oil and natural gas, and their ever in- Sukhatme studied the combined free and forced convection from a
creasing prices, solar power is increasingly being considered as one horizontal cylinder and suggested that the expansion parameter for
of the viable alternate energy resources in electricity generation. One a free-convection dominant flow is the inverse of the Richardson
of the preliminary designs of solar power plants contains a central number.
receiver which is constructed by a series of vertical pipes and looks Three-dimensional combined free and forced convection has not
like a large vertical cylinder [1]. The receiver is designed as a once received much detailed study. Young and Yang [7] studied the effect
through heat exchanger to boil the water by solar heating. The surface of small cross-flow and surface temperature variation on laminar free
temperature of the receivers is much higher than the ambient tem- convection along a vertical plate. By coincidence, Eichhorn and Hason
perature. A significant free convection is induced by the density [8] and Plumb [9] report their studies of the mixed convection about
stratification of air inside the thermal boundary layer. Simulta- a vertical surface in cross-flow at the same time. The asymptotic so-
neously, a breeze may pass by the receiver and disturb the free con- lutions for a forced-convection dominant flow for both internal flow
vection around the receiver. The resulting free and forced convection (pipe flow) [10] and external flow [11, 12] have been obtained by
flow is three-dimensional. The distributions of temperature and flow perturbations on the forced convection. The expansion parameter is
are needed to design the structure for rigidity and to estimate the still the Richardson number. The numerical solution of the boundary
free-forced convective heat loss. layer along a heated horizontal cylinder shows that the range of va-
Due to the large dimension of the receiver and the likely high am- lidity of the asymptotic solution is limited to a small region close to
bient turbulence level, the boundary-layer flow around the receiver the leading edge of the cylinder [13] where the free-convection effect
is expected to be turbulent. In particular, the vertical pipes of small is small. However, the importance of the free convection grows
radius, which form the receiver, behave as roughened surfaces. In this downstream even for a small Richardson number.
paper, however we concentrate on the interaction of the free and the In this paper, we study the free-convection dominant flow on a
forced convections around the receiver by studying the laminar flow partially heated infinite vertical cylinder. The forced convection is
along an infinitely long vertical cylinder. considered due to a steady breeze and is treated as a perturbed effect.
The free-forced convection is a fundamentally interesting problem
in convective heat transfer. Oosthnizen and Leung [2,3] have studied
the problem for an infinite cylinder with a finite heating length ex- z,w
perimentally and numerically. They have found that the accuracy of
the numerical solution deteriorates in the region of the thermal
leading edge. Their computation was started a short distance above
the thermal leading edge. Consequently, an error due to mistreating
the leading-edge conditions is to be expected. A detailed analysis is
carried out in this paper to describe the flow development and heat
transfer mechanism in this region. Before we start describing the
physical model, a brief review of some relevant works is summa-
rized.
Free convection has been extensively studied for the past half
century, since Schmidt and Beckmann [4] measured the thermal
boundary layer along a heated vertical plate. Pohlhausen demon-
strated that the similarity solution exists for the free-convection
boundary layer. Since then, series solutions have been thoroughly
developed for free convections over arbitrary shape bodies. A com-
prehensive review can be found in Lin and Chao [5].
Two-dimensional combined free and forced convection has also
attracted substantial attention. Most of them are either by a method
of parameter perturbation on the forced convection, or by a numerical
method. The expansion parameter is the Richardson number. A re-
view of the works can be found in Joshi and Sukhatme [61. Joshi and

Contributed by the Heat Transfer Division for publication in the JOURNAL

OP HEAT TRANSFER. Manuscript received by the Heat Transfer Division
September 9,1981. Fig. 1 Physical model and coordinates

546 / VOL. 103, AUGUST 1981 Transactions of the ASME

Copyright © 1981 by ASME
Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 The perturbation parameter is the inverse of the square root of the d2v 1 du v 1 d2v 2 dud2v]
Richardson number; Reynolds number, Re, and Grashof number, Gr, + " -z^2 + z-z-
dr r dr — + — — - + - + —— (ic)
are the implicit parameters for this problem. In this paper, we assume r2 r2 d<j> r2 d(f> dz 2 ]
Re is large to insure the existence of a forced-convection boundary _ dw v dw _dw —1 dp
u— + 3 r w—— = -+ v d2w 1 dw 1 d2w r>2r
layer. It can be shown that the displacement effect of the free-con- dr r d<t> dz p dz
vection boundary layer is 0(Re/Gr 3 / 2 ). In the following analysis, we dr2 f dr 2
r d0 2

assume that Re is smaller than Gr 3/2 . This rules out the possibility that + Bg(T-T„) (Id)
the axial pressure gradient induced by the free-convection boundary _dT <vdT _dT d2T 1 dT 1 d2T
u —3 + — 1- w — = a He)
layer becomes 0(1). Besides the restriction stated above, the pertur- dr r d</> dz dP + r dr + r2 d(p2
bation solution is not limited by the value of e; instead, it is valid as
The fluid density variation due to heating is considered to be only
long as e • z 1 / 2 is small. This implies that the smaller e is the larger
important in the form of the buoyancy force, equation (Id). This is
region (or large z) where the perturbation solution can be applied. It
known as the Boussinesq approximation.
is worthy to point out that f • z1/2 can be rewritten as (Re/VGrj).
The flow and thermal conditions considered in the paper are that
(2/a) 2 where Gr? is the Grashof number in terms of 2. This indicates
the Grashof number is large to insure the existence of the thermal
that the characteristic length of the free convection is 2. x Since the
boundary layer; however, its value should not be too large to cause the
thermal leading edge provides a division line for 2 > 0 and 2 < 0, and
free-convection boundary layer to be turbulent. The magnitude of
the flow development for 2 < 0 has little practical interest, our study
the Reynolds number (defined in equation (2)) is not explicitly re-
is concentrated for 2 > 0.
stricted, as long as its value is not too large to make the Richardson
The asymptotic solution shows that the free convection dominates number small. This implies that Re 2 < Gr.
the heat loss in the region close to the thermal leading edge. However, The flow close to the leading edge (2 = 0) is very complicated. Since
the importance of the forced convection increases as one moves ver- the heat conduction in the 2 direction is important, the thermal
tically upward away from the thermal leading edge even when the boundary layer starts a small distance ahead of the leading edge. The
magnitude of the breeze is small compared with the buoyancy effect. thermal penetration depth, 5, is proportional to \f)g • (Tw — T„)/v]~1/s
The region in which the free convection is the dominant flow com- and is independent of the geometric length (the variables are defined
ponent shrinks as the magnitude of the breeze increases. No flow in equation (2)). The induced velocity is proportional to v/8 = [Bg(Tw
separation along the aft body of the cylinder is found in the region - T„)v 2 ] 1 / 3 . This is much smaller than the thermal boundary layer
where the free convection is dominant. However, the solution does velocity whose magnitude is proportional to y//3g{Tw — Ta)z. This
indicate the flow separation can occur as one moves vertically upward has been verified by comparing Pohlhausen's solution with the
where the forced convection gradually becomes important. Numerical Schmidt and Beckmann's data. In the analysis, we do not consider
solution [14] shows that, however, the separation starts from the rear the leading-edge effect; therefore, the result cannot be applied to the
stagnation point about 2 = 0.0056 • ah2 for Prandtl number, Pr, equal immediate neighborhood of the leading edge.
to 10, and z = 0.0044 • ale2 for Pr = 0.733.
For a free-convection dominant flow, the magnitude of the
buoyancy force should be the same order as that of the inertial force.
2 Physical Model and Analysis The dimensionless variables are introduced in equation (2).
The physical model considered is an infinite vertical cylinder of
radius a, shown in Fig. 1. A uniform horizontal free stream is flowing 2
by the cylinder. The velocity and the temperature of the free stream 2 = _
, (coordinates)
are U„ and To, respectively. The surface temperature of the cylinder
u,v,w = u, v, wl (velocities)
for 2 > 0 is held constant at Tw, which is larger than T„. For 2 < 0, it
is unheated. The governing equations in cylindrical polar coordinates, (VGi-v/a)
Fig. 1, are p = P/(p Gr v2/a2) (pressure)
6= (T- Ta)/(TW - T„) (temperature) (2)
du u 1 dv dw
0 (la)
_ +- + +—: Re = (Vaa)/p (Reynolds number)
dr r r d<j> dz
_ du v du v2 _du -ldp Gr = [8g(Tw - T„)a 3 ]/i/ 2 (Grashof number)
u— +- — + 10— = —
dr r ixj> r dz pdr e = Re/VOT (the inverse of the square root
d2u 1 du u 1 d2u 2 dU d2u\ of Richardson number)
+ v (lb)
dr2
r dr ' r 2 2
r d<t>2 ._ + __ Pr = via (Prandtl number)
r2 d(j> dz''
_dv v dv uv _ du =
—1
—
dp where v is the kinematic viscosity; a, thermal diffusivity; B, the
u—+ 1- — + w—z r—
dr r d(/> r dz pr d$ thermal expansion coefficient.
Inviscid Flow. Complete solution of equations (1) may be ob-
tained by a numerical method, such as finite-difference method. It
is well known that the required number of nodes for flows of large
1
The authors wish to acknowledge that the referee pointed out this important Reynolds numbers is usually beyond the capacity of the current
fact. generation computers; in particular, a thin boundary layer exists along

.Nomenclature-
Re = Reynolds number
a = radius of the cylinder T = temperature cj> = azimuthal coordinate
/ = stream functions, equations (17-21) u,v,w — boundary-layer velocities r = shear stress
8 = gravitational acceleration U, V, W = inviscid flow velocities
G = temperature functions, equations (17- 2 = axial coordinate Subscripts
21) B = thermal expansion coefficient co = surface
Gr = Grashof number, equation (2) f = expansion parameter, equation (2) <*> = free stream
Nu = Nusselt number rj = similarity variable, equation (17e)
P = pressure 6 = dimensionless temperature, equation Superscripts
Pr = Prandtl number (2) — = dimensional quantities
r
- radial coordinate v = kinematic viscosity • = derivative with respect to 7/

Journal of Heat Transfer AUGUST 1981, VOL 103 / 547

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 the cylinder for the considered problem. On the following, we apply
the method of matched asymptotic expansion to solve equations dr dz r d<l> r r d<j>
(1). dP2
The zeroth-order outer solution is the potential flow with the slip Ui 1- W i 1 = -, (10c)
dr dz r dc/> dz
conditions on the cylinder. The zeroth-order inner solution is the
boundary-layer flow with the nonslip conditions on the cylinder and dU2 U2 dW2 ldV2
(lOd)
matching with the zeroth-order outer solution far away from the dr r dz r d$
cylinder. The expansion procedures for the outer and the inner flows
The boundary conditions associated with equations (86) and
are described below in detail.
(lfta-lOc) are
The dimensionless equations governing the flow outside of the
thermal boundary layer can be obtained by substituting equations Wi and Vi cos 0 + Ui sin <l> -»• 0,
(2) into equations (1) and are and V\ sin ip — Ui cos (j> —• 1 as r —*• °°; (11a)
du u 1 du dw
— +- + + — =0 (3a)
dr r r d<t> dz C/i = 0 at r = 1 (116)
du du v du v2 dp , ,—
U— +W — + = - — + 0(lA/Gr) (36) For an infinite cylinder, the inviscid flow does not depend on the
dr dr r d<j> r dr z -coordinate; its solution is well known. The velocity and the pressure
dv du v dv uv gradient along the surface of the cylinder (r = 1) are
u hio 1 1 : =i^+0(l/V&) (3c)
dr dz r d<t> r r d0
Vi = 2 sin (f>, (12a)
dw dw v dw dp , , ,—,
u — +w — + = - — +0(l/VGr) (3d) Wi = 0, (126)
dr dz r d<j> dz
Since it is isothermal outside of the thermal boundary layer, the en- and
ergy equation can be ignored. The associated boundary conditions dP2
are = 4 sin <>
/ cos ij> (12c)
dcj>
as r -* °°:tu = v cos 0 + u sin 0 —>• 0, and v sin 0 — u cos < When Re is not very small, the forced-convection boundary layer
(4a) separates along the aft body of the cylinder. The flow separation can
affect the inviscid flow; therefore, the solution (12) is valid only along
a t r = l:u = 0 (slip-condition). (46) the forebody of the cylinder for a forced-convection flow. We will,
Boundary conditions (4a) and (46) suggest that the solution of however, demonstrate later that the boundary layer does not separate
equations (3) can be expanded in a series of e, which is the inverse of in the region where the free convection is dominant.
the square root of the Richardson number. Knowing that the value The solution (12) shows that forced convection affects the free-
of t is small, the solution of equations (3) is sought in the following convection boundary layer in two ways. The viscous shear, equation
form. (12a) drives the boundary layer along the if> direction. Its order of
magnitude is e, and is proportional to sin <t>. The pressure gradient
u = U0+eU1 + e2U2 + e3C/3 + . . . along the 0 direction, equation (12c), has a smaller effect than that
w = W0 + eWt + e2W2 + esW3 + ... of the viscous shear. Its order of magnitude is e2 and is proportional
(5) to sin 20. This feature probably shares by the cavity flows driven by
v = V0+fVi + €2y2+e3V3 + ...
the surface forces or by the body forces near the boundary.
p = P 0 + ePi + e 2 P 2 + « 3 P 3 + • • • The equations of e3 with equation (lOd) can be solved to obtain U2,
The governing equations of the components in equation (5) are W2, V2 and P3. However, the homogeneous boundary conditions in-
obtained by substituting equation (5) into equations (3) and then sure a zero solution, i.e.,
collecting the terms of equal t order. The equations and their solution
are given below. U2 = V2 = W2 •0 (13)

v/=_dPo Boundary Layer. The boundary-layer equations can be obtained

(e1): UQ 1- Wo 1 (6a) by stretching the coordinate normal to the surface of the cylinder. The
dr dz r d<j> r dr
stretched coordinate is
IT0 dW°^u, 1- jy 0dW° , 1 VodWa . d P 0
U (66)
dr dz r d0 dz y : ( r - l ) - G r 1 / 4 and «(, = {/-Or 1 ' 4 (14)

(Jo + Wo h 1 (6c) and the rest of the dimensionless variables are identical to those de-
fined for the inviscid flow. Substitution of the dimensionless variables
dr dz r d<j> r r d<j>
into equations (1), after evaluating the pressure gradient from the
dU0
t | U0 | 1 dV 0 dW Q _
(6d) inviscid solution, gives
dr r r d<j> dz dui, dw dv
The solutions of equations (6) which satisfy the homogeneous — - + — + — = 0, (15a)
boundary conditions are zero, i.e., dy dz d 0
dv dv dv d 2 i;
Vo = Uo = Wo = 0 and Po = constant. (7) Ub l-u hu — e2 • (2 sin 20) + + 0(Gr-1<'4), (156)
dy dz d(j> dy 2
dPi dPi dPt 2
(f 1 ) 0 (8a) dw dw dw „ d w „,_, , „ , ,„_ ^
dr dz d<j> ub — + w — +v — = 9 + — 2- + 0(Gr- 1 / 4 ), (15c)
dy dz d0 dy
dUi lh dWi ldVi
(86) dS dd dd 1 d20 ,„ ,,,
dr r dz r d<j> ub — +w — +v — = + OtGr- 1 / 4 ). (15d)
dy dz d0 Pr dy 2
Equations (8a) and (7) indicate that
The momentum equation in the r direction shows that the pressure
Pi = 0 (9) gradient normal to the surface of the cylinder is of order of Gr~ 1/4 and
is negligible. The associated boundary conditions are
dP2
(e2): Ui f- Wi 1- (10a) ;
dr dz r d(j> r dr ' 0 and v = t • (2 sin 0) at z = 0. (16a)

S48 / VOL. 103, AUGUST 1981 Transactions of the ASME

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 ub = v = w = 0, and 6 = 1 at y = 0, (16b)
Therefore, the perturbation is valid for arbitrary value of e as long as
the value of ezx/2 is small. The upper limit ez1/2 for which the pertur-
w = 6 = 0, and u -»• c • (2 sin <j>) as y -»• <». (16c)
bation solution may be applied can be found by comparing the per-
Following the expansion procedure used in solving the inviscid flow, turbation solution with the downstream numerical solution which is
the solution of the boundary layer can be expressed as a series of e. discussed in [14].
In terms of the stream functions, they are Equations (18-20) with the boundary conditions (21) are solved
1 2
by numerical integration. The results are given and discussed in the
w = (42) ' • f0'(v) + (e) • 5 • (42) • /Y(t)) • (2 cos 0) next section.
+ (£2) • (42) 3 / 2 • \fac'(v) • (4 cos 2 0) + hs'Cn) • (4 sin 2 0)]
3 Results and Discussion
+ ... (17a)
Equation (17a) shows that the axial velocity can be estimated in
u = («) • A ' W • (2 sin 0) + (£2) • (42) 1 ' 2 • fi'it)) • (2 sin 20) + . . . terms of the stream functions fo, / i , / 3 c and / 3 s . The values of/ 0 and
A' are presented in Fig. 2 for Pr = 0.01,1 and 10, respectively. This
(17b)
solution is known as the Schmidt-Beckmann-Pohlhausen solution,
ub = (42)-i/" • (vfo' - 3/0) + (e) • (4z)!« • (ijA' - / x + / 2 ) • (2 cos 0) and represents the free convection along a heated vertical cylinder
+ (<E2) • (42)3/* • \(r)hc' - lhc + f2) • (4 cos 2 0) (or plate). The magnitude of the free convection is inversely propor-
tional to the Prandtl number. Functions / i and f{, given in Fig. 3,
+ ( Vfas' ' 7/ 3s + A) • (4 sin 2 0)] + . . . (17c) represent the first-order effect of the forced convection on the free
2
8 = GO(TJ) + (e) • (4 Z )i/ • Gi(ij) • (2 cos 0) convection. Values o f / i and /V are negative. This is a consequence
+ (f2) • (42) • [G2c(v) • (4 cos 2 0) + G 2s • (4 sin 2 0)] + . . . , (17d) of the forced convection suppressing the free-convection boundary
layer in the neighborhood of the forward stagnation point (0 = 0), and
where thus thinning the boundary layer. The amplitudes of/i and f{ are also
inversely proportional to the Prandtl number. Values of / 3 c , f3c', f3s
V = y/(4z)V4 (lie)
and /3s', given respectively in Figs. 4 and 5, are the effects of the forced
and the prime denotes the derivative with respect to i). The governing convection driven by the azimuthal pressure gradient. The amplitude
equations of the stream functions, /'s, and the temperature functions, of these functions increase with Prandtl number. The functional
G's, are obtained by substituting equations (17) into equations (15) shapes of/3c and / 3 s are similar, but are different in magnitudes. This
and collecting the terms of equal t. They are shows that the second-order solution of w is not proportional to cos
20.
(€°):/o"' + 3/o/o"-2(/o') 2 = - G 0 , (18a)
The azimuthal velocity, u, can be estimated in terms of/2 and fi as
G 0 " + 3 - P r - / 0 - G o ' = 0, (18b) shown in equation (17b). Values for f2, / 2 ', fi and /Y are shown in Figs.
6 and 7, respectively. / 2 and f{ represent the forced convection driven
(e1): A ' " + 3/ 0 /i" - 6/o'/i' + 5/ 0 "A = -Gi - A"A, (19a) by the viscous force; fi and /Y are due to the azimuthal pressure gra-
dient. Function / / shows a wave structure which pushes the fluid
A'" + 3/oA" = 0, (19b)

0.6
G x " + 3 / o d ' - 2/o'Gi = - G 0 ' • (6/i + f2) (19c) / • \
Pr
fo
(e 2 ):/ 4 '" + 3/o/ 4 " - 2 / 0 ' / 4 -/ \ fo' - 0.5

= - 1 - 5vf2"fi' + (vfi - A t A) • A " + (A') 2

(20a) / \
O.OI-A ^^ 0.4
A c " ' + 3/oA c " - 8/o'/3c' + 7/o"Ac x
-; \
= - G 2 c - /o" • A - 5 - (4r/A' + / i ) - / i "
1 s
s
+ 5./i"-/2+100-(A')2 (20b) ^ \ - 0.3
1 1 X
s
fas'" + 3/o/ 3 s " - 8/o'A s ' + 7/ 0 "/ 3 s = - G 2 s - 5A'A' - h"h (20c) 1
r
> \ /
~ 0.2
— • G 2 c " + 3/oG 2c ' — 4/o'G 2c
Pr

= - G o ' • (A + 7 • / 3 c ) - (4T/A' + A - A) • Gi' + 10 • A' • Gi (20d)

KX1
l yC \ 1
"-^ "**•* O.I

10
Pr G 2 s " + 3 • /o • G 2 s ' - 4/ 0 'G 2 s = - G o ' • (A + 7 • / 3 s ) - U • Gi-
Z^Sz:
I***-—r •"••ihu,, — 1 1 1
10 12
1
0.0

•v
(20e)
Fig. 2 Stream functions f0 and f0'
The transformed boundary condition associated with equations
(18-20) are -2.0, ,-0.2

atr, = 0:/ 0 = / 0 ' = / i = / i ' = A = A' = Ac = Ac' = A s = A /

= A = A' = Gx = G 2c = G 2s = 0, and G 0 = 1; (21a)

as n — <*>: A', Ac', fa/, A', G 0 , G 2c , G 2s - 0, and A ' - 1. (21b)

The solution of equations (18) is the free-convection flow, and is
known as the Schmidt-Beckmann-Pohlhausen solution. In the limit
that the Reynolds number approaches zero, equations (17) show that
the flow is free convection.
Equations (17) show that the effects of the forced convection are 0 2 4 6 10 12 14
proportional to &z1/2. This indicates that e is not the actual parameter
for the perturbation solution; instead, the parameter should be ez1/2. Fig. 3 Stream function f, and f-,'

Journal of Heat Transfer AUGUST 1981, VOL. 103 / 549

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 •1.6
_^I0 '3C

f3c
-I 2
/to \ JO

If \ ^ ^ -08-S
I w^ \ ^ -^obi
// //\.

10 12 14

Fig. 4 Stream functions t3c and f3c'

'3S Fig. 7 Stream functions f4 and U

- »3S
10 | -*• V
-—
10 12 14 16 18 20,

/ / 0.01 „-
/ / ^--
1.0 0.2
~ aoi
0.0
2 4 6 8 10 12 14
»-f)

Fig. 5 Stream functions /3s and /3s'

Fig. 8 Temperature functions G0 and G-,

14 12 10
0.00
'
'"^N
' ^Ky/l
G2S *• .
0.02
^-*'*~-\
"•j G2C ,
10 12 0.08
0.01, y /
/ V/l.o
Fig. 6 Stream functions f2 and fj 0.4 / 0.12

along the positive direction close to the cylinder surface and retards
the fluid close to the free stream.
0,2

0.0
zx^
fX° v
/
'OOI
0.16

0.20
Function Go, given in Fig. 8, is the temperature distribution due 10 12 14
to the free convection. The first-order temperature distribution, Gj., -*-v
due to the forced convection is also shown in Fig. 8 for Pr = 0.01,1, Fig. 9 Temperature functions G2c and G2s
and 10, respectively. The second-order temperature distribution due
to the azimuthal pressure gradient is contributed to two parts: G2,. <<t> < 180 deg. The first order azimuthal shear stress, Tr^,, is propor-
represents the temperature distribution proportional to cos 2 <> / and tional to sin 0 and the second order effect is proportional to sin2 0. The
G2, is proportional to sin 20, as shown in Fig. 9. second-order effect is due to the azimuthal pressure gradient whose
S h e a r Stress and Nusselt N u m b e r . The shear stresses along z value is positive for 0 deg < 0 < 90 deg and is negative for 90 deg <
and 0 directions can be estimated from equations (17a) and (17b), 0 < 180 deg. This indicates that the forced-convection separation for
respectively. They are 90 deg < 0 < 180 deg can occur when the value of e2z becomes large,
as expected.
: G,Tzlf. (42)1/4. {^//(o) + i0((4z)^2 • fx"{0) • cos 0 Nusselt number with respect to a and (Tw - 7\») can be derived
pv2/a2
from equation (17rf) and is
+ 4e 2 (4z). [/ 3c "(0) • cos 2 0 + f 3s "(0) • sin 2 0] + . . .) (22a)
Nu = -[Gr/(4z)] 1 '' 4 • (Go'(0) + 2e(4z) 1/2 G 1 / (0) • cos 0
-ll±-2 = Re . Grirt - (4z )-W • [2 • / 2 "(0) • sin 0 + 4e2(4z) • [G 2c '(0) • cos 2 0 + G 2s '(0) • sin 2 0] + . . .) (23)
pi> /a 2
With Table 1, equation (23) shows that the first-order forced-
+ 2€(4z) 1 / 2 ./ r 4 "(O)-sin 2 0 + . . . ] (226)
convection effect increases the heat transfer rate for 0 < 0 < 90 deg
Coefficients of / " s are listed in Table 1. The last two terms of and decreases it for 90 deg < 0 < 180 deg. The second-order forced-
equation (22a) represent the effects of the azimuthal pressure gra- convection effects, due to the azimuthal pressure gradient, constantly
dient. Except for Pr = 0.01, the values of / V t O ) and f3c"(0) are neg- decrease the heat transfer rate. Since Go'(0) represents the free-
ative. The azimuthal pressure gradient has the effect of decreasing convection heat transfer, the heat transfer rate due to the free and
the value of rrz. The second term on the right-hand side of equation forced convection is always smaller than that due to the free convec-
(22a) is contributed due to the azimuthal inviscid velocity, which tion alone for 0 > 90 deg. This type of Nusselt-number distribution
increases the value of Trz for 0 < 0 < 90 deg and decreases it for 90 deg has been demonstrated by the numerical solution for Pr = 10 [3].

550 / VOL. 103, AUGUST 1981 Transactions of the ASME

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 Table 1 Coefficients for shear stress and Nusselt number
Pr
Functions 0.01 0.1 0.733 1.0 10
/o"(0) 0.987543 0.858704 0.674175 0.642178 0.419160
A"(0) -0.001392 -0.036508 •0.059129 -0.059208 -0.044931
/3c"(0) 0.030153 -0.029042 •0.130256 -0.139573 -0.202793
/s."(0) -0.014563 -0.052309 •0.074680 -0.076463 -0.133325
/2"(0) 0.768148 0.705665 0.591001 0.567147 0.377088
A"(0) 0.480285 0.483977 0.547409 0.575214 1.00187
Go'(0) -0.080623 -0.230175 •0.507904 -0.567134 -1.16926
Gi'(O) -0.016694 -0.040675 •0.063274 -0.065802 -0.085290
G2c'(0) 0.079404 0.180094 0.224638 0.210649 0.246207
G 2 /(0) 0.018833 0.044503 0.063753 0.059811 0.051344

However, the numerical solution for Pr = 0.7 indicates that the pp. 19-24.
Nusselt number along 0 = 90 deg is larger than that due to the free 4 Schlicting, H., Boundary-Layer Theory, McGraw-Hill, New York, 1968,
convection which is inconsistent with that of Pr = 10. This is due to p. 300.
5 Lin, F. N., and Chao, B. T., "Laminar Free Convection over Two-Di-
the fact that the numerical solution was not started at 2 = 0 ; instead, mensional and Axizymmetric Bodies of Arbitrary Contour," ASME JOURNAL
it was started at a small value of z. The error induced by mistreating OF HEAT TRANSFER, Vol. 96,1974, pp. 435-442.
the initial condition propagates along the cylinder. The error might 6 Joshi, N. D., and Sukhatme, S. P., "An Analysis of Combined Free and
be gradually damped out as the value of z increases; however, the error Forced Convection Heat Transfer from a Horizontal Circular Cylinder to a
Transverse Flow," ASME JOURNAL OF HEAT TRANSFER, Vol. 93, No. 4,1971,
can be significant in the neighborhood of the thermal leading edge. pp. 441-448.
Equations (17) show that the length scale further upward along the 7 Young, R. J., and Yang, K. T., "Effect of Small Cross Flow and Surface
cylinder is a -Gr/Re 2 . It can be shown that the solution in the region Temperature Variation on Laminar Free Convection Along a Vertical Plate,"
z ~ 0(a • Gr/Re 2 ) is valid at z = 0, and the similarity solution of the JOURNAL OF APPLIED MECHANICS, Vol. 30,1963, pp. 252-256.
8 Eichhorn, R., and Hasan, M. M., "Mixed Convection about a Vertical
free-convection boundary layer is the initial condition for this prob- Surface in Cross-Flow: A Similarity Solution," ASME Paper No. 80-H-129,
lem. However, z = 0 is a singular point, and a stable numerical solution 1980.
can only be found by transforming the governing equations to the 9 Plumb, O. A., "The Effect of Crossflow on Natural Convection from
parabolic coordinates which remove the singularity at z = 0. The Vertical Heated Surfaces," ASME Paper No. 80-HT-71.
computational procedure is very similar to the case of free-forced 10 Yao, L. S., "Free-Forced Convection in the Entry Region of a Heated
Straight Pipe," ASME JOURNAL OF HEAT TRANSFER, Vol. 100, 1978, pp.
convection on a longitudinal and horizontal cylinder [13]. 212-219.
11 Yao, L. S., "Variable Viscosity Effect on the Laminar Water Boundary
Layer on Heated Cones," Journal of Applied Mechanics, Vol. 45, 1978, pp.
References 481-486.
1 Clausing, A. M., Clark, G. L., and Mueller, M. H., Modeling, Simulating, 12 Yao, L. S., and Catton, I., "Buoyancy Cross-Flow Effects in the Boundary
Testing, and Measurements for Solar Energy System, edited by Nash, J. M,, Layer on a Heated Longitudinal Horizontal Cylinder," ASME JOURNAL OF
et al., ASME, New York, 1978, p. 73. HEAT TRANSFER, Vol. 99,1977, pp. 122-124.
2 Oosthuizen, P. H., and Leung, R. K., "Combined Convective Heat 13 Yao, L. S:, Catton, I., and McDonough, J. M., "Free-Forced Convection
Transfer from Vertical Cylinders in a Horizontal Flow," Paper No. 78-WA/ along a Horizontal Cylinder," Number of Heat Transfer, Vol. 1, 1978, pp.
HT-45, ASME Winter Annual Meeting, San Francisco, Calif., 1978. 255-266.
3 Oosthuizen, P. H., "Numerical Study of Combined Convective Heat 14 Yao, L. S., "Buoyancy Effects on a Boundary Layer Along an Infinite
Transfer from a Vertical Cylinder in a Horizontal Flow," Proceedings of the Cylinder with a Step Change of Surface Temperature," ASME Paper No. 80
6th International Heat Transfer Conference, Toronto, Canada, MC-4,1978, WA/HT-25,1980.

Journal of Heat Transfer AUGUST 1981, VOL. 103 / 551

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 Deieloping Laminar Free
M. A. 1. El-Shaarawi Confection in an Open Ended
A. Sarhan Vertical Annulus with a Rotating
Asst. Lecturer.

Mechanical Engineering Department,

Inner Cylinder
Alazhar University,
Nasr City, Cairo
Egypt
A finite-difference scheme is developed for solving the boundary layer equations govern-
ing the laminar free convection flow in open ended vertical concentric annuli with rotat-
ing inner walls. Numerical results are presented for a fluid of Pr = 0.7 in an annulus of
radius ratio 0.5 under the thermal boundary conditions of one wall being isothermal and
the opposite wall adiabatic. Comparing the present results for the development of <5# (tan-
gential boundary layer displacement thickness) with those corresponding results of
forced flows shows that heating the inner cylinder has always stabilizing effects while
heating the outer cylinder has either destabilizing or stabilizing effects.

Introduction These investigations showed that the presence of a superimposed low

Free convection heat transfer and its associated fluid motion in an axial flow has a stabilizing effect on the onset of Taylor vortices.
open ended vertical concentric annulus whose inner cylinder rotates In reality, the tangential velocity profile for an axial viscous flow
is an idealization of many frequently occurring engineering situations. in an annulus with a rotating inner cylinder is a function of the axial
In vertical electric machines, the heat resulting from the iron and distance, not withstanding the superimposed axial velocity profile
copper losses is conducted to the rotor surface where it is transferred [9-12]. Therefore, there is an axial length required for the tangential
to the air in the annular gap between the rotor and the stator. The velocity profile to approach the well known fully developed Couette
transfer of heat by natural convection and radiation is always a factor profile. Astill [9-10] showed, by means of a visualization technique
in the cooling of an electric machine and may be the sole means for and hot wire measurements, that transition in a tangentially devel-
cooling some small types of electric apparatus. Therefore, the design oping flow starts near the rotating inner cylinder as oscillating waves
of cooling systems for such types of electric machines requires a within the tangential boundary layer. He also developed the unique
knowledge of the hydrodynamic behavior of the free convection flow stability criterion available to date for the onset of instability in
to limit the rotor temperature to less than the maximum permitted tangentially developing flows. His empirical stability criterion, which
value. Beside other applications in chemical mixing processes, fiber has recently been confirmed by Payne and Martin [13], is a Taylor
coating applications and drying machinery, there are possibilities of number based on the tangential boundary layer displacement
future applications for compact rotary heat exchangers and com- thickness (&$) rather than the gap width.
bustion chambers. According to AstilPs stability criterion, knowledge of the axial
These possible future applications depend on the fact that when growth of 5# for tangentially developing flows is essential to locate the
an isothermal viscous fluid is contained in an annular duct with a axial position of the point of origin of instability. Therefore, one of
rotating inner cylinder, there is a possibility for the existing Couette the main aims of the present investigation is to determine the devel-
flow to become hydrodynamically unstable and Taylor vortices [1], opment of the tangential boundary layer displacement thickness for
a secondary laminar flow regime, may appear. This instability occurs tangentially developing flows created by buoyancy effects only. Also,
when the rotational speed of the inner cylinder becomes sufficiently the lack of either theoretical or experimental investigations concerning
large such that the ratio of destabilizing centrifugal forces to re- the development of laminar free convection in open ended vertical
straining viscous forces reaches a critical value at which breakdown concentric annuli, with either stationary or rotating inner walls,
of the original Couette laminar regime occurs, and vortices appear. represents another motive for the present work. Two thermal
These vortices, occuring within the fluid on a macroscopic scale, will boundary conditions are considered in this investigation; namely, case
increase the efficiency of convective heat transfer mechanism if one / in which the inner wall is isothermal while the outer wall is adiabatic,
or both of the annulus boundaries were heated. and case 0 in which the outer wall is isothermal while the inner wall
However, heating one or both of the annulus boundaries creates is adiabatic.
a radial temperature gradient which affects the stability of a Couette
flow between two rotating concentric cylinders. In the absence of axial Governing Equations
fluid flow, neglecting gravitational body forces (i.e., neglecting free The fluid is assumed to have constant physical properties but obeys
convection effects), and assuming constant physical fluid properties the Boussinesq approximation, according to which its density is
except the density which is allowed to vary according to the Boussi- constant except in the buoyancy term of the vertical momentum
nesq approximation in only the centrifugal acceleration term, Yih [2], equation. The flow is steady, axisymmetric and without internal heat
Becker and Kaye [3] and Walowit, et al. [4] have shown that positive generation. Viscous dissipation and axial diffusion of heat are ne-
and negative radial temperature gradients are destabilizing and sta- glected. Further, applying the Prandtl boundary layer assumptions
bilizing, respectively. and checking the order of magnitude of each term [10], the equations
The presence of an axial velocity component has also significant governing the free convection fluid motion and heat transfer in the
effects on the hydrodynamic stability of a Couette flow. Most of the entry region of a vertical concentric annulus with a rotating inner
available investigations into this problem are limited to the case of cylinder reduce to the following equations
the onset of hydrodynamic instability in fully developed Couette flows
o(rv) d(ru)
with a superimposed steady or fully developed axial flow (e.g., [5-8]). •+ -1-^ = 0 (1)
dr dz
Contributed by the Heat Transfer Division for publication in the JOURNAL w2 dp
OP HEAT TRANSFER. Manuscript received by the Heat Transfer Division, July (2)
7,1980. r or

552 / VOL.. 103, AUGUST 1981 Transactions of the ASME

Copyright © 1981 by ASME
Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 dw dw\ 1 d for case / and T = 1 for case 0
Po\v — + u — • V -— (™) (3)
dr r dr for Z = L and R = 1, P = 0 (14)
dr dz]

I du
or
du\
u— + u— = •
dz)
dp
dz
ot
-Pog[l

ot
/3(t-t0)]+^Mr^
r dr \ or
(4) It is noteworthy that the initial pressure at the duct entrance in the
above conditions (14) is obtained by applying Bernoulli's equation,
i.e., taking into consideration the acceleration of the fluid to t h e
poCp \v Yu (5)
dr oz, r or\ dr} channel inlet.
Defining a pressure defect p ' as Using the boundary conditions (14), equation (9) can be written
in the following integral form
p' =p-Ps (6)
Q = ( i • _ N2)U0 =2 f RUdR (15)
equations (2) and (4) can be written JN
w dp' The tangential boundary layer displacement thickness, the heat
Pa— = - (7)
r dr absorbed by the fluid from entrance up to a particular elevation, and
the mixing cup temperature can respectively be given, in dimen-
I du
Po [v — + " —
\ dr
du!

dz
dp'
dz
+ PagPtt •t0)+-
u. d
—
r or
(8)

Introducing the dimensionless parameters given in the Nomenclature,

(-a sionless forms, by the following equations

8t*= f WdR/(l-N) (16)

equations (1,3,5,7) and (8) can be replaced by the following dimen- JN

sionless forms
H= CuTRdR (17)
dV V dU JN
—+-+—=0 (9)
dR R dZ
= ^UTRdRI f URdR (18)
8(1 - iV)6 Gr 2 dP JN JN
(10)
R 1 + JV Ta dfl Also, the local Nusselt number and the average Nusselt number
dW dW 2
dW IdW over the annulus height can respectively be given by the following two
W_
V +U = (H) equations
dR dZ dR2 R dR R2
2 dT
dU at/ dP T d // ____^
1 dU Nu = 2(1- Tm)
V—+ U— = (12) •N) —
dR id- (19)
dR dZ N) +dR R dR 4 2

dZ 16(1 -2 Nu Ra*HD/Dw (20)

dT rrdT 1 dT Id
y.
dR
+U =_ +
dZ P r \dfl 2 R dR,
(13)

These five coupled equations (9-13) are subject to the following

3 The fully developed flow, which occurs if the channel is sufficiently
high and uniform temperature is achieved before the fluid reaches
boundary conditions: the top, provides an analytical check on the finite difference solution
to be obtained. Noting that for fully developed flow, T = 1, d/dZ =
forZ = OandiV <R < 1, V = W = T •• 0, U = t/o 0 and V = 0, the inertia terms on the left hand sides of equations (11)
a n d P = Po = - t / o 2 / 2 and (12) vanish and the resulting reduced two equations give, re-
spectively, the following fully developed tangential (Couette) and axial
for Z > 0 and R = N, U = V = 0, W = 1 and T = 1 velocity profiles
dT
for case / or — = 0 for case O W / a - T 2- ^ ( r - * ) (21)
dR l-N \fi I
2 2
dT l-R -q-N )(\nR/lnN)
for Z > 0 and R = 1, U = V = W = 0 and ufd> (22)
dfl 64(1 - JV)4

•Nomenclature
Po = dimensionless pressure defect at annu-
lus entrance, Por24lpol2y2 Gr* 2
a = local heat transfer coefficient H = dimensionless heat absorbed from en- Pmo = dimensionless pressure defect at outer
based on area of heated surface, trance up to any particular elevation, hi wall, (pwo-Ps)r24lp0l2y2Gt*2
_ wpocplyGv*(ta -t0) q = volumetric flow rate,
l\ttn tw) H = dimensionless heat absorbed from en-
dr trance up to annulus exit, i.e., value of// at 2irru dr
a = average heat transfer coefficient over the z = l
annulus height, h/irDwl(tm — to) Hfd = fully developed value of H Q = dimensionless volumetric flow rate,
b = annular gap width {r^ — r{) k = thermal conductivity of fluid q/irly Gr*
cp = specific heat of fluid at constant pres- £ = height of annulus Qfd = fully developed value of Q
sure L = dimensionless height of annulus, 1/Gr* r = radial coordinate
D = hydraulic diameter of annulus (2b) N = annulus radius ratio, ri/r2 r\ = inner radius of annuius
Dw = diameter of the heated wall n = number of radial increments in the nu- r 2 = outer radius of annulus
g = gravitational body force per unit mass merical mesh network R = dimensionless radial coordinate, rlr<i
h = heat absorbed by fluid from p = pressure of fluid at any point t = fluid temperature at any point
entrance up to any particular p ' = pressure defect at any point, p — ps tm = mixing cup temperature over
elevation in the annulus, Po = pressure of fluid at annulus entrance any cross section,
PwO = pressure of fluid at the outer wall of

JT2

ri
2wcppQru(tm - t0) dr annulus
ps = hydrostatic pressure, —pogz J
T2

utr dr/ 1 ur dr
<^r2

h = heat absorbed by fluid from entrance up P = dimensionless pressure defect at any ri Jr\
to annulus exit, i.e., value of h at z = I point, p'r2 4 /po' z T 2 Gr* 2 (Continued on next page)

Journal of Heat Transfer AUGUST 1981, VOL. 103 / 553

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 Also, since for fully developed flow, T = 1, equations (15) and (17) 4
show that the limiting values of the dimensionless volumetric flow
rate and heat absorbed (Q and H) are equal and given by 1+1 n rm

Qfd=Hfd = 2 C RUfddR Z=L (channel

exit)
UN
1-/V2 i - m J+2
1 + AT2 + (23)
128(1 - iV) 4 IniV )+l

j Z=(j-1)*Z
N u m e r c i a l A n a l y s i s a n d M e t h o d of S o l u t i o n
Considering the mesh network of Fig. 1, and by an indirect exten- j-1-
sion of the work of Bodoia and Osterle [14], equations (9-14) can be
J-2
written in the following finite difference forms
Vj+ij+i ~ Vjj+i Vj+ij+i + Vjj+i
/ /
AR 2[N+(i-$AR] 3

2 z=o
(channel inlet)
2AZ
R=0 R=N ^ V
R=N+(i+1)«R Nr R=1 R
WjjWj,,j-n sq-N^G^Pij+i-Pi-u+i (inner waU) (outer watt)
(25)
N+(i- l)AR 1 + N Ta AR Fig. 1 Mesh network for finite difference representation
y. Wi+U+1 + Wi+lj ~ Wj-ij - Wj-iJ+1 | Wjj+i - WjJ
iJ ij
4AR AZ
Q = 2Aii £ [ 7 ; j [ i V - l - ( i - l ) A R ] (29)
Wi+1J+1 + Wi+1J - 2WiJ+1 - 2Wj,j + Wj-u+i + Wj-u
2(AR)2 It should be noted that the above finite difference equations are
1 linearized by assuming that, where the product of two unknowns oc-
curs, one of them is given approximately by its value at the previous
N+(i- 1)AR
axial step. The variables with subscript; + 1 represent the unknowns
Wi+1J+1 + Wj+u - WJ-IJ - Wt-u+1 WiJ+1 + Wjj and those with subscript j are knowns. Also, applying the available
AAR 2[N + (i - 1)AR)2 numerical stability theories summerized in [15] shows that the finite
difference equations (24-28) are consistent representations of the
ViJ + U i boundary layer equations (9-13) and are stable for all mesh sizes as
2A^ - AZ AZ
long as the downstream axial velocity is non-negative. Moreover, the
chosen finite difference approximations are not perfectly symmetrical
16(1 - AT)4 (Afl) 2 nor are they of the same form in all the equations. This is done so as
to insure stability of the numerical solution and to enable the equa-
1 t/j+ij+i ~ UJ-IJ+I , .
tions to be solved in the following manner.
N+(i-l)AR 2Afl
For an annulus of a given N, the numerical solution of the foregoing
finite difference equations is obtained by first selecting values of Pr,
K,V
2Afl +UiJ
AZ Q and Gr 2 /Ta. Using the value of Q and applying (15) we get the inlet
velocity Uo and hence the inlet pressure Po is obtained. Then, starting
at the 7 = 1 column (annulus entrance cross section) and applying (26)
' Pr \ (Afl) 2 (AiJ) 2 to points i = 2, 3 , . . . , n and (28) to points i = 2, 3 , . . . , n + 1, we get
+ 1 TWi-T,-!,^ (2n — 1) simultaneous linear equations which when solved (e.g., by
iV+(i-l)AB 2Afl / Thomas' method [16]) give the values of the (2ra — 1) unknowns (1^2,2,

-Nomenclature Continued-
V = dimensionless radial velocity compo-
tm = mixing cup temperature over the exit nent, 11^2/7 displacement thickness,
cross section, i.e., value of tm at z = I w = tangential velocity component at any
to = fluid temperature at annulus entrance point w dr/Un
tw = isothermal wall temperature W = dimensionless tangential velocity com-
T = dimensionless temperature at any point, ponent, w/Qri &g* = dimensionless tangential boundary
(t - t0)/(tw - to) Wfd = dimensionless fully developed tan- layer displacement thickness, dg/b
Tm = dimensionless mixing cup temperature, gential velocity component Gr = Grashof number, gfi(tw — to)D3/y2
(tm ~ to)/(tm — to) z = axial coordinate Gr* = modified Grashof number, D Gr//
u = axial velocity component at any point Z = dimensionless axial coordinate, z/£ Gr* Nu = local Nusselt number, aD/k
Ufd = fully developed axial velocity compo- /3 = volumetric coefficient of thermal ex- Nu = average Nusselt number, aD/k
nent pansion Nu; = average Nusselt number for case / , (1
uo = entrance axial velocity, 6 = dimensionless axial coordinate, 2z(l — _rN)HRa*/N
N)/r2 Re Nuo = average Nusselt number for case 0 , (1
2irru dr/ir(r22 — r\2) p = fluid density -N)7tRa*
Po = fluid density at inlet fluid tempera- Pr = Prandtl number, ncp/k
U = dimensionless axial velocity at any point, ture Ra = Rayleigh number, Gr • Pr
ur-^lly Gr* ix = dynamic viscosity of fluid Ra* = modified Rayleigh number, Gr* • Pr
Ufd = dimensionless fully developed axial 7 = kinematic viscosity of fluid, jt/po Re = axial Reynolds number, uoD/y
velocity component, u/dr 2 2 //7 Gr* fi = angular velocity of inner cylinder Ta = Taylor number, 2ft 2 ri 2 6 3 /7 2 (ri + r 2 )
v = radial velocity component at any point be = tangential boundary layer Ta* = modified Taylor number, Ta (b/l)2

554 / VOL. 103, AUGUST 1981 Transactions of the ASME

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 W3,2 Wn,2, T2,2, T3i2, . . . , T„+i,2) at all points of the second veloped values, it was impossible to compute a value of Z at which the
column. Now, applying (25) with i = 2, 3 n + 1 and (27) with i pressure P returns back to zero. Therefore, the present case was solved
= 2, 3 , . . . , n and (29) to the entire column we obtain 2ra equations for values of Q ranged from 0.001 to 0.015, at various values of the
which when solved (e.g., by a special form of Gauss-Jordan elimination parameter Gr 2 /Ta.
scheme [15]) give the values of the In unknowns (t/2,2, U•&,?., • • •, Un,2, On the other hand, more than 80 computer runs were made but it
Pi,2, Pi,2 Pn+1,2) at all points of the second column. Using the has been found that, at sufficiently low values of Gr 2 /Ta, a numerical
computed values of U's and applying (24) we get the values of V s at solution, for a given Q, is available up to a certain axial distance from
all points of the second column. Repeating this procedure, we advance the annulus entrance afterwhich a divergence from the laminar so-
column by column along the annulus until the pressure P at the outer lution, similar to that reported in [18], occurs. For such unsuccessful
wall of the annulus (Pn+ij) returns to zero. This cross section (at computer runs the dimensionless pressure defect (P) failed to return
which Pn+ij = 0) establishes the unknown dimensionless heated back to zero and hence the determination of the annulus dimen-
length L, which, by definition, is the reciprocal of the unknown sionless height (L) was impossible. Also, the point at which the solu-
modified Grashof number. Hence, both the dimensionless heat ab- tion starts to diverge, in such unsuccessful runs, always moves toward
sorbed by the fluid from entrance up to the channel exist (H) and the the entrance as the value of Gr 2 /Ta further decreases. Moreover, as
average Nusselt number (Nu) can be calculated from equations (17) the value of Q decreases, the lowest value of Gr 2 /Ta, for which a
and (20), respectively. laminar solution could be obtained over the entire annulus height,
It may be worth mentioning that, in practice, for a confined free increases. Taking into consideration the relationship which links
convection flow, such as that given under consideration, the channel together the parameters Gr 2 /Ta, Q and Re 2 /Ta, the reader may refer
height £ is normally known (i.e., L is given) while the volumetric flow to [18] for a discussion on the phenomenon of a nonexistant laminar
rate q (hence Q) is unknown. In fact, the present model and method solution, for a given Q, at low values of Gr 2 /Ta. Also, Schlichting [19]
of solution are handling the problem in a reversed manner, i.e., ob- explained the physical reasons due which one "must not expect"
taining an unknown channel height for a given volumetric flow rate. laminar solutions to be available "for arbitrary large Qri/uo," i.e., for
Therefore, the last condition stated in equation (14), i.e., P = 0 at Z high spin, low buoyancy cases.
= L and R = 1, is not explicitly imposed on the solution, but contin- Apart from those nonexistant solutions at sufficiently low values
ually checked for satisfaction. However, taking into consideration that of Gr 2 /Ta, the following sections present samples of the most im-
its satisfaction establishes the unknown channel height, this condition portant results which have been obtained from the successful com-
could be regarded as being implicitly imposed. puter runs.
On the other hand, it is known that near the annulus entrance large Effect of the Parameters Q and Gr 2 /Ta on the Characteristics
gradients exist. For this reason, all the computer runs were made by of the Developing Flow. It is worth mentioning that choosing a
taking 40 radial increments and very small axial steps (AZ = 1 0 - 7 for value of Q « Qfd implies a short annulus of a large hydraulic diameter
values of Q «= 0.0025 and AZ = lO" 6 for values of Q > 0.0025) near the with a large temperature difference (tw — to). In other words, it implies
entrance. Further downstream only 20 radial increments were used a case of a large modified Grashof number (Gr*) and that the flow
and the axial step was increased several times as the flow moves away leaves the annulus top exit before reaching the fully developed flow
from the entrance. conditions. Thus, for a given value of Gr* 2 /Ta* (for N = 0.5, Gr* 2 /Ta*
= 4 Gr 2 /Ta), decreasing the value of Q implies also an increase in the
value of Taylor number.
Results and Conclusions
Figures 2(a) and 2(b) show, for Q = 0.005 and 0.014, respectively,
Comparing the dimensionless equations and boundary conditions the variations of the dimensionless outer wall pressure defect (Pwo)
(9-14) with those corresponding equations and boundary conditions and the dimensionless heat absorbed by the fluid (H) with the di-
governing the case of developing laminar forced flows [17-18], three mensionless axial distance (Z), for both thermal boundary conditions
main differences might be noted. Firstly, the energy and momentum (/) and (O), at various selected values of the parameter Gr 2 /Ta. It
equations are coupled in the present case. Secondly, the inlet velocity could be seen from these two figures that, for a given Q and for either
U(s and the inlet pressure Po are not predetermined initial conditions case / or case O, increasing the value of the parameter Gr 2 /Ta causes
independent of the channel height as in the case of forced flows. an increase in the length required for the pressure defect at the outer
Rather, each of them is dependent upon the channel height and the wall to return back to zero, i.e., an increase in the dimensionless an-
initial driving temperature difference (tw — to). Thirdly, the number nulus height (L). This means that, for a given mass flow rate, in-
of similarity parameters describing the present case exceeds by one creasing the rotational speed of the inner cylinder reduces the required
those corresponding similarity parameters describing the purely annulus height. However, provided that laminar solutions exist, it has
forced convection case. been found that the Q-L variation in the range of 10 3 =S Gr 2 /Ta < °>
For this last reason, it was possible, for a given Pr and a given N, could be considered, with a maximum deviation of less than 10 per-
to obtain solutions for those equations which govern the forced flow cent, equivalent to the Q-L relationship shown in Fig. 3 for the sta-
by varying only one parameter, viz Re 2 /Ta [17-18]. On the other hand, tionary walls case.
numerical solutions for the present case are only possible by varying Another conclusion to be drawn from Fig. 2(a) is that, the inner
the magnitudes of two parameters, namely Gr 2 /Ta and Q. However, cylinder rotation causes the heat absorbed by the fluid (H) to be
there are limitations, as will be clarified, on the magnitude of each of slightly increased for case / and slightly decreased for case O. At large
these two parameters. values of Q, such an effect is so slight that it is unremarkable in Fig.
Computations were carried out for an annulus of N = 0.5 with Pr 2(6). Therefore, the H-L variation for the stationary walls case, given
= 0.7. For this particular radius ratio, equation (23) yields the fol- in Fig. 3, could also be used, irrespective of the value of the parameter
lowing limiting fully developed value for both Q and H Gr 2 /Ta, for the case with rotating inner walls. These slight effects of
the core rotation on H are attributed, as may be seen from Fig. 4, to
Qfd = Hfd = 0.015748 (30)
the increase in the outer wall and the decrease in the inner wall ve-
Equation (30) merely states that it is impossible, in the purely locity boundary layer thicknesses [18].
laminar free convection regime and irrespective of the value of Gr 2 /Ta, Other unpresented results indicate that, for a given Q, the param-
to have a dimensionless volumetric flow rate (Q) greater than 0.015748 eter Gr 2 /Ta has very slight effects on the local Nusselt number, the
through an annulus of N = 0.5, otherwise there must be an external mixing cup temperature, the development of the tangential velocity
force aiding the buoyancy driving force. In other words, the dimen- component and the dimensionless tangential boundary layer dis-
sionless volumetric flow rate (Q) reaches an upper asymptotic value placement thickness. In general, the slight effects of the parameter
(Qfd) as the annulus height approaches infinity. A special computer Gr 2 /Ta on the characteristics of the developing free convection flow
run was made for Q = Qfd and Gr 2 /Ta = °° and it has been found that, are qualitatively similar to those effects of the parameter Re 2 /Ta on
even though both U and T approach asymptotically their fully de- the characteristics of developing forced flows [17-18].

Journal of Heat Transfer AUGUST 1981, VOL 103 / 555

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 lb -n-rM! 1 ,—i-r-T-rwr T r « | J-L

Q = 0.015 lUh-J^y^"^
- Cast (1) . ' /

12 r - Case (0) //f -

10 / /
*bn p '// -
X / 1/ /
O.
tt
6 - •

4
" °' // /
-•
2
„,-•- ,^> '
Fig. 2(a) Variations of the heat absorbed by fluid and the outer wall pressure 0 sSSSi , i . , i l • • ' » " i

defect with the axial distance at various values of Gr2/Ta, N = 0.5, 0 = 0.005, to"3 io" io J
io' id"1
Pr = 0.7 L=VOr=<U&rXl/D)

1 1 - 1 1 1 1 1 I 1 Fig. 3 Variations of dimensionless volumetric flow rate and dimensionless

H
H heat absorbed with the dimensionless annulus height, Pr = 0.7, N = 0.5,
J^-^ Gr2/Ta = <»
\ KT-oo
f//~zr-
- *ill/i'—..
a
*r/y~-^
v/
- Ilfl 10

— Case (O)
IK< \ \ \\ N^5 , —
- 8
-f'"~x\
v V* 2xltf/\\>V
6

4
2x1ffX\\. 10V
\\ 2

[ 1 1 i i i\ \ i • \ W i n
10 12 W 16
ZxlO 3

Fig. 2(6) Variations of the heat absorbed by fluid and the outer wall pressure
defect with the axial distance at various values of Gr2/Ta, N ~ 0.5, Q = 0.014, Fig. 4 Effect of the parameter GrVTa on the development of the axial velocity
Pr = 0.7 component, W = 0.5, Q = 0.0025, Pr = 0.7, Case /

Figure 5 present the developing tangential velocity profiles corre- fluid from regions close to the rotating inner wall to the region close
sponding to two chosen values of Q. For Q = 0.001, this figure shows to the heated outer wall and thus transporting tangential momentum
that the zero tangential velocity profile at the annulus entrance is from fluid near the inner rotating cylinder to the fluid far from it. On
developed at the exit into a profile which is similar to that about a the other hand, it can also be seen that, for this value of Q, the radial
single vertical cylinder rotating in a flow parallel to its axis. On the velocity component for case / mainly transfers fluid from regions close
other hand, for Q = 0.012, Fig. 5 shows that the tangential velocity to the stationary outer wall to the region close to the rotating inner
profile near the annulus exit tends to approach the fully developed wall. Therefore, the tangential momentum transport due to the radial
profile given by equation (21). velocity is bigger in case O than in case / and hence one can easily
The effect of the parameter Q on the development of the radial understand why the tangential velocity profiles, presented in Fig. 5
velocity component could be seen from Fig. 6. While, for high values for Q = 0.001, develop in case O quicker than in case /.
of Q, the radial velocity transfers fluid from regions close to the two Also, at high values of Q (e.g., Q = 0.015), the same previously
boundaries to the core fluid, it mainly transfers, for low values of Q, mentioned effects of the thermal boundary conditions on the devel-
fluid from the region close to the adiabatic wall to that region close oping radial and tangential velocity components are present, as could
to the heated wall. be seen from Figs. 5 and 6. However, such effects become decreasingly
Effect of Thermal Boundary Conditions on the Characteris- significant as the value of Q increases, i.e., as the value of Gr* de-
tics of the Developing Flow. In Fig. 7 a comparison is made, for creases.
a given value of Q and a given value of Gr2/Ta, between the pressure The dimensionless tangential boundary layer displacement
distribution across the annulus for case / and for case O. It is clear thickness is drawn in Fig. 8 against the dimensionless axial distance
from this figure that, for the same value of Z, the effect of inner cyl- ( for various values of the parameter Q. The dimensionless axial dis-
inder rotation on the pressure distribution across the annulus is more tance e, rather than Z, has been chosen in order that the forced flow
strong for case O than for case /. Actually, this is due to the fact that, results [17] could easily be compared with the present free convection
for a given Q, the dimensionless annulus height (L) for case O is results. It is clear from this figure that for case / the value of 5#* is
shorter than that for case /, i.e., the modified Grashof number for case lower, at the same value of e and for all values of Q, than its corre-
O is larger than that for case I. Hence, for a given Gr2/Ta, i.e., for a sponding value of the forced flow case. Also, from the same figure, for
given Gr*2/Ta*, case O has a higher value of Ta* than case /. a given annulus under thermal boundary conditions (/) and at the
In order to explain the effect of thermal boundary conditions on same value of 6, decreasing the value of Q, i.e., increasing the value
the development of the tangential velocity component, it should be of Gr* and consequently increasing the corresponding value of (tw —
noted that the tangential momentum is transported by two mecha- to), causes a decrease in the value of 89*. So, combining these results
nisms. Firstly, by the transport created by the radial velocity com- with AstilPs stability criterion for tangentially developing flows [9-10]
ponent and, secondly, by the molecular diffusion due to the viscosity leads to the conclusions that: (1) A purely laminar free convection flow
of the fluid. At low values of Q (e.g., Q = 0.001), it can be seen from under thermal boundary conditions (/) has, for all values of Q and at
Fig. 6 that the radial velocity component for case O mainly transfers the same value of e, a more stable tendency than a forced flow. (2) At

556 / VOL. 103, AUGUST 1981 Transactions of the ASME

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 the same value of e, this tendency for stability increases as the value
D.001 Q=0.012
\ Gr/Ta =10° GrVTado''
of Q decreases, i.e., as the value of (tw — to) and hence the corre-
- Cased) • Case(I)
sponding radial negative temperature gradient increase. These results
and conclusions are as might be expected since in case of a pure free
\ \ - CaselO)
convection under thermal boundary conditions (/) a negative radial
^\\ \ \ ^-^ temperature gradient and an axial flow are present; both have stabi-
30
\ N \ lizing effects as was mentioned in the introduction of the paper.
\N \ 20
On the other hand, in case of a pure free convection under thermal
boundary condition (O) a positive radial temperature gradient, which
15 has destabilizing effects, is present beside the axial flow, which is
stabilizing. Therefore, depending upon the relative magnitudes of the
stabilizing axial flow and the destabilizing positive radial temperature
10
gradient, i.e., depending upon the values of both Q (which is propor-
tional to Re/Gr) and Re (which is inversely proportional to t), it is
yv^ 5
expected in case O to have either stabilizing or destabilizing effects.
Again, the combination of Astill's stability criterion and the present
i\\> A predictions, shown in Pig. 8, for the development of 8$* for case O,
confirm these expectations. As could be seen from this figure, for a
3 given annulus under thermal boundary conditions (O) and at the same
! \
\ value of e, decreasing the value of Q causes an increase in the value
2 of 8e* i.e., an increase in the tendency of the flow for destabilization.
\V Also, for very low values of Q (e.g., Q = 0.001) the purely laminar free
\ ZX1C
6
=1 " convection flow under thermal boundary conditions (O) has a more
i unstable tendency than a forced flow, while for high values of Q (e.g.,
0 0.5 1 0.5 Q = 0.015) the reverse is true.
(r-ijWrj-r,) (r-r^-r,)
Finally, in order to check the adequacy of the present numerical
Fig. 5 Effect of the thermal boundary conditions and the parameter 0 on
the developing tangential velocity profiles, N = 0.5, Pr = 0.7 predictions against the present fully developed analytical solution
and also against the available numerical and analytical results for
parallel plate channels and circular tubes [14, 20, 21] and [22], the
practically important correlation between the average Nusselt number
' 2 5 and the modified Rayleigh number is drawn in Fig. 9. It is noteworthy
GU0.015 , Gr/Ta=2.5*10 Q=0.001 , ---*. Gr^TfclO6
1 5 1 5 that this figure could be used for any value of Gr 2 /Ta, provided that
/ \
<o/ the flow remains in the purely laminar free convection regime. This
%"/ \ is because it has been found that, for a given value of Q, the parameter
10 10 - 4' ^"- \
Gr 2 /Ta has slight negligible effects on Nu. From this figure, it is clear
/' / JK \
\ that at low values of Ra*, the two curves of the present results for case
5 5 x \ \ J and case O approach respectively the lines A and A which are plots
/// ,-2°-. X \ \ * of equation (20) at full development. Also, at high values of Ra* these
x^x\A two curves approach, respectively, lines B and B which represent the
0 0 4£t
\$S>>^-30 limiting case of an infinite hydraulic diameter (D) and thus the \
\AX\~-L5- power law of heated vertical plates or ciruclar tubes applies.
-5
\K y - -5 Sy / On the other hand, the present results for case O are in good
\ \ i ^ agreement at low values of Ra* with the results of [22] for circular
-10 Casetir 10 \—§^ tubes, but at high values of Ra* the results of [22] are higher by about
Case(If
CaselO) CaselO) 8 percent than the present results. Also, the present results for case
I I I ! l 1 / are clearly different, at low values of Ra*, from the results of [21]
02 04 06, 08 02 04 06 08
<r-r, )/(r2-r,) (r-r^/frj-f,) for parallel plate channels. However, at high values of Ra*, the present
Fig. 6 Effect of the parameter Q and thermal boundary conditions on the results for case / are, in fact, lower by about 2 percent than the results
developing radial velocity component, W = 0.5, Pr = 0.7 of [14] and [20].

s' ^-Forced Flow(Q=w) ,---"

0.3 z
— ,' \^ _ — —"
%

0.2 -

0,1 - ff^' — Cased)

Mf Case (0)

i 1 , 1 i l . 1 .

Fig. 7 Effect of thermal boundary conditions on the pressure distribution Fig. 8 Dlmensionless tangential boundary layer displacement thickness
across an annulus of W = 0.5, Pr = 0.7, O = 0.0025, Gr2/Ta = 105 against dimensionless axial distance, N = 0.5, Pr = 0.7

Journal of Heat Transfer AUGUST 1981, VOL. 103 / 557

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 Flow Between Rotating Coaxial Cylinders With Fully Developed Axial Flow,"
Journal of Fluid Mechanics, Vol. 81,1977, pp. 641-655.
8 Hasoon, M. A., and Martin, B. W., "The Stability of Viscous Axial Flow
in an Annulus with a Rotating Inner Cylinder," Proceedings of Royal Society
of London, Series A, Vol. 352,1977, pp. 351-381.
9 Astill, K. N., "Studies of the Developing Flow Between Concentric
Cylinders With the Inner Cylinder Rotating," ASME JOURNAL OF HEAT
TRANSFER, Vol. 86,1964, pp. 383-391.
10 Astill, K. N., "Modes of Adiabatic Flow in the Entrance Region of an
Annulus with an Inner Rotating Cylinder," Ph.D. dissertation, Massachusetts
Institute of Technology, 1961.
11 Astill, K. N., Ganley, J. T., and Martin, B. W., "The Developing Tan-
gential Velocity Profile for Axial Flow in an Annulus with a Rotating Inner
Cylinder," Proceedings of the Royal Society of London, Series A, Vol. 307,1968,
pp. 55-69.
12 Martin, B. W., and Payne, A., "Tangential Flow Development for
Laminar Axial Flow in an Annulus with a Rotating Inner Cylinder," Proceedings
Log Ra of the Royal Society of London, Series A, Vol. 328,1972, pp. 123-141.
Fig. 9 Comparison of Nu against Ra* with data of [14, 20, 21, 22] 13 Payne, A., and Martin, B. W., "Heat Transfer to Laminar Axial Flow
in a Concentric Annulus from a Rotating Inner Cylinder," Proceedings of the
Fifth International Heat Transfer Conference, Japan Society of Mechanical
Engineers, and Society of Chemical Engineers, Japan, Heat Transfer 1974, II,
Acknowledgment FC 2.7, pp. 80-84.
T h e financial s u p p o r t of A l a z h a r U n i v e r s i t y , w h i c h allowed u s t o 14 Bodoia, J. R., and Osterle, J. F., "The Development of Free Convection
c a r r y o u t t h i s i n v e s t i g a t i o n o n t h e I C L 1900 c o m p u t e r of C a i r o U n i - Between Heated Vertical Plates," ASME JOURNAL OF HEAT TRANSFER, Vol.
84,1962, pp. 40-44.
versity, is gratefully a c k n o w l e d g e d .
15 Carnahan, B., Luther, H. A., and Wilkes, J. O., Applied Numerical
Methods, John Wiley & Sons, New York, pp. 449-475.
References 16 Lapidus, L., Digital Computation for Chemical Engineers, McGraw-
1 Taylor, G. I., "Stability of a Viscous Liquid Contained Between Two Hill, New York, 1962, pp. 254-255.
Rotating Cylinders," Philosophical Transactions, Series A, Vol. 223,1923, pp. 17 Coney, J. E. R., and El-Shaarawi, M. A. I., "A Contribution to the Nu-
289-343. merical Solution of Developing Laminar Flow in the Entrance Region of Con-
2 Yih, C , "Dual Role of Viscosity in the Instability of Revolving Fluids centric Annuli With Rotating Inner Walls," ASME Journal of Fluids Engi-
of Variable Density," The Physics of Fluids, Vol. 4, No. 7, July 1961, pp. neering, Vol. 96, No. 4, Dec. 1974, pp. 333-340.
806-811. 18 Coney, J. E. R., and El-Shaarawi, M. A. I., "Laminar Heat Transfer in
3 Becker, K. M., and Kaye, J., "The Influence of a Radial Temperature the Entrance Region of Concentric Annuli With Rotating Inner Walls," ASME
Gradient on the Instability of Fluid Flow in an Annulus with an Inner Rotating J O U R N A L O F H E A T T R A N S F E R , N O . 4, Vol. 96, Nov. 1974, pp. 560-562.
Cylinder," ASME JOURNAL OF H E A T TRANSFER, May 1962, pp. 106-ilO. 19 Schlichting, H., "Laminar Flow About a Rotating Body of Revolution
4 Walowit, J., Tsao, S., and DiPrima, R. C , "Stability of Flow Between in an Axial Airstream," NACA, TM 1415,1956.
Arbitrarily Spaced Concentric Surfaces Including the Effect of a Radial Tem- 20 Aung, W., Fletcher, L. S., and Sernas, V., "Developing Laminar Free
perature Gradient," ASME Journal of Applied Mechanics, Vol. 31, Dec. 1964, Convection Between Vertical Flat Plates With Asymmetric Heating," Inter-
pp. 585-593. national Journal of Heat and Mass Transfer, Vol. 15, No. 11, Nov. 1972, pp.
5 DiPrima, R. C , "The Stability of a Viscous Fluid Between Rotating 2293-2308.
Cylinders with an Axial Flow," Journal of Fluid Mechanics, Vol. 9,1960, pp. 21 Aung, W., "Fully Developed Laminar Free Convection Between Vertical
621-631. Plates Heated Asymmetrically," International Journal of Heat and Mass
6 Elliot, L., "Stability of a Viscous Fluid Between Rotating Cylinders with Transfer, Vol. 15, No, 8, Aug. 1972, pp. 1577-1580.
Axial Flow and Pressure Gradient Round the Cylinders," The Physics of Fluids, 22 Davis, L. P., and Perona, J. J., "Development of Free Convection Flow
Vol. 16, No. 5, May 1973, pp. 577-580. of a Gas in a Heated Vertical Open Tube," International Journal of Heat and
7 Chung, K. C , and Astill, K. N., "Hydrodynamic Instability of Viscous Mass Transfer, Vol. 14,1971, pp. 889-903.

558 / VOL. 103, AUGUST 1981 Transactions of the ASME

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 Laminar iixed Confection in a
S. Acharya
S. V. Patankar
Shrouded Fin Array
Department of Mechanical Engineering,
University of Minnesota, An analytical study is made to investigate the effect of buoyancy on laminar forced con-
Minneapolis, MN 55455
vection in a shrouded fin array. Two heating conditions are considered; in one, the fins
and the base surface are hotter than the fluid, and in the other, they are colder. The re-
sults are obtained numerically for a wide range of the governing buoyancy parameter. It
is found that with a hot fin and base, the secondary flow pattern is mostly made up of a
single eddy. The influence of buoyancy is significant and leads to Nusselt numbers and
friction factors which are much higher than for pure forced connection. With a cold fin and
base, the presence of a tip clearance between the fins and the shroud generates a multiple
eddy pattern. The resulting stratification is responsible for the existence of high axial ve-
locity and temperature in the clearance region relative to that in the inter-fin space. Com-
pared to the hot fin case, the secondary flow is weaker, and therefore a relatively smaller
increase in the friction factor is obtained. The Nusselt number is found to increase only
in the absence of tip clearance. The distribution of the heat transfer coefficient along the
fin and the base for both heating situations is found to be highly nonuniform.

Introduction Analysis
The heat transfer from fins is often strongly influenced by the de- The geometrical symmetry of the physical situation shown in Fig.
tails of the flow field in which they are placed. An example of such a 1 can be used to confine the analysis to a typical module, which is also
situation is the forced convection through a shrouded fin array, which shown in Fig. 1, where one may observe the use of the thin fin as-
was analyzed by Sparrow, et al. [1]. They demonstrated that the heat sumption t « H and t « s.
transfer from the fin is greatly altered by the presence or the absence The first step in the specialization of the conservation equations
of a clearance between the fin tip and the shroud. Another factor that for mass, momentum and energy is the formulation of the buoyancy
is likely to exert a substantial influence on the fin heat transfer is the term. This term is contained in the equation expressing the conser-
presence of secondary flows induced by the buoyancy forces. Such vation of momentum in the y direction and may be written as
influence has been established for laminar convection in a circular
tube with circumferentially uniform heating [2-6] or nonuniform dp
•—-Pg (i)
heating [7] of the tube walls. The purpose of this paper is to evaluate <>y
the effect of buoyancy on the flow and heat transfer through a The density appearing in (1) can be related to temperature by means
shrouded fin array. of the Boussinesq model, which gives
The geometry to be considered is shown schematically in Fig. 1.
Fluid flows in the direction perpendicular to the plane of diagram. •• pw (1 - 0(T - Tw)) (2)
Although a fin thickness t is shown in the schematic representation, Here Tw denotes the temperature of the fin and the base surface, and
the flow field is calculated from the thin-fin assumption, i.e., the pw is the corresponding density. At this stage, a modified pressure p '
thickness t is regarded as negligibly small in comparison with the fin is defined as
height H or the inter-fin spacing s.
The analysis is performed for laminar flow and for thermally and P' = P + Pwgy (3)
hydrodynamically developed conditions. The shroud is considered
With this definition, expression (1) can be rewritten as
as adiabatic, while the base surface and the fins transfer heat at a
uniform rate per unit axial length to (or from) the fluid flowing
-~+pgi3(T-Tu:) (4)
through the array. Compared to the convection resistance, the con-
duction resistance of the solid material of the base and the fins is
where, in accordance with the constant-density assumption, pm has
supposed to be negligible, so that, at a given cross section, there is a
been replaced by p.
uniform temperature along the fin and the base surface. Since the
The fully developed conditions imply that the velocity components
buoyancy effects will depend on whether heat is transferred from or
u, v, and w do not change with the axial coordinate z, and that the
to the fluid, two separate situations are considered. In one, the fins
. axial pressure gradient itp/dz is equal to dp/dz, where p is the mean
and the base are maintained at a temperature higher than that of the
pressure over a cross section. For the thermal boundary condition
fluid. This situation will be referred to as the "hot fin" case. In the
other situation, designated as the "cold fin" case, the fins and the base considered, all temperatures rise linearly with z, so that
are maintained at a temperature lower than that of the fluid. Further, dT dTw Q'
it is interesting to note that the effects of buoyancy are identical for (5)
dz dz pc p (s/2) (H + c)w
the hot fin case for the geometry shown in Fig. 1 and the cold fin case
with the geometry inverted, and vice versa. Computation
Sluoud
The governing equations for the situation were solved numerically.
The procedure described in [8], in particular the SIMPLER algorithm,
was used. Solutions were obtained over a wide range of values of the
" l" " " " " 11M"
C
" ! 11" " I (

Fin
n
module

governing buoyancy parameter, such that substantial natural con-

vection effects were encountered at the high end of the range. The
t V
gravity
Prandtl number was assigned the value of 0.7 (for air).

Contributed by the Heat Transfer Division for publication in the JOURNAL

OP HEAT TRANSFER. Manuscript received by the Heat Transfer Division
October 7,1980. Fig. 1 Shrouded fin array and the computation module

Journal of Heat Transfer AUGUST 1981, VOL. 103 / 559

Copyright © 1981 by ASME
Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 where Q' is the heat input per unit axial length for the computation The boundary of the computation module shown in Fig. 1 is made
module shown in Fig. 1. Equation (5) has been obtained from an 1 of solid walls and symmetry lines. All velocities are equal to zero on
overall heat balance for the domain. the solid walls. At a symmetry line, the normal velocity is zero, and
With the dimensionless variables defined by the normal gradients of the two other velocity components are zero.
For the temperature equation, the appropriate boundary conditions
X = x/H, Y •- S = s/H, C = c/H (6a) are given by <j> = 0 on the fin and the base surface and the normal
gradient d<j>ldn = 0 on the shroud and on the symmetry lines.
U =- V=- W • (66)
(v/H) • (-dp/dz) (H2/n) Computational experiments were conducted to determine the grid
spacing required for sufficient accuracy. It was found that a 22 X 28
T-Tu,
P =- (6c) grid gave solutions for which the reported results are accurate to 0.5
p(v/H)2'
</> = •
•"' Q'/k percent. A higher density of points was employed close to the solid
the equations expressing the conservation of mass, momentum, and boundaries and near the fin tip (Y = 1).
energy become
Results and Discussion
dU W
0 (7) Results are presented for the two kinds of heating situation and for
dX dY~ Grashof numbers ranging from 104 to 107. Results beyond a Gr + of
dP d2U dHJ 107 have not been shown, since it is unlikely that the flow will remain
(8) laminar under those conditions. The inter-fin spacing S(= s/H) is held
dX dY dX+ dX2 +
dY 2
fixed at 0.5, while the tip clearance C(= c/H) is assigned the values
dP d2V i>2V
(9) of 0, 0.4 and 1.0. As stated before, all results pertain to a Prandtl
dX dY — + —— + + Gr + 0
dY dX2 2 dY2 2 number of 0.7.
dW dW Streamline Patterns. Information about the secondary flow
1 + — 2- + (10)
dX dY dX dY2 pattern can be best obtained from the streamline maps, which will
now be presented. The stream function is obtained from the velocity
W/W
(11) field solution by evaluating the integral
dX dY Pr [dX2 dY2, P r S ( l + C)
Equations (7) to (11) contain two independent parameters, Gr + and UdY (13)
Pr. Gr + is a modified Grashof number defined as Jo
along constant-X lines and with \p = 0aXX = Y = 0. The streamline
g PQ'H3
Gr+ = (12) patterns are shown for both kinds of heating situations—the hot fin
v2k case in Figs. 2, 3 and 4 and the cold fin case in Figs. 5, 6 and 7. The
For a fixed duct height and inter-fin spacing, it is the magnitude of contours have been drawn for all three values of the clearance pa-
Gr + alone that governs the heat transfer results for a particular fluid. rameter (C = 0,0.4,1) and for Grashof number Gr+ of 10 4 ,10 e and 107.
The second parameter is the Prandtl number Pr, which has been set The value of \pm below each streamline map stands for the maximum
equal to 0.7 in this investigation. value of \[/, which is a measure of the vigor of the secondary flow.
It is worth noting that, since a fully developed flow is considered For the hot fin case, the streamlines corresponding to Gr + of 104
here, the cross-stream flow field and the associated heat transfer are and 106 (Figs. 2 and 3) have a single eddy structure. At Grashof
governed only by the Grashof number and are not influenced by the number of 107 and for C = 0 (Fig. 4), a small counter-rotating eddy,
Reynolds number of the axial flow. This is a characteristic of all fully akin to a "thermal" is created. For greater clearances, the stronger
developed flows (see [7] for an additional example), in which, since secondary flows produce a more homogeneous distribution of axial
the axial velocity does not change with the axial distance, the inertia velocity and temperature, thereby restoring the single eddy pat-
of the axial flow plays no role. In a developing flow, however, the tern.
Reynolds number must be considered in addition to the Grashof In the cold fin case, depicted in Figs. 5, 6 and 7, there is a tendency
number. for temperature stratification as evidenced by the multiple eddies

-Nomenclature*
Af = fin surface area ihf = mass flow rate in inter-fin space TB = bulk temperature
AM = heat transfer area of fin and base Nu = average Nusselt number for fin and T,„ = fin and base surface temperature
C = clearance parameter, c/H base surface, hH/k t = fin thickness
c = tip clearance (Nu)o = forced convection value of Nu U,V,W = dimensionless velocities, equation
cp = specific heat at constant pressure P = dimensionless pressure, equation (6c) (6b)
De = equivalent diameter, equation (16) Pr = Prandtl number of the fluid u, v, w = velocity components in x, y, z di-
/ = friction factor, equation (15) p = pressure rections
Gr + = modified Grashof number, p' = modified pressure, equation (3) W = mean value of W
gPQ'HWv* p = mean pressure over a cross section
g = acceleration due to gravity w = mean axial velocity
Q' = prescribed uniform heat transfer rate for
H = fin height X, Y = dimensionless coordinates, x/H,
the computation module per unit axial y/H
hb = local heat transfer coefficient along the
length
_ base, qb/(Tm - TB) x, y = cross-sectional coordinates
hb = average heat transfer coefficient along Ql' = the part of Q' that is transferred by the 2 = axial Coordinate
base, cjb/(Tw - TB) fin /3 = thermal expansion coefficient
h[ = local fin heat transfer coefficient, Qb — local base surface heat flux li = viscosity
_ qf/(Tw - TB) qb - average base surface heat flux v = kinematic Viscosity
hf = average fin heat transfer coefficient, (// = local fin heat flux p = density
_ <jf/(Tw - TB) cjf = average fin heat flux <l> = dimensionless temperature, equation
h = average heat transfer coefficient for fin Re = Reynolds number, WDe/v (6c)
and base surface, equation (14) S = inter-fin spacing parameter, s/H ip = dimensionless stream function, equation
k = thermal conductivity of fluid s = inter-fin spacing (13)
m = total mass flow rate T = temperature <//m = maximum value of i/'

560 / VOL. 103, AUGUST 1981 Transactions of the ASME

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 O1.0

O0.4 C=Q4

c=o
oo

Fig. 2
i'ow *m*o.i8 *m=a33
Streamline pattern for the hot fin case (Gr' = 104)
Mi
Fig. 4 Streamline pattern for the hot fin case (Gr + = 107)

O1.0

C=Q4 C=Q4

C=0

u
fmom $m°o.m *mo.i9
Fig. 3 Streamline pattern for the hot fin case (Gr + = 106) Fig. 5 Streamline pattern for the cold fin case (Gr' = 104)

Journal of Heat Transfer AUGUST 1981, VOL. 103 / 561

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 C=I.O
O1.0

Fig. 6 Streamline pattern for the cold tin case (Gr+ = 106) Fig. 7 Streamline pattern for the cold fin case (Gr + = 107)

formed in the clearance region. The stratification is further enhanced

with an increase in either Gr + or C. Such an effect reduces the cross-
sectional temperature difference in the inter-fin region and thus di-
minishes the strength of the secondary flow.
Centerline Profiles of Axial Velocity. Figure 8 shows the nor-
malized axial velocity profile along the base bisector for the hot fin
case. For both values of C (i.e., 0 and 1) the curve for Gr + = 104 lies
remarkably close to the curve corresponding to pure forced convec-
tion, thereby implying the presence of only a very weak secondary
flow. This effect can be repeatedly observed in many of the graphs
to be presented later and indicates that a certain threshold value of
Gr + (around 104) has to be crossed before buoyancy can cause any
detectable change in the results relative to those for pure forced
convection.
For zero tip clearance (C = 0), the symmetrical distribution in ab-
sence of natural convection becomes left-skewed (Fig. 8) at higher
Fig. 8 Axial velocity distribution along the base bisector for hot fin case
Grashof numbers. This is caused by the downflow leg of the main eddy
carrying with it the high axial velocity fluid from the interior of the
duct. With tip clearance (C = 1) and in the absence of secondary flow,
the fluid in seeking the path of least resistance attains higher axial
velocity in the clearance region than in the inter-fin space. With
stronger secondary currents (at higher Gr + ), the axial velocity dis-
tribution becomes more homogeneous as evidenced in Fig. 8.
For the cold fin case (Fig. 9), the secondary flow is downward along
the fin and upward along the base bisector. Thus, for C = 0, the
symmetrical distribution in absence of natural convection becomes
right-skewed when buoyancy is present. In the presence of tip clear-
ance, high axial velocities in the clearance region are preserved due
to the occurrence of temperature stratification which confines the
main secondary eddy in the inter-fin region.
Overall Heat Transfer Results. The overall heat transfer
characteristics of the system will be presented through two dimen-
sionless groups: the ratio of average Nusselt number to its forced
convection value, Nu/(Nu) 0 and the ratio Q/'/Q', where Q/' is the
surface integrated fin heat loss (or gain) and Q' is the total heat loss Fig. 9 Axial velocity distribution along the base bisector for the cold fin
(or gain) from the fin and the base surface. The average Nusselt case

562 / VOL. 103, AUGUST 1981 Transactions of the ASME

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 number is evaluated as 28 |2.25

hH 24 Hot fin 2.0

h=- ,Nu (14)
(H + s/2)(Tw - TB) k Cold fin / -
20 r 1.75
where TR is the bulk temperature of the fluid.
Figures 10 and 11 show the distributions of Nu/(Nu) 0 and Qf'/Q'
with respect to Gr + . The solid lines correspond to the hot fin case and
_ 16 - C=0/ - 1.5
are referred to the left hand ordinate. The dashed curves are for the
Nu // SI
GR 1 2 // 1.25(Nu)o
cold fin case and are keyed to the right hand ordinate. s /
For the hot fin case, the Nusselt number (Fig.10) increases with Gr + 8- . #"""'/ 1.0 —. 1.0
and also with the tip clearance, in accordance with the strengthening . -Jf-y? Q4
of the buoyancy-induced secondary flow. For the results presented, 4- -~5A 0.75
a Nusselt number as high as 24 times its forced-convection value is
—c-o
encountered. In the cold fin case, Nusselt number increases only in O | 1 1 1 1 0.5
absence of tip clearance. Both for C = 0.4 and C = 1.0 the curves fall 10s 10' 105 ^ 106 10'
Gr+
slightly below the Nu/(Nu) 0 = 1 line thereby implying a small adverse
effect of buoyancy on the overall heat transfer. Such an effect is un- Fig. 10 The Nusselt number ratio as a function of the Grashof number
derstood by noting that the stratification at the top is responsible for
reduced temperatures and axial velocities in the inter-fin region. In
1J0
addition, the major contribution to the value of the bulk temperature .iO Hoi fin
TB, used in the definition of Nusselt number Nu, comes from the hot 0.9 _ r>n - Cold fin
fluid in the clearance region, which however does not play an active
role in the heat transfer process from the thermally active walls.
The ratio Qf'/Q' plotted in Fig. 11 for the hot fin case is seen to
0.8 - •k\S
decrease with Gr + up to Gr + = 106, followed by a slight increase. The Q; 0.7 1.0 1.0
initial decrease is due to the presence of secondary flow which causes
Q' 0.6 04
the fluid to become thermally saturated as it flows along the base, and 0.95
thus, the fluid reaching the fin is not particularly heat receptive. As
c-o,. --'" Q;
the velocity of the secondary flow increases, the thermal saturation 05 - 0.9 Q>
effect diminishes, and this manifests itself beyond Gr+ of 106 through
an increase in Qf'/Q'- 0.4 - 0.85
It should be noted, that up to a Gr + of about 5 X 106, Qf'/Q' exceeds
03 1 1 1 i 0.8
Af/Aht = 0.8, thereby implying that, on a unit area basis, the fins 103 10* 10° 10" 10'
constitute a more effective heat transfer surface than the base. Gr+
However, beyond a Gr + of 5 X 105, the presence of strong secondary
Fig. 11 Fractional fin heat loss as a function of the Grashof number
flows make the base a more effective heat transfer surface.
In the cold fin case, the fin is always a more effective heat transfer
surface than the base. This is because the base receives fluid which
is precooled by the fin. With increasing Gr + , this thermal saturation
effect becomes particularly pronounced leading to higher Qf'/Q':
Friction Factor. The friction factor is evaluated from the stan-
dard definition
f=-(dp/dz)DJ(\pw2) (15)

in which the equivalent diameter D e for the computation module

is
De = 4(s/2) (H + c)/(H + s) (16)

If equation (15) is rewritten in terms of the dimensionless variables,

there results
103 10" 10'
. 10" 10'
/Re = (8/W) |S(1 + C)l(l + S)) 2
(17) Gr+
Fig. 12 The friction factor ratio as a function of the Grashof number
ith
Mass Flow Ratios. To characterize the flow distribution, the ratio
Re = wDg/v (18)
rhf/m was evaluated, where rhf is the mass flow rate in the inter-fin
Figure 12 shows the variation of (/Re)/(/Re) 0 with Gr + . Here (/Re) 0 space and rh is the total mass flow rate in the module.
is the forced convection value of/Re. The results for the hot fin case For C = 0, the mass flow ratio rhf/m is obviously equal to unity (Fig.
are shown by solid lines and are referred to the left hand ordinate, 13). For the hot fin case with C ^ 0, the secondary flow becomes more
while the results for the cold fin case drawn with respect to the right vigorous with higher Gr + , leading to a more homogeneous axial ve-
hand ordinate are shown by dashed lines. locity distribution and thus a higher rhf/m value. For the cold fin case,
Stronger secondary flows are associated with increasing values of an increase of Gr + produces more stratification thereby retaining high
Gr + and higher (lower) values of clearance C for the hot (cold) fin case. axial velocities in the clearance zone and thus reducing the rhf/m
As expected, the greater the vigor of the secondary flow, the higher values.
is the axial pressure gradient or the value of/Re. Since the buoyancy Local Heat Transfer Rates. The distributions of the normalized
effects are more vigorous for the hot fin case, the increase in /Re values heat transfer coefficient along the fin hf/hf are presented in Figs. 14
is also proportionately higher. Further, it is interesting to note that, and 15 for the hot fin case and the cold fin case respectively. The re-
whereas the Nusselt number for the hot fin case is increased by as sults for pure forced convection were found to lie very close to those
much as 24 times, the corresponding friction factor is increased only for Gr + = 104, while the results for Gr + = 106 display a trend exactly
about four times. similar to that predicted with Gr + = 107. Hence results have been

Journal of Heat Transfer AUGUST 1981, VOL. 103 / 563

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 Fig. 13 Fraction of mass flow rate through the inter-fin region as a function
of the Grashof number

presented for Gr + values of 104 and 107 and for all three values of
C.
The heat transfer coefficient along the fin hf and the average fin
heat transfer coefficient hf were evaluated from the following defi-
nitions

h/ = qfl(Ta - TB) hf = q,l(Tw ~ TB) (19)

where only qf varies along the isothermal fin. It should be noted that
hf/hf = q/lqf.
For the hot fin case, the results for pure forced convection or for low 0 0.2 0.4 0.6 OJB 10
Gr + are identical to those presented by Sparrow, et al. [1]. For C = Y
0, hf has its maximum value midway along the fin where the fluid axial Fig. 14 Variation of the fin heat transfer coefficient for the hot fin case
velocity is a maximum. In the presence of tip clearance, hf increases
monotonically to the tip, which is washed by high axial velocity fluid.
Under the action of buoyancy (Gr+ = 107), the maximum heat transfer
coefficient in the presence of tip clearance occurs at two locations
characterized by the presence of high axial velocities. One such lo-
cation is at the fin tip and the other is between the mid span of the fin
and the base. With C = 0, the maximum axial velocity and therefore
the maximum value of hf/hf occur at the midplane of the fin.
For the cold fin case (Fig. 15), the secondary flows are rather weak
at a Gr + of 104, resulting in a hf/hf distribution similar to that in Fig.
14. With a more vigorous secondary flow, the heat transfer coefficient
in the presence of tip clearance increases monotonically to the tip.
This effect is attributable to the occurrence of temperature stratifi-
cation, which confines the high axial velocity fluid in the proximity
of the fin tip. With C = 0, the upflow leg of the secondary flow eddy
carries the high axial velocity from the midplane towards the fin tip.
Since at the fin tip the resistance to flow is relatively large, the max-
imum heat transfer coefficient occurs close to, but not at, the fin
tip.
The distribution of hf,/h/,, which represents the ratio of the local
base heat transfer coefficient hi, to the average base heat transfer
coefficient hi,, is presented in Figs. 16 and 17. The coefficients hb and
hi, are defined in the same way as the fin heat transfer coefficients hf
and hf.
For the hot fin case (Fig. 16), the flow is in a clockwise direction.
Thus, the fluid becomes less and less amenable to heat transfer as it
flows towards the fin, which explains why ht,/h/, decreases away from
the base mid point. For Gr + = 107 and in the absence of any clearance,
hb/hb is maximum at the separation streamline between the two
eddies (i.e., at x/H c^ 0.125. See Fig. 4). This is due to the convection
of high axial velocity and low temperature fluid along the separation
streamline. With tip clearance, secondary flows are strengthened
thereby removing the small eddy and the effect of the separation
streamline on the nature of the h^/h/, distribution.
Figure 17 shows the hb/hb distribution for the cold fin case. Since 0.2 0.4 0.6 Q8 10
the fluid flows down the fin and then over the base, regions of the base Y
closer to the fin can transfer more heat than the regions away from Fig. 15 Variation of the fin heat transfer coefficient for the cold fin case

564 / VOL. 103, AUGUST 1981 Transactions of the ASMH

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 1.8

u
0.05 0.1 0.15 0.2 0.25
X
Fig. 17 Variation of the base heat transfer coefficient for the cold fin
Fig. 16 Variation of the base heat transfer coefficient for the hot fin case case

the fin. Thus a more uniform distribution of hb/hi, relative to the ENG 75-18141 AOl.
forced convection distribution is obtained.

Concluding Remarks References

The results of this investigation have demonstrated that the 1 Sparrow, E. M., Baliga, B. R., and Patankar, S. V., "Forced Convection
buoyancy forces significantly affect the heat transfer characteristics Heat Transfer from a Shrouded Fin array with and without Tip Clearance,"
of laminar forced convection in a shrouded fin array. JOURNAL OF HEAT TRANSFER, NOV. 1978, Vol. 100, pp. 572-579.
2 Cheng, K. C, and Hong, S. W., "Combined Free and Forced Laminar
For the hot fin case, the flow pattern is in general made up of a Convection in Inclined Tubes," Applied Scientific Research, Vol. 27,1972, pp.
single eddy. Secondary flows, which become stronger when either Gr + 19-38.
or C is increased, produce a more homogenized axial velocity and 3 Newell, P. H., and Bergles, A. E., "Analysis of Combined Free and Forced
temperature distribution. The strengthening of the secondary flows Convection for Fully Developed Laminar Flow in Horizontal Tubes," ASME
JOURNAL OF HEAT TRANSFER, Vol. 90,1970, pp. 83-93.
produces a substantial increase in both friction factor and the Nusselt 4 Farris, G. N., and Viskanta, R, "An Analysis of Laminar Combined
number relative to their forced convection values, but causes a decline Forced and Free Convection Heat Transfer in a Horizontal Tube," Interna-
in the relative effectiveness of the fin as a heat transfer surface. tional Journal of Heat and Mans Transfer, Vol. 10, 1967, pp. 1625-1629.
For the cold fin case, the presence of a tip clearance produces a 5 Hussain, N. A., and McComas, S. T., "Experimental Investigation of
Combined Convection in a Horizontal Circular Tube with Uniform Heat Flux,"
multiple-eddy flow pattern. The secondary flows are weaker than Heat Transfer, Vol. 4, Paper No. NC 3.4, Elsevier Publishing, Amsterdam,
those in the hot fin case and are further dampened with increased 1970.
stratification at high Grashof numbers. The axial velocity and tem- 6 Bergles, A. E., and Simonds, R. R,, "Combined Forced and Free Con-
perature are higher in the clearance region than in the interfin space. vection for Laminar Flow in Horizontal Tubes with Uniform Heat Flux," In-
ternational Journal of Heat and Mass Transfer, Vol. 14, 1971, pp. 1989-
While the friction factor value increases with Gr + , the Nusselt number 2000.
does so only in the absence of any clearance. The fin is almost always 7 Patankar, S. V., Ramadhyani, S., and Sparrow, E. M., "Effect of Cir-
a more effective heat transfer surface than the base. cumferentially Non-Uniform Heating on Laminar Combined Convection in
a Horizontal Tube," ASME JOURNAL OF HEAT TRANSFER, Vol. 100, Feb. 1978,
pp. 63-69.
Acknowledgment 8 Patankar, S. V., Numerical Heat Transfer and Fluid Flow, McGraw-
This research was performed under the auspices of NSF Grant Hill—Hemisphere, 1980.

Journal of Heat Transfer AUGUST 1981, VOL. 103 / 565

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 Combined Free and Forced
C. Prakash
Convection in Vertical Tubes
S. V. Patankar
Department of Mechanical Engineering,
University of Minnesota,
with Radial Internal Fins
Minneapolis, MN 55455
An analysis is made of the fully developed laminar flow and heat transfer in vertical tubes
with radial internal fins to determine the influence of the buoyancy forces. The governing
equations for velocity and temperature are solved by a finite difference technique which
incorporates a special scheme for treating the two coupled variables. Results are present-
ed for a range of the Rayleigh number and for various values of the fin height and the
number of fins. The buoyancy force is found to increase significantly both friction and
heat transfer in the finned tube; augmentation factors in the range of 5 to 10 are encoun-
tered. The effect of buoyancy is particularly strong when the number of fins is small and
the fins are short.

Introduction Analysis
Internally finned tubes have found wide use in compact heat ex- Statement of the Problem. The problem to be investigated is
changers. The main purpose of the fins is to provide additional surface that of fully developed laminar combined free and forced convection
area for augmenting the heat transfer. An extensive review of heat through a vertical internally finned circular tube, the cross section
transfer enhancement techniques has been presented by Bergles [1], of which is shown in Fig. 1. The fins are straight, continuous, and
in which different internal fin arrangements are described. The fins equally distributed around the tube circumference. The thickness of
are usually radial and they may be straight or spiral. The present work the fins is assumed to be negligible. The tube receives a uniform heat
is restricted to straight radial fins. input per unit axial length, and at any axial location, a uniform tem-
The problem of fully developed laminar flow in circular tubes with perature is taken to prevail at all points on the fins and the tube wall.
radial internal fins has been studied analytically [2] and numerically This assumption implies that the fins and the tube are made of a
[3,4]. However, the effect of buoyancy on the forced convection has highly conducting material. The fluid properties are assumed constant
not been investigated so far. For finless circular tubes, the problem except for the variation of density in the buoyancy term. Finally,
of fully developed combined free and forced convection has been viscous dissipation and compression work terms are neglected in the
solved analytically [5, 6] and investigated experimentally [7], The energy equation. For a fully developed flow, the velocity components
purpose of the present paper is to obtain numerical solutions for the in the cross-sectional plane are zero and the axial velocity is itself
free and forced convection in vertical internally finned circular independent of the axial distance. Further, for the thermal boundary
tubes. condition considered, all temperatures rise linearly in the axial di-
Only certain thermal boundary conditions admit fully developed rection at the same rate.
solutions for the buoyancy-affected flow. One such boundary condi- Governing Equations. Under the foregoing assumptions, the
tion is that of uniform heat input per unit axial length. Because of its equations governing the problem are:
practical relevance, this boundary condition was chosen for the
present study. Further, the tube wall and the fins were supposed to
be of high thermal conductivity so that they would assume a uniform drr ))
r dr V d r2 d0 2
2
a- (i)

temperature over any cross section.

+ ad
2
T
2
„ .dT
=w ,_-
— (2)
For a heated tube, the buoyancy forces assist the wall-near flow r dr\ dr) r d0 dz
if the fluid flows upwards. In a cooled tube, on the other hand, an
where W is the axial velocity and T is the temperature; other symbols
upward flow is opposed near the wall. Further, it may be noted that
are defined in the Nomenclature. The effect of natural convection is
a heated upflow behaves identically to a cooled downflow. Similarly,
introduced through the Boussinesq approximation:
a cooled upflow is equivalent to a heated downflow.
For finless circular tubes, the cooled upflow (or heated downflow) p = pw (1 + p (7V - T)) (3)
is known [5] to give useful solutions only for very small values of the
Rayleigh number. At a Rayleigh number of about 33, the centerline where Tw(z) is the wall (or fin) temperature at an axial location z, pw
velocity approaches infinity and the solution is said to "run away." the corresponding density, and /5 the coefficient of volumetric ex-
Since a similar behavior is expected also for the finned tubes, the pansion. Since Tw rises linearly with z, pw is a function of z. However,
present investigation was restricted to only heated upflow (or cooled in accordance with the constant-density approximation, pw will now
downflow). For this situation the effect of buoyancy is to accelerate be interpreted as a constant density corresponding to some reference
the flow near the wall while retarding the flow in the core. At a critical location.
value of the Rayleigh number, this retardation tends to cause reverse
flow. Since fully developed solutions involving reverse flows are not
Computational
expected to occur in practical situations, the numerical computations
Domain
were not carried out beyond the critical Rayleigh number.
In addition to the Rayleigh number, two geometrical parameters,
namely the fin height and the number of fins govern the problem. The
number of fins was varied between 4 and 32, and the fin height was
increased up to 80 percent of the tube radius.

Contributed by the Heat Transfer Division for publication in the JOURNAL

OF HEAT TRANSFER. Manuscript received by the Heat Transfer Division
October 7,1980. Fig. 1 Circular tube with fins and the computational domain

566 / VOL. 103, AUGUST 1981 Transactions of the ASME

Copyright © 1981 by ASME

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 For an arrangement of N equally spaced fins, the required calcu- where the integrals are taken over the tube cross section. Another
lation domain is shown in Fig. 1. It can be seen to be a sector of angle important quantity appearing in equation (10) is the Rayleigh number
TT/N extending from one fin to a location half-way towards the next Ra defined by
fin. The boundary conditions for this domain are straightforward. On
the tube wall and the fin surface, the velocity W is zero and the tem- Ra = (pwPga*dTb/dz)/(<xii) (14)
perature equals Tw- The remaining boundaries of the domain (shown The presence of 7rfl in equations (10) and (11) can be suppressed
by dashed lines) are lines of symmetry implying dW/d0 and dT/dB by introducing a new variable ij> defined by
to be zero there.
The axial gradient of temperature dT/dz can be related to the heat 4> = tf-ne (15)
input Qt per unit axial length by an overall heat balance:
As a result, equations (10) and (11) become
dT _dTw _dTb _Qt
(4) V2Q + Ra0 + 1 = 0 (16)
dz dz dz mc
where Tb is the bulk temperature defined by: and

Tb = SJWTrdrdd/SSWrdrde (5) V2<j> - Q = 0 (17)

2
the integration being over the whole cross section of the tube, m is the where V is the two-dimensional Laplacian operator given by
mass flow rate through the tube and is given by v^il[,i| + I^ (18)
m = pW (TTO2) = p SSWrdrdd (6) RdR\ dRj R2d82
As regards the boundary conditions, fi and cj> are both zero on the
where a is the radius of the tube. Incidentally, equation (6) defines tube wall and the fin surfaces, while their normal gradient is zero on
the mean velocity W as well. the remaining boundaries of the domain.
The problem can be nondimensionalized by introducing Heat Transfer and Pressure Drop Parameters. Results of
pressure drop due to friction may be presented through the value of
G = (T TwWQt/k) (7) /Re where,
; dp dp
(Wp)l
and
dz
• Pwg (8) friction factor / = De
dz
Pwg
'/£ 2
Pw W \ (19)

and
R = r/a (9) Reynolds number Re = pwWDe/n (20)

where k is the thermal conductivity of the fluid. These definitions lead Here De is the equivalent hydraulic diameter defined by:
to the following dimensionless form of the governing equations.
De = 4 X cross-sectional area/wetted perimeter
1 d / dn\ 1 d2Q which for the finned tube is given by:
\R— + 2 + wSlRaO + l = 0 (10)
R dR dR R dd2 '
De = 2a/((Nh/w) + 1) (21)
and where h is the dimensionless fin height £/a. A combination of equa-
i a ae 1 a2e „„ _ tions (12, 19) and (20) leads to
\R — + —2 — 2- - Q(TTQ) = 0 (11)
R dR \ dR R d9 fRe = 2De2/(Qa2) (22)
in which 0 is the mean value of Q and is related to W by: Thus, after a solution for fi has been obtained, /Re can be obtained
from the value of Q.
dp
Q = Wfi/ • -— Pwg (12) The heat transfer coefficient is based on the difference between the
dz wall temperature and the bulk temperature. This temperature dif-
From a given distribution of A, Q, can be calculated from ference is given by

Q= SSQRdRde/SSRdRdd (13) Tw-Tb = -e Qtlk (23)

-Nomenclature-
a = tube radius Nu = Nusselt number z = axial coordinate
Af = fin heat transfer area Nuo = Nusselt number at zero Rayleigh a = thermal diffusivity of the fluid
At = total (fin plus circular tube) heat number /3 = coefficient of volumetric expansion of the
transfer area Qf = heat transferred by the fin per unit axial fluid
c = specific heat of the fluid length 8 = angular coordinate
D = tube diameter la Qt = total heat input per unit axial length
#o = half the angular spacing between the
De = equivalent hydraulic diameter of r = radial coordinate
fins
channel, equation (21) R = dimensionless radial coordinate r/a
/ = friction factor, equation (19) Re = Reynolds number (pw WDe)/p. 0 = dimensionless temperature, equation
(/Re)o = /Re at zero Rayleigh number Ra = Rayleigh number (7)
g ~ acceleration due to gravity (Pw^gaidTb/dz)l(afi) p. = fluid viscosity
h = dimensionless fin height T = fluid temperature p = fluid density
hr = heat transfer coefficient, equation Tb = bulk temperature of the fluid, equation pw = fluid density at wall temperature
(25) (5) (p = the variable irQQ, equation (15)
k = thermal conductivity of the fluid Tw = tube wall and fin temperature Q = dimensionless axial velocity, equation
£ ~ fin height W = axial velocity of the fluid _(8)
>n = total mass flow rate W = average velocity of the fluid, equation Q = average dimensionless velocity, equation
N = number of fins (6) (13)

Journal of Heat Transfer AUGUST 1981, VOL. 103 / 567

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 where where the subscript i has been dropped for convenience and the
quantities ej and fj are given by
9 = SfQQRdRde/ffaRdRde (24)
ej = nj + PijQi+ij + qijtoi-ij (31)
the integration being performed over the tube cross section.
With external heating being applied on the circular tube of diameter fj = Pij<t>i+ij + qu<t>i-ij (32)
D (= 2a), the heat transfer coefficient hr is defined as
The fl and 0 values entering in the definitions of ej and fj are the
hT = Qt/(TD(TW - T t )) (25) current values available in the iterative procedure.
Equations (29) and (30) with the current estimates of ej and fj
and thus the Nusselt number Nu is given by
represent sufficient equations for finding the values of Q and 4> at all
the grid points along the chosen line. After the simultaneous solution
Nu = — = - l / ( i r 9 ) (26)
for these equations has been obtained (by the method of Appendix
k
A), the next line is handled in a similar manner. In this way, lines for
all values of i are visited and the fl and <j> values updated there. The
Computational Procedure procedure is then repeated for lines identified by the values of ;'.
The computational task consists of the solution of equations (16) Sufficient repetitions of this scheme finally lead to the converged
and (17). These can be seen to be the familiar Poisson equations, solution. It may be noted that, since the line equations treat 0 and
which are coupled through their source terms. They can be solved by <j> as simultaneous variables, convergence is obtained even when the
any standard finite-difference technique. At first it appears that the coupling is strong.
coupling between the equations can be handled by an iterative tech- Although the procedure described so far converged satisfactorily
nique with successive substitution. Thus, with a guessed field of <j>, in most cases, rather slow convergence was encountered when the
equation (16) is solved to obtain a field of fl; this, in turn, is substituted number of fins was large and the fin height was small. This results
into equation (17), which is then solved to obtain new values of <t>. The from the fact that, when the angular width of the calculation domain
process is repeated until the values of fi and <j> cease to change. is small, the circumferential coefficients pij and qij in equations (27)
Such a simple-minded scheme, however, was found to converge for and (28) are much greater than the radial coefficients by and cy.
only small values of the Rayleigh number. At large values of Ra, for Further, with small fin height, the boundaries normal to the 6 direc-
which the equations are strongly coupled, the successive-substitution tion are largely symmetry lines and do not provide direct information
technique tended to diverge. At times, convergence could be obtained about the values of Q and <j> there.
by heavy underrelaxation, but then the approach to convergence was This difficulty was eliminated by supplementing the line-by-line
extremely slow resulting in excessive computer time. scheme by a block-correction procedure, which uses the concepts of
To describe a more satisfactory solution procedure, the differential "additive corrections" generalized by Settari and Aziz [10]. The al-
equations (16) and (17) are first written in finite-difference form as gebraic details of the block correction are given in Appendix B. The
follows. central idea of the procedure is to update given fields of fly and <j>tj
by adding corrections fl'y and 4>'j which are uniform over a line of
ttifoij - Pijtii+lj + Qij&i-lJ + byQij+i + Cyflij_i constant r (i.e., for a given value of j). These corrections are obtained
+ dij^ij + rtj (27) by solving the integral conservation equations derived by summing
equations (27) and (28) for all values of j . The integral equations for
fl'i and (p'j are also coupled in the same manner as the line equations
and (29) and (30) and are therefore solved by the procedure of Appendix
A. It is possible to use block correction also in the other direction with
ai;'0i; = Pij<t>i+lJ + Qij<t>i-lj + bijfcj+i + Cijfaj-i + gijttij (28)
Q'j and </>'{ as uniform corrections over a radial line. However, for the
Here, the subscripts i and j denote grid locations in the 6 and r di- geometry and the boundary conditions of the present problem, this
rections respectively. Equation (27) expresses fly in terms of the fl did not seem particularly helpful.
values at the four neighboring grid points and in terms of the value To summarize, the coefficients in the finite-difference equations
of <t>ij. Similarly, equation (28) gives <j>ij in terms of its neighbors and (27) and (28) are first calculated. These equations are then solved by
Qij. All the coefficients in these equations, except dy, depend simply an iterative technique. One overall iteration cycle includes the block
on geometrical properties of the grid, and the expressions for them correction and the line-by-line solution of the coupled equations on
can be obtained from any suitable finite difference formulation. The all radial and circumferential lines. The iteration process is continued
actual formulation used in this study has been described in detail in till convergence.
[8]. The term r;y arises from the constant source term of unity in
equation (16); there is no corresponding term, it is worth noting, in Results
equation (28). The coefficients d;y and gij represent the coupling be- The computations were performed on a 32 X 14 grid in the r — d
tween the two equations; dij depends on the Rayleigh number Ra. coordinates. A nonuniform grid spacing was used; the grid lines were
The procedure for solving the algebraic equations (27) and (28) is closely packed near the fin, the fin tip, and the tube wall. Exploratory
based on the practice of solving them by a line-by-line method [8]. The calculations on finer grids and on grids of different nonuniformity
method involves solving simultaneously for all the variable values were used to establish that the presented results are accurate to at
along one grid line, which can be accomplished by an efficient algo- least 0.1 percent. No underrelaxation was necessary while solving the
rithm. In the present work, the line method is extended to the si- coupled equations by the method just described. About 25 iterations
multaneous solution of two coupled variables. The complete details were needed to attain convergence to at least five significant figures.
of the algorithm (which may also be found in other places, for example It is interesting to note that the number of iterations was independent
[9]) are given in Appendix A. The central idea of the algorithm is as of the value of the Rayleigh number; this is a direct result of the si-
follows. multaneous solution for the two coupled variables.
If equations (27) and (28) are written for a line identified by a cer- A range of values of fin height and number of fins was considered.
tain value of i, then the algebraic equations connecting the grid points For each value of fin height and number of fins, solutions were ob-
along that line read tained for increasing values of the Rayleigh number. The Rayleigh
number was gradually increased till reverse flows started appearing
ajQj = bjfly+i + Cjilj-t + dj<j>j + ej (29)
in the solution. As stated before, fully developed solutions beyond this
and critical Rayleigh number are not expected to occur in practical sit-
uations.
aj4>j = bj<pj+1 + Cj<j>j-! + gjUj + fj (30) Velocity Profiles. Figure 2 shows the effect of natural convection

568 / VOL. 103, AUGUST 1981 Transactions of the ASME

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 b 90

80 _
4 h=0.5
N=I6 - h=Q8

_50O_
e--e0
3 60 \ - 0 ^
w. (f Re),
'o
W J000__ 50 -
So2"-~^
2
4 0 -- \«-rr0.6

1 \> " 500 "sNA 30

•^^\«^h=0.5
Ra=0 ~^\>M
20 - h=Q2^
0- ^
10, I i i i l 1 l 1
20 24 16 28 32
Fig. 2 Effect of natural convection on velocity profile on a radial line midway N
between two adjacent fins Fig. 4 ffle as a function of fin height and number of fins for zero Rayleigh
number, i.e., no natural convection

32

28 -
—h=0.8
24 -

20 -
Nuo /'N.
16
f \^
k(Tw-T)' 12 \ .
-# :
:z^i
--^hjO.6
Q,
8 -/

4
h=0 h=0.2
I i i i i I I
C 4 8 12 16 20 24 28 32

Fig. 3 Effect of natural convection on temperature distribution on a radial Fig. 5 Nusselt number as a function of fin height and number of fins for zero
line midway between two adjacent fins Rayleigh number, i.e., no natural convection

on the flow velocity for a typical case. Here, W/W is plotted as a

function of the radial coordinate Ratd = B0 which corresponds to the The results for non-zero Rayleigh number are normalized with these
symmetry line between two adjacent fins. The dimensionless fin basic values. The (/Re)o curves are the same as those given by Hu and
height h is 0.5. At zero Rayleigh number (no natural convection), the Chang [2]. The Nusselt number Nu 0 cannot be directly compared with
smaller velocities in the outer region are caused by the greater resis- the values in [2] because of a different thermal boundary condition
tance provided by the presence of the fins. Most of the flow, therefore, used there. The data presented in Figs. 4 and 5 have also been given
remains in the main core. As the Rayleigh number increases, the in a tabular form by Masliyah and Nandakumar [4]. (In Fig. 5 only
hotter fluid near the tube wall and the fins is accelerated by the a part of the curve for h = 0.8 has been included; the remaining part
buoyancy force. Consequently, for a given flow rate, the velocity in would go far beyond the scale chosen for the figure.)
the center of the tube decreases in comparison. At very large Rayleigh Figure 6 shows the effect of natural convection on Nu and /Re for
numbers, value of W/W in the center of the tube can be seen to be- a moderate number of fins {N = 6) for four different fin heights. For
come very small and even negative. Further, it can be noted that an each case, the Rayleigh number range is limited by the appearance
increase in the Rayleigh number makes the velocity profile steeper of reverse flow. The larger the value of the fin height, the higher is the
at the tube wall leading to greater friction. value of this critical Rayleigh number. Both /Re and Nu increase with
Temperature Profiles. Temperature profiles for the same lo- the Rayleigh number with /Re increasing faster. At a given Rayleigh
cation and conditions are shown in Fig. 3. For the shown case of six- number, shorter fins tend to have a higher /Re/(/Re)o and Nu/Nuo
teen fins, the temperature of the wall (and fins) seems to prevail in than longer fins. This behavior can be understood by noting that the
the outer region beyond R = 0.5; this is expected because of the close shorter fins require, due to their smaller area, greater nonuniformity
proximity of the fins, which have a dimensionless height h = 0.5. In of cross-sectional temperature to transfer a given heat input Qt. The
the main core, the temperature is significantly different from the wall effect of buoyancy is thus stronger for shorter fins. The curves marked
temperature. At low Rayleigh numbers, the central region of the tube h = 0 in Fig. 6 correspond to the finless tube, for which the friction
is able to sustain much lower temperatures since the velocity there factor is seen to rise very rapidly with the Rayleigh number. Also,
is high. As the Rayleigh number increases, the temperature distri- compared to the finned tubes, a finless tube is seen to sustain a rather
bution tends to become more uniform because the low-velocity core small Rayleigh number, which is around 600.
region is now unable to maintain substantially lower temperatures. Figure 7 presents the same results as Fig. 6 except that the number
Friction and Heat Transfer. It is convenient to present the re- of fins is 16. For such a large number of fins, the fin height used in
sults for the friction factor and the Nusselt number with reference practice is expected to be smaller. Therefore, the values of h beyond
to their values for zero Rayleigh number. These reference values, 0.5 were riot included in the computations. The results for sixteen fins
therefore, are first presented in Figs. 4 and 5, where the variation of can be seen to be similar to those in Fig. 6.
(/Re) 0 and Nu 0 with the number of fins N and the fin height h is The striking conclusion from Figs. 6 and 7 is that the buoyancy
shown. The subscript 0 identifies the case of zero Rayleigh number. forces have a significant effect on friction and heat transfer. Aug-

Journal of Heat Transfer AUGUST 1981, VOL. 103 / 569

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 3.0
v-""*"V^t" • • ••*» H=r " . - « - - V

2.5 -2
— fRe/(fRe)0 \\
1 2.0 — Nu/Nu 0 ^
Nu N=6
Nuo
1.5 6|
fRe
Wei
1.0 8
1
0.5 25 10
10 10 KT
Ra
10"
— fRe/(fRe)0
\ \ \r N = 2 4
Fig. 6 Effect of natural convection on (Re and Nusselt number for tubes with 0.5 — Nu/Nu n
six fins of varying fin heights
h=0.2
14
10 102 I0 3 10"
Ra
Fig. 8 Effect of natural convection on /Re and Nusselt number for tubes with
fins of dimensionless fin height h = 0.2 and different number of fins

F0.5

6 I 10
fRe
(fRe)0

1
Nuc

2-

Fig. 7 Effect of natural convection on fRe and Nusselt number for tubes with
sixteen fins of varying fin heights — fRe/ffRe),
— Nu/Nu,
mentation factors of 5 to 10 are easily encountered. h = 0.5
Figures 8 and 9 compare the effect of natural convection on tubes
with different number of fins for the same fin height. The fins con- 10' 10°
Ra
sidered in Fig. 8 are rather short, the dimensionless fin height being
Fig. 9 Effect of natural convection on fRe and Nusselt number for tubes with
0.2. As can be seen, for a given Rayleigh number, the quantities / R e /
fins of dimensionless fin height h = 0.5 and different number of fins
(/Re)o and Nu/Nuo are not very strongly dependent on the number
of fins. Figure 9 shows the results for longer fins, with h = 0.5. The the fin out of a total amount of heat Qt, then the ratio Qf/Qt represents
overall behavior is the same as in Fig. 8. The curve for N = 6 showing the relative contribution made by the fin. Further, if the separate
the variation of /Re/(/Re)o, however, indicates a faster rise of/Re with areas of the fin and the tube plus fins, Af and At respectively, are
the Rayleigh number. The combination of h = 0.5 and N = 6 is such taken into account, (QflAf)/(Qt/At) stands for the fin effectiveness
that there is substantial heat transfer in the circumferential direction. on an equal area basis. Table 1 shows these quantities for the situation
This leads to greater nonuniformity of temperature and consequent in the absence of natural convection. The subscript 0 has been added
stronger buoyancy forces. An increase in N and a decrease in h both to emphasize that the quantities are for zero Rayleigh number. The
tend to make the heat transfer predominantly radial. Further per- first set shows that, for a given fin height, increasing the number of
spective is added by the curves marked N = 0 in Figs. 8 and 9, which fins reduces the fin effectiveness. The second set shows that, for a
represent a finless tube. specified number of fins, shorter fins are more effective than longer
In summary, Figs. 6-9 illustrate that the effect of natural convection ones.
is to increase both the /Re and the Nusselt number Nu. At any given Figure 10 shows the effect of natural convection on fin effectiveness
Rayleigh number, the ratio /Re/(/Re)o is greater than the ratio Nu/ for all combinations of fin height and number of fins given in Table
Nuo- The effect of natural convection, seen in terms of /Re/(/Re)o and 1. As can be seen, the fin effectiveness is considerably reduced in the
Nu/Nuo, seems to be more pronounced on tubes with a small number presence of natural convection. For a given Rayleigh number and fin
of fins and small fin heights. Another noteworthy characteristic is that height, the smaller the number of fins, the greater is the reduction in
the influence of the Rayleigh number on the heat transfer perfor- the fin effectiveness expressed in terms of (Qf/Qt)/(Qf/Qt)o- For a
mance for both finned and finless tubes is about the same; however, given Rayleigh number and number of fins, the effect of fin height is
the frictional resistance of the finless tube increases more rapidly than not very strong, but the general trend is that fins with greater fin
that of the finned tubes. In this sense, a finned tube becomes even height have a greater reduction in effectiveness. These trends can be
more attractive as an augmentation device when the natural con- understood from the fact that the buoyancy forces accelerate the flow
vection effects are significant. near the tube wall, while retarding the flow near the tube axis. Thus,
Of the total heat input to the tube, a part is carried by the fin and the tube wall carries a greater share of the total heat transfer as the
the remainder by the tube wall. If Q/ is the amount of heat carried by Rayleigh number increases.

570 / VOL. 103, AUGUST 1981 Transactions of the ASME

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 Tab] e 1 F i n effectiveness at zero Rayleigh number
(no n a t u r a l c o n v e c t i o n )

N h (Qf/Qt)0 Af,At ^

6 0.5 0.7183 0.4884 1.4708

12 0.5 0.9077 0.6563 1.3832
16 0.5 0.9471 0.7179 1.3192
24 0.5 0.9768 0.7925 1.2326
16 0.2 0.7671 0.5045 1.5205
16 0.3 0.9053 0.6043 1.4981
16 0.4 0.9494 0.6707 1.4156
16 0.5 0.9471 0.7179 1.3192

Concluding Remarks
0.8
The equations governing fully developed laminar combined free 10 100 1000 10000
Ra
and forced convection through vertical internally finned circular tubes
have been solved using a finite difference technique. For all values Fig. 10 Effect of natural convection on fin effectiveness
of fin height, number of fins, and Rayleigh number, the Nusselt
number and the friction factor are found to increase significantly with aJVJ •• brfj+t + Cj(j>j-i + gjQj + fj (A-2)
the Rayleigh number. The effect of natural convection is particularly
pronounced when the number of fins is small and when the fins are for j = 2, M, internal points
short. For each geometry, there exists a critical value of the Rayleigh where it is assumed that there are M + 1 grid points along the line.
number beyond which a reverse flow would occur at the tube center- These equations have been written such that the boundary values of
line. The critical Rayleigh numbers for the finned tubes were found Q and 4> at j = 1 and j = M + 1 do not appear explicitly. In other
to be greater than the critical Rayleigh number for a finless tube. This words,
means that, for a given mass flow rate and heat input, the formation
;
of recirculating cells is less likely with a finned tube than with a finless 0 and bfj = 0 (A-3)
tube. Also, since the friction of the finless tube is more strongly en-
hanced by buoyancy, the finned tubes appear as more attractive This can be achieved in the following manner. If a boundary value is
augmentation devices when the influence of natural convection is known, it can be absorbed in the constant term ey or fj. If the
significant. boundary value is not known, the corresponding boundary condition
(such as zero gradient) can be used to eliminate the boundary value
from the finite difference equations for the internal points. The first
Acknowledgment
step in the algorithm is to recast equations (A-l) and (A-2) as
This research was performed under the auspices of NSF Grant
ENG 75-18141 A01. fiy=PyQy+i + fly0y+1 + Qy (A-4)

References <j>j = Pj-fij+i + Tyfly+i + Uj (A-5)

1 Bergles, A. E., "Survey and Evaluation of Techniques to Augment When the above equations are written for j' — 1 and used to substitute
Conveetive Heat and Mass Transfer," Progress in Heat and Mass Transfer,
Vol. 1, Pergamon Press, Elmsford, N. Y., 1969, pp. 331-424. for Oy-! and </>y_i into equations (A-l) and (A-2), the result is
2 Hu, M. H., and Chang, Y. P., "Optimization of Finned Tubes for Heat
Transfer in Laminar Tube Flow," ASME JOURNAL OF HEAT TRANSFER, Vol. (ay - cj Pj-i)Qj - (cjRj-! + dj)<j>j = bj fiy+i + (ey + cyQy-i)
95,1975, pp. 332-338.
3 Nandakumar, K., and Masliyah, J. H., "Fully Developed Viscous Flow (ay - cj Pj-iHj - (CJTJ-I + gj)Qj = bj qbj+i + (fj + CJUJ-I)
in Internally Finned Tubes," Chemical Engineering Journal, Vol. 10, 1975,
pp. 113-120. These two simultaneous equations can be solved forflyand <j>j in terms
4 Masiliyah, J. H., and Nandakumar, K., "Heat Transfer in Internally of Qy+i and 4>j+\. Comparison of the expressions so obtained with
Finned Tubes," ASME JOURNAL OF HEAT TRANSFER, Vol. 98, 1976, pp. equations (A-4) and (A-5) leads to the recurrence relations for Pj, Rj,
257-261. Qj, Tj and Uj. These are given as follows:
5 Morton, B. R., "Laminar Convection in Uniformly Heated Pipes," Let
Journal of Fluid Mechanics, Vol. 12,1959, pp. 227-240.
6 Tao, L. N., "Heat Transfer of Combined Free and Forced Convection
in Circular and Sector Tubes," Applied Scientific Research (A), Vol. 9,1960, aj = aj - CjPj-i (A-6)
pp. 357-368.
7 Kemeny, G. A., and Somers, E. V., "Combined Free and Forced-Con- 13j = dj + cjRj-i (A-7)
vective Flow in Vertical Circular Tubes—Experiments with Water and Oil,"
ASME JOURNAL OF HEAT TRANSFER, Vol. 84,1962, pp. 339-346. Vj = Si + cyTy-j (A-8)
8 Patankar, S. V., Numerical Heat Transfer and Fluids Flow, McGraw- 2
Hill—Hemisphere, New York, 1980. & = («y -ftiy) (A-9)
9 Deshpande, M. D., and Giddens, D. P., "Direct Solution of Two Linear
Systems of Equations Forming Coupled Tridiagonal-Type Matrices," Inter- then
national Journal of Numerical Methods in Engineering, Vol. 11, 1977, pp.
1049-1052. Pj = bjCtj/tj (A-10)
10 Settari, A., and Aziz, K., "A Generalization of the Additive Correction
Methods for the Iterative Solution of Matrix Equations," SIAM Journal of Rj = bjfljltj (A-ll)
Numerical Analysis, Vol. 10,1973, pp. 506-521.
Qj = (aj(ej + cyQy-i) + ft(/y + CjUj. - i ) ) / & (A-12)

Tj = bjVj/Hj (A-13)
APPENDIX A Uj = (Vj(ej + CJQJ-I) + aj(fj + CjUj- -I))/?,' (A-14)

Simultaneous Solution for T w o Coupled Variables Since C2 = 0, the values of Pi, R2, Q2, etc. can be calculated without
Here the task is to solve the following equations: knowing P i , Ri, Qi, etc. Then the recurrence relations can be used for
j = 3, 4 . . . M. Also, since btd is zero, PM, RM and TM would turn out
ayfiy = bjQj+i + cyfiy-i + dj4>j + ej (A-l) to be zero. Hence Q.M = QM and 4>M = UM-

Journal of Heat Transfer AUGUST 1981, VOL. 103 / 571

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 S u m m a r y of t h e A l g o r i t h m where
i Calculate P,, Qj, Rj, Tj and Uj for j = 2 to M using equations
(A-6) through (A-14).
Vj = E (au - Pa -" Qij) (B-5)
ii Set QM — QM and 4>M = UM- 1=2
iii Back-substitute into equations (A-4) and (A-5) to obtain Qj N
and (j>j for; = M - 1, M - 2 , . . . , 4, 3, 2. Xj = E bij (B-6)
i=2

APPENDIX B N
Wj = E dj (B-7)
;=2
Block-Correction Procedure N
The equations for the block-correction procedure are derived from Yj = E du (B-8)
i=2
the basic finite difference equations (27) and (28) for the Q and 4> N
fields. Further, it is assumed that these equations have been modified Mj = E Bij (B-9)
such that the boundary values do not appear in them. It is proposed 1=2
that
Zj = E (PijQ*i+tj + qijSi*i-ij + bij J2* lV+1
t=2
««y i« ij i «« j (B-l)
+ cij Q*ij-i - aij U*ij + dij <t>*ij + rij) (B-10)
(B-2)

for all internal grid points.

Here Q*ij and 4>*ij are the current values of Qij and <f>ij, and Q'j andNj = i=2E (Pij <l>*i+ij + qij <t>*i-ij + bij <t>*iJ+1
<t>'j are the proposed block corrections. Equations (B-l) and (B-2) are
first substituted into equations (27) and (28), and then the resulting + Cij <P*ij-i - dij 4>*ij + gij Q*y) (B-ll)
equations are summed over all internal values of i, i.e., from i = 2 to
i = N; the result is: It is interesting to note that the expressions for Zj and Nj represent
the summations of the residuals of equations (27) and (28) respec-
Vjil'j = X}Q'j+1 + WjQ'j-i + Yj<t>'j + Zj (B-3) tively; when these residuals are zero, no block correction is obviously

Vj4>'j = X ^ ' j + i + Wj4>'j-i + MjQ'j + Nj (B-4) needed.

The coupled equations (B-3) and (B-4) are solved for the values of
for; = 2 t o M Wj and 4>'j by the use of the algorithm given in Appendix A.

572 / VOL. 103, AUGUST 1981 Transactions of the ASME

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 Krishna P. Singh Some Fundamental Relationships for
Vice President—Engineering ^ ^
j
Tarc:ro^ Tubular Heat Exchanger Thermal
Mem. ASME %#

Performance
Some basic relationships to characterize tubular heat exchanger thermal performance
are derived in terms of the well-known state variables. It is shown that the knowledge of
T) (NTU), R (heat capacity rate ratio), and partial derivatives of the temperature efficien-
cy P with respect to r\ and R enables complete characterization of the exchanger perfor-
mance around an operating point. Thus, the exchanger performance can be readily pre-
dicted for the so-called "subdesign" conditions. Likewise, additional criteria to compare
various exchanger styles for a given range of operating conditions can be developed. Two
such criteria are developed in this paper.

1 Introduction
Heat exchangers are designed to deliver a certain heat transfer rate actual heat duty to the thermodynamically maximum possible heat
for a certain specified condition of flow rates and temperatures. This duty, i.e.
specified condition is often labeled as the "design point" for the heat
exchanger [1]. However, there are few exchangers which are confined Q
e (1)
to operating at the design point only. Conditions of service result in o
variation in the input quantities to the exchanger (namely, flow rates
where
and inlet temperatures) resulting in quite different outlet tempera-
tures and heat duty. Furthermore, the heat exchanger operator may Q=Ms(T1-T2)=Mi(t2-t1) (2)
seek to set the input quantities to obtain a desired output. Thus, the
"subdesign" characteristics of heat exchangers are undoubtedly of and
great value to the operator and the system engineer.
Qmax = M m i n ( T i - ti) (3)
A similar motivation for quantifying subdesign characteristics arises
where
when selecting the exchanger style [2] during the design stage. If an
exchanger is designated to perform under more than one set of con- M m i „ = Min (Ms, Mt) (4)
ditions, then it is obviously meaningful to compare the candidate
designs for their subdesign characteristics rather than focusing the c can be conveniently expressed in terms of commonly used dimen-
attention exclusively on the design point. In order to make a mean- sional parameters; namely temperature efficiency P, and heat capacity
ingful comparison between various exchanger geometries "around" rate ratio R, defined as follows:
a design point, the concept of the "exchanger effectiveness gradient
vector" and "LMTD correction factor gradient vector" are proposed. P =^ ^ (5)
Their practical significance is illustrated by comparing some com- Tx - h
monly used designs. Mt 7 \ - T2
R= — = — (6)
In contrast to other process equipment such as pumps and turbines, Ms t2 - ti
operating curves for heat exchangers are seldom generated. This is
It can be shown using the foregoing equations, that
presumably due to the large number of variables involved (two flow
rates, four terminal temperatures, heat transfer coefficients and heat e = P; R <1
transfer surface). However, it is possible to derive explicit relation- = RP; R>1 (7)
ships to predict exchanger performance when any or some of the
principal variables are changed. Such expressions are obviously of Equation (7) leads to some interesting corollaries: (a) Since e must
value to the user: they also help the designer evaluate a host of oper- be less than or equal to 1, it follows that P must always be less than
ating conditions which ordinarily necessitate use of a computer pro- the lesser of 1 and 1/R. (b) The temperature efficiency P is always less
gram. than or equal to the effectiveness, e.
The expression for the exchanger effectiveness gradient vector is Another important dimensionless variable that occurs in the for-
derived in the next section, followed by the development of expres- mulations of heat exchanger performance analysis is the so-called
sions for predicting performance under subdesign conditions in number of transfer units (NTU) or r),2 defined as
Section 3. The concept of the LMTD correction factor gradient vector
is introduced in Section 4. These concepts are further illustrated in
Section 5, followed by a brief conclusion in Section 6.
In fact, it can be shown that r] and R can be treated as two independent
2 Exchanger Effectiveness variables that completely characterize the thermal performance of
There are several definitions of efficiency associated with heat a heat exchanger; i.e.
exchanger performance. Of these, exchanger heat transfer effective-
ness 1 as defined by Kays and London [3] is perhaps most meaningful. P=f(v,R) ©a)
The effectiveness of a heat exchanger is defined as the ratio of the

1
Also referred to as "temperature efficiency" in the literature.
2
Contributed by the Heat Transfer Division for publication in the JOURNAL Some texts [3] define NTU as UA/Mmin\ where Mmjn is lesser of Mt and Ms.
OP HEAT TRANSFER. Manuscript received by the Heat Transfer Division In this paper the definition implied by equation (8) is used throughout the
October 23,1980. text.

Journal of Heat Transfer AUGUST 1981, VOL 103 / 573

Copyright © 1981 by ASME
Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 It thus follows from equation (7) that e can also be considered to be variables. We note that a and b are readily computed if the derivatives
a function of?) and R; i.e. of P with respect to TJ and R are known.
Before proceeding to illustrate the usefulness of the above rela-
£ = h(i),fl) (9b) tionships, we will consider similar concepts for another measure of
Thus exchanger performance: the so-called LMTD temperature correction
factor.
„ dh dh ,
d€ = dh = — dri A dR
i>7] dR 3 Temperature Correction Factor
Whereas e is a measure of the exchanger effectiveness corresponding
dt = dP = — dn + — dR; R<1 (10) to thermodynamically maximum attainable heat transfer rate, the
dri dR
LMTD temperature correction factor F, is the gage of exchanger
de = PdR + RdP performance vis-a-vis a countercurrent heat exchanger. It is defined
IdP dP as the ratio of the heat transfer surface area required in a counter-
= PdR + R\— dij + — dR current heat exchanger to that in the subject heat exchanger for a
Id?) dR
given set of heat duty and flow conditions. It is expressed in terms of
I dP\ dP the foregoing parameters rj and R as follows
o r d e = P + fi—\dR + R — dV; R>1 (11)
\ dRj dr]
Equations (10) and (11) define the incremental change in £ as a P =- ln- (15)
ri(R - 1) "" 1 - PR
function of partial derivatives of P with respect to ?) and R. We
have P appears explicitly in equation (15). However, F can also be viewed
as an implicit function of rj and R by virtue of equation (9a).
de= a dR* + b dri* (12) Derivatives of F with respect to 77, P , and R are directly obtained
from equation (15). We have
where
dP dF 1 . 1-P -F
R—, R<1 (16)
dR d-q HR- l) 1 -PR f)
(a) dF 1
(17)
R \P + R- R>1 dP 7,(1 -P)d -PR)
dR,
dF F P
dP 4 (18)
j) — , R < 1 dR (R v(R~ D ( l - PR)
dr; -l)
b =< (b)
dP
r)R — ; R > 1 (13) dF dF dF
dr} dF = — dP + — d« + — dR (19)
dP dri dR
and Substituting for dP from equation (10), we have
dR IdFdP bF\ IdFdP dF\ ,„
dR* (c) dF = + — d?)+ + —IdR
R \dPdri dy \dPdR dR

dF' + A\dr) + [B— + CldR (20)

(d) I dri
n
where
Thus dR* and dri* are fractional changes in R and TJ, respectively.
Hence, we have dF dF
(A,B,C)
dr)' dP' dR}
de = Ve. n
Thus
where
dP dP
dF = r) [IB — + A\dr]* + R\B — +C\dR* (21)
(n) = (dr)*,dR*) and dV j \ dR I
Vf is the gradient of e surface in the coordinate system consisting of
or dF = VP. n (22)
dR* and dri*. The magnitude of this gradient vector is given by
VP is the gradient of P vector in the coordinate space consisting of
X = |Vf| = (a 2 + i 2 ) 1 ' 2 (14)
?,* and R *. The norm of this vector can be used as a direct measure
As discussed later, x c a n be used as a direct measure of the pertur- of the susceptibility of the exchanger performance due to perturba-
bation in exchanger performance due to small variations in the input tions in the input data. It is to be noted that computation of the norm

•.Nomenclature

A = total heat transfer surface area, m 2 P = temperature effectiveness, dimension- tures, respectively, K
a = coefficient of dR* less U = overall heat transfer coefficient,
b = coefficient of d7,* Q = total heat duty, W W/m 2 K
Qmax = thermodynamic maximum heat duty, 6 = exchanger heat transfer effectiveness
P = LMTD temperature correction factor, W
dimensionless (equation (1)), dimensionless
R = heat capacity rate ratio (equation (6)),
Ms = shellside heat capacity rate (mass flow r\ = number of heat transfer units (NTU)
dimensionless
rate times the specific heat of shellside (equation (8)), dimensionless
P i , P2 = shellside inlet and outlet tempera-
fluid), W/K tures, respectively, K X = norm of exchanger effectiveness gradient
ti, £2 = tubeside inlet and outlet tempera- vector
Mt = tubeside heat capacity rate, W/K

574 / VOL. 103, AUGUST 1981 Transactions of the ASME

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 of VF requires evaluating dP/d?j and dP/dR as was the case for Ve (c) Two Shell Pass—Four Tube Pass. The functional rela-
(equation (13)). tionship is [4]
The functional relationship for most heat exchanger configurations
r
have the general form | - 1 - R + - [(1 - P ) ( l - PR)]i/ 2 + R;
•n = — In (35)
V = <t>(P,R) (23) R'
- - 1 - R + - [(1 - P ) ( l - Pfl)]l/2 _ R<
where </> is an explicit function in P and R. dri/dP (and hence dP/drj)
can be obtained directly from equation (23) by differentiation. dP/dR In this case evaluation of the partial derivatives is quite tedius. It is
is computed by observing that preferable to obtain the derivatives numerically on a digital computer.,

\l/(n,P,R) = t)-<t>(P,R) = 0 (24)

M_ _ 7?(P+AP) ~ V(P)
(o)
Treating \p to be a function of t\ and R directly, and also explicitly dP AP-O AP
through P , we have dy__ ,. THR+AR) — V(R)
(b) (36)
dtp d\p dP dR Afl—o Afl
- + -1-— = 0 (25)
dR d P dR Having determined drj/dP and dr]/dR; equation (26) can now be uti-
Noting that lized to evaluate dP/dR.
Prior to illustrating the application of the concepts of effectiveness
d\p —d0 d\p — d<f> gradient and LMTD gradient vectors, we will derive general expres-
dR " dR' dP dP sions for predicting subdesign performance of tubular exchangers.
we have These expressions are also shown to require evaluation of the same
set of partial derivatives, namely dP/drj and dP/dR.
-d0 d<t>dP_
dR d P dP 4
Subdesign Relationships
First of all, we observe that P is independent of temperatures
dP - d 0 / d 0 (equation (9)). It thus follows that P will remain invariant with respect
l\ (26) to changes in the inlet or outlet temperatures. 3 Thus, it should be
possible to determine exchanger performance for subdesign cases
d<t>/dR and d $ / d P are directly obtained from equation (23). In this involving changes in temperature data if the value of P for the design
manner, the expression for dP/dR is obtained. point is given. Explicit expressions for the four possible cases are
We will now examine some common heat exchanger geometries. readily obtained as follows:
(a) Countercurrent Heat Exchanger. The functional rela- (a) Input temperatures Ti and ti varied {to new values, say 7 \ *
tionship is given by [4] oradti*): 4
1-P Noting that
ln-
1-PR tz-ti t2* - U*
(27)
v- fl-1 Ti - ti T i * - t i *
We have, by direct differentiation
We have
dri/dP = (a) t2* = P ( 7 V - ti*) + ti* (37)
(1 - P ) ( l - PR)
evW-»[v(R - 1) + 1 - e'f*- 1 '] Heat balance yields
dP/dR (b) (28)
[l-fle"<«-»]2
R(ti* - ii*) = Ti* - T2* (38)
The norms of Ve and VP can now be directly obtained by substituting
above relationships into equations (13) and (20), respectively. This, together with equation (37) yield
(b) One Shell Pass—Two Tube Pass Heat Exchangers. The
To* = P i * -PR(T\* ~t\*) (39)
functional relationship is well-known [4].
(6) Hot fluid inlet T\ and cold fluid outlet t2 varied {to new values
1 2 - P[R + 1 - (1 + R2)1'2]
In (29) T\* and t2*, respectively).
v- (1 + fl2)1'2 .2 - P[R + 1 + {R2 + 1) 1/2 ] Equation (37) gives
This expression can be recast to give P as a function of r/ and R P T i * - ti*
t,*=- (40)
2[1 - e"K'] P - l
(30)
R+ 1 - R' - e»R'(l + R + R') To* is expressed in terms of T\* and to* by using equation (38) as
where follows:

R' = (1 + R2)1'2 (31) To* = — ^ — [(P - 1 + PR)TX* - PRto*} (41)

The derivatives dP/drj and dP/dR are obtained by direct differen-
tiation. We have, after some manipulation (c) Hot fluid outlet T2 and cold fluid inlet t\ varied: Similar
analysis is performed to give the following relationship for T{* and
2
dP/dj; = 1 - PR - P + 0.5P R (32) to*:
dP
—- = [2i)Re»fl' + P|fl'(l - e»R')
dR YR
- R{1 + e»R') - r)Re*R'{R + 1 + R')}] (33) 3 It is presumed that the coefficient U is not affected by temperature changes.
For most practical cases, this assumption can be adopted without significant
where error as long as temperature changes are small.
4
Subdesign temperatures are indicated with a * superscript in this sec-
Y = R+1-R'- e"R'(R + 1 + R') (34) tion.

Journal of Heat Transfer AUGUST 1981, VOL. 103 / 575

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 insight into the stability and sensitivity characteristics of a heat ex-
Ti* = l-PRtx* + T2*\ (42) changer at the operating point under consideration. Tube pluggings
a-PR) and tube surface fouling are a fact of life in heat exchanger operation.
t2
*=7; —[PT2*-ti*{PR + P-l)] (43) A pertinent question is: How will the heat duty hold up in face of a
(1 - PR) declining U (fouling) or decreasing surface area (plugged tubes)? The
(d) Both outlet temperatures varied: Given the new values of T2* magnitudes of Ve and VP provide direct answers to this question. To
and ti*, the associated values of T\* a n d t i * are obtained by a similar verify this fact we refer to Pig. 1 which gives x ( = I Ve|) and e as a
procedure. The final results are: function of r\ for countercurrent (curve A) and one shell pass-two tube
pass (curve B) designs for a specific value of R. We note that x for
h*RP - (1 - P)T2* "one-two" design is smaller than that for the countercurrent design
T1* = (44)
P + PR - 1 for the same value of e. For example, corresponding to P = e = 0.5, XA
PT2* + PRt2* - *2* = .312, XB - -285. This indicates that the effectiveness for the one-two
h* (45) design will vary less rapidly with respect to changes in input data than
P + PR - 1
the countercurrent design. Table 1 gives numerical values of the de-
Equations (37,39-45) give the eight required relationships to compute rivatives to enable a direct comparison. By way of comparison, let us
two unknown terminal temperatures for each of the four possible cases compute the decline in the exchanger effectiveness for a 1 percent
of combinations of specified temperatures. reduction in U (and hence ?/) in a countercurrent heat exchanger
(e) Flow rates varied: As opposed to "modest" variation in terminal operating at P = e = 0.5 (R = 0.5). Table 1 yields (by interpola-
temperatures, changes in flow rates, and surface area directly affect tion)
the values of r\ and/or R (and also ij indirectly through possible
changes in U). Thus flow rate changes alter the heat exchanger per- de/d?j = 0.375
formance variables in an explicit manner. It is important to know how
the exchanger will respond to changes in flow rates within the ex- Equation (10) gives
pected range of variation. Using Taylor series [5], assuming P to be
continuous in r\ and R, we have de = 7) — drj* = -(0.8)(0.375)(0.01) = -0.0030

*>P „ dP „ „ A one-two exchanger operating at identical conditions (P = 0.5, 7) =

P(ri*,R*) =P(i),R) + — Ari + — Afl
0.867, R = 0.5), subject to 1 percent reduction in n has (by interpola-
1 d2P tion, Table 1)
+ - • Ai) 2 + 2 A-qAR
2 drj' dridR de
— = 0.278
d P 2 dr)
+ — - AR2\ + Terms of higher order (46)
ZA-3J2-
For moderate changes in ?j and R (AR and A?) small), sufficiently
accurate value of P (ij*, R*) (referred hence forth as P*) can be ob-
tained by using only the first order derivatives. Thus

dP dP
P* =P + — A?j + — AR (47)
drj dR

The partial derivatives dP/di? and dP/dfl are evaluated in accordance

with the procedure described in the preceding section. Thus, P* is
LEGEND
computed. Having determined P*, the two unknown terminal tem-
[ A ] COUNTERCURRENT FLOW
peratures can be determined using the appropriate set of equations [B] I SHELL PASS 2 TUBE PASS
(out of equations (37-45)).

5 Effectiveness and Correction Factor Gradient

Vectors
It was shown in the foregoing that computation of the partial de- 1 —
rivatives of P with respect to r\ and R directly leads to the determi-
nation of Ve and VP. Evaluation of Ve and VP provide important Fig. 1 £ and x versus ij (R = 0.5)

Table 1 D e r i v a t i v e s of P w i t h r e s p e c t to ij a n d -Rt, c o u n t e r c u r r e n t a n d o n e - t w o h e a t e x c h a n g e r s ( i ? = 0.5)

Countercurrent Heat One Shell Pass--Two Tube
Exchanger Pass Heat Exchanger
de de de de
NTU, v P P
drj dfi d?; dfl

0.1 0.093 0.865 -0.425 E-2 0.093 0.863 -0.44 E-2

0.2 0.174 0.754 -0.146 E-l 0.173 0.748 -0.155 E-l
0.4 0.307 0.587 -0.440 E-l 0.304 0.567 -0.492 E-l
0.6 0.412 0.467 -0.763 E-l 0.403 0.436 -0.894 E-l
0.8 0.496 0.379 -0.107 0.480 0.337 -0.130
1.0 0.565 0.312 -0.133 0.540 0.263 -0.169
1.1 0.595 0.285 -0.145 0.565 0.233 -0.187
1.2 0.622 0.260 -0.155 0.587 0.206 -0.205
1.4 0.670 0.220 -0.173 0.623 0.162 -0.236
1.6 0.710 Q.187 -0.186 0.652 0.128 -0.264
1.8 0.745 0.160 -0.196 0.675 0.101 -0.288
dP
* Note: — = — — (from eq uation (10)) for R < 1.0.
bri d-q dR dR

576 / VOL. 103, AUGUST 1981 Transactions of the ASME

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 Hence value of F. Figures 2 and 3 show VF plotted against F for two different
heat exchanger geometries. We note that VF -* 0 as F -» 1 for both
de = -(0.867)(0.278)(0.01) = -0.0024 designs. This is seen to be true for all heat exchanger configurations.
VF is thus a second measure of deviation from countercurrency. A
Therefore, the loss of effectiveness in the one-two design is 80 sharp decrease in F coincides with a sharp increase in VF (Fig. 4).
percent of that in a countercurrent design when both designs are If two exchanger thermal geometries possess identical values of F
subject to 1 percent reduction in the NTU. Thus the one-two design at the design point, then VF provides the basis for judging their rel-
may be considered to be more stable and less sensitive to perturba- ative merits for subdesign conditions. In this sense, VF provides a tool
tions in the input data than an identical service countercurrent heat for sensitivity studies in the same manner as Ve. Figure 5 shows the
exchanger. coefficients of VF vector as a function of P. We note that both de-
To demonstrate another variation of the same concept, let us in- rivatives are negative and both increase with P (as F decreases, refer
vestigate the increase in the shellside flow rate required to nullify the to Fig. 4), although dF/dR has a well defined maximum point. These
loss in heat duty due to a given decrease in NTU. Let us consider the curves point to the plausible conclusion that the LMTD correction
following operating condition: factor in inefficient designs respond more sharply to perturbation in
input data. Thus if the heat transfer surface area in a "low F" is re-
R = 0.5, P = 0.565, Ar; = -0.05 duced, the attendant increase is F will be greater than that in a "high
The data for the countercurrent design is found from Table 1: F " unit subject to a similar surface area reduction. On the other hand,
the decrease in F due to input data change (such as decreased shellside
de de flow rate) is also liable to be greater in a low F geometry compared to
n = 1.0, — = 0.312, — = -0.133
dr; dR a high F one. The conditions for each heat exchanger are unique to
From equation (10) its function. Therefore, the considerations in regard to subdesign
performance will dictate as to how to interpret the coefficients of the
de , de LMTD gradient vector, or the effectiveness gradient vector. Our ob-
de = Q = — dr) + — dR
dr] dR jective herein is merely to formulate their definition and illustrate
their possible use. Perhaps not all designs require their evaluation;
Hence
nonetheless their knowledge certainly yields additional insights into
the hardware characteristics.

6 Conclusions
^1 Expressions to evaluate the performance of a heat exchanger under
dR
subdesign operating conditions have been derived. It has been shown
that the partial derivatives dP/drj and dP/dR are central to the
Substituting numbers we have

dR\cc = -0.1173
Hence

R* = R + dR = 0.3827
Percentage increase in Ms required

= I — - 1 (100) = 30.65 percent

Similar procedure for one-two design yields

de de
i\ = 1.1, — = 0.233; — = 0.187
dr; dR

dfl|i-2 = - 0 . 0 6 2 3
R* = 0.5 - 0.0623 = 0.4377

Thus, percentage increase in Ms required is computed to be 14.23

percent.
Fig. 2 F versus V F for 1 shell pass-2 tube pass heat exchanger
This comparison clearly shows the superiority of one-two design
over its countercurrent counterpart from the standpoint of stability
of performance.
Questions like the ones raised above can be readily answered if Ve
is known, which merely implies computing dP/dr) and dP/dR. i.o-
Evaluation of these derivatives is a simple matter, especially since
most design work is performed on digital computers.
The above example mathematically proves what is intuitively ob- 9-

vious. The decline in performance of a countercurrent heat exchanger

due to deterioration in U, reduction in A or flow rates is steeper than
.8-
it is in a one-two exchanger. The magnitudes of x and the components
of Ve mathematically confirm this intuitive result. Ve is a simple, yet
definitive measure of relative sensitivities between two non-coun- .7-
tercurrent designs. It is also a convenient tool to predict subdesign
performance of operating equipment.
Knowledge of VF provides understanding of the equipment re- 0 .I .2 .3 .4 .5 .6 .7 .8
sponse from another vantage point. VF for a countercurrent heat —VF
exchanger is by definition zero. Large values of VF correspond to small Fig. 3 V F versus F; two shell pass-4 tube pass (R = 0.S)

Journal of Heat Transfer AUGUST 1981, VOL. 103 / 577

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 ^TEMR EFFICIENCY P
. - TEMP EFFICIENCY P
Fig. 5 Derivatives of F versus P (two shell pass-4 tube pass H.X.)
Fig. 4 F and | V F | versus P (two shell pass-4 tube pass H.X.)

characterization of the exchanger response in the vicinity of an op- References

erating point. The effectiveness gradient vector or the temperature 1 Standard For Closed Feedwater Heaters, Heat Exchange Institute, Third
Edition, 1979.
correction factor gradient vector can be readily evaluated if dP/drj 2 Standard for Tubular Exchanger Manufacturers Association, Sixth
and dP/dR are known. The components of these vectors (in the r\ — Edition, New York, 1978, p. 5.
R plane) and their norms yield valuable insight into the exchanger 3 Kays, W. M., and London, A. L., Compact Heat Exchangers, McGraw-
characteristics around the design point. Comparison of these quan- Hill, New York, 1964.
4 Kern, D. Q., Process Heat Transfer, McGraw-Hill, New York, 1950.
tities may be useful in selecting the most suitable candidate design
5 Hildebrand, F. B., Advanced Calculus for Applications,Prentice Hall,
for a given set of operating data. New York, 1962, p. 348.

578 / VOL. 103, AUGUST 1981 Transactions of the ASME

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 A General Theory of Wet Surface
I. L. Maclaine-cross
Broken Hill Division,
Heat Exchangers and its Application
University ol New South Wales,
Broken HIM, N.S.W. 2880, Australia
to Regenerate Efaporative Cooling
P. J. Banks
Wet surface heat exchangers such as cooling towers and wet plate heat exchangers are im-
Division of Mechanical Engineering,
Commonwealth Scientific and Industrial Research portant in air conditioning. A linear approximate model of wet surface heat exchangers
Organization, is proposed. The equations of the model are rearranged, enabling solutions for wet bulb
Highett, Victoria 3190, Australia depression and wet bulb temperature to be obtained independently, by analogy from pub-
lished solutions for dry bulb temperature in dry surface heat exchangers. Performance
predictions by this method for a crossflow cooling tower are found to agree with those from
a prevous finite difference solution. Published performance measurements for a crossflow
wet plate heat exchanger are lower than predicted by the method possibly due to poor
wetting or excessive water flow. Excellent performance is predicted for a proposed regen-
erative evaporative cooler using such an exchanger.

1 Introduction
In wet surface heat exchangers, heat is transferred between a fluid of the processes occurring. If results are required for a different design
stream and a moist air stream through a water film on the moist air at another place or time, another computer program must usually be
side of the heat transfer surface. In air conditioning, simple evapo- written, which can be laborious.
rative coolers, water cooling towers, finned coils with condensation The theory of wet surface heat exchangers presented here is in-
and sprayed coil or evaporative condensers are commercially impor- tended to improve understanding and simplify calculations. A linear
tant types of wet surface heat exchanger [1]. The fluid stream is either approximation to the saturation line, together with assumptions like
water or a primary or secondary refrigerant in these examples. those of Mickley [6] mentioned above, are used to give linear partial
Air-to-air plate heat exchangers with the surfaces of one set of differential equations (Section 2). These equations are rearranged,
passages wet are used as indirect and regenerative evaporative coolers. enabling solutions for wet bulb depression and wet bulb temperature
Regenerative evaporative coolers offer lower energy consumption than to be obtained independently, by analogy from published solutions
vapor compression or closed cycle coolers and better performance and [8] for dry bulb temperature in dry surface heat exchangers (Section
lower water consumption than simple evaporative coolers. These were 3). This approximate theory is found to agree with Baker and
sold in the south western United States during the 1930's [2]. Later Shryock's [7] finite difference solutions for a crossflow cooling tower
they did not compete with vapor compression or simple evaporative (Section 4).
coolers. The large size of the metal plate heat exchangers was an im- Predictions of the efficiencies of Pescod's plate heat exchangers
portant disadvantage. with wet passages, using dry surface results [4, 5] and the theory
Pescod [3] has more recently developed compact crossflow plastic presented here, are found to be significantly greater than reported
plate heat exchanger designs. His pin fin surfaces have small diameter measurements [5] (Section 5). Poor wetting of Pescod's plastic plates
pins at large spacings to minimize pressure drag. At high laminar flow is suspected. The wettability of parallel plates of woven cloth has been
Reynolds numbers, he measured pressure drops and heat transfer found excellent in evaporative coolers [9,10]. The performance of a
coefficients over three times those for parallel plates [4, 5]. The ratio proposed regenerative evaporative cooler (Section 6) using a plate heat
of measured pressure drop to heat transfer coefficient was similar to exchanger with a Pescod surface for the dry passages and woven cloth
that for parallel plates. Pescod [5] has also reported measurements parallel plates for the wet passages is predicted (Section 7). Detailed
of effectiveness for his plate heat exchangers with the surfaces of one calculations are given in Section 8 to illustrate the use of Sections 2
set of passages wet. and 3. The predicted performance compares favorably with that of
Mickley [6] recognized that all wet surface heat exchangers could vapor compression and simple evaporative coolers.
be represented by a single mathematical model which he solved using All the authors' calculations in Sections 3, 4, 5, 7 and 8 were made
a partly graphical method. Mickley assumed that the Lewis relation using a Hewlett-Packard 97A programmable electronic calculator.
is satisfied which is approximately true for water-air systems with The program for calculating crossflow heat exchanger temperature
good wetting, and that the water mass flow rate and the specific heat efficiency used Baclic's equation [11] with all Bessel functions eval-
of moist air are constant. These assumptions allow the differential uated simultaneously by a recurrence relation [12]. For capacity rate
equations of the model to be expressed and solved in terms of moist ratios less than unity, temperature efficiency equals the effectiveness
air specific enthalpy, which is equivalent for the degree of approxi- defined in [8].
mation used to solution in terms of adiabatic saturation or thermo- The theory of wet surface heat exchangers presented here was
dynamic wet bulb temperature. suggested to the authors by the analogy theory of coupled heat and
Baker and Shryock [7] used an electronic digital computer to solve mass transfer in porous media, which they have applied to regenera-
finite difference equations for a cooling tower with water film heat tive heat and water vapour exchangers [13, 14, 15]. In this analogy
transfer resistance. Like Mickley they assumed that the Lewis relation theory, the coupled heat and mass transfer is described by two com-
is satisfied, but, except in their equations (9,10) for film resistance, bined potentials acting independently, which correspond to wet bulb
did not assume constant water mass flow rate or moist air specific temperature and web bulb depression in a wet surface heat exchanger.
heat. The water film heat transfer resistance for their examples was The possible accuracy of the theory presented here was indicated by
small so that the implied assumptions in their equations (9, 10) af- the close agreement found between performance predictions from the
fected their results only slightly. Computer results such as these can analogy theory in linearized form and from finite difference solutions
be as accurate as the equations solved, but they give no understanding for practical energy recovery regenerators used in air conditioning [14,
15].
Contributed by the Heat Transfer Division for publication in the JOURNAL
OF HEAT TRANSFER. Manuscript received by the Heat Transfer Division 2 A Model of the Wet Surface Heat Exchanger
October 7,1980. Partial differential equations including terms for all significant

Journal of Heat Transfer AUGUST 1981, VOL. 103 / 579

Copyright © 1981 by ASME
Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 effects in a wet surface heat exchanger may be solved using an elec-
tronic digital computer. Solution using a hand calculator and litera-
ture tables or charts requires simplified equations. Therefore the
actual wet surface heat exchanger will be replaced by a mathematical
model satisfying conditions which give a set of linear algebraic and \ A A A A A A / V W V X
partial differential equations to be solved simultaneously. The model ^PERFECT INSULATOR V A / N / V A A
structure is shown schematically in Fig. 1. \ / \ / \ / \ / \ / \ / \ / \ A A A A.
The model satisfies the following conditions, which correspond to AIR t k
those used for heat transfer alone in heat exchangers: ; =•» ta,Wa,Ca,hQ
1 zero wall, air and fluid thermal and moisture diffusivity in the ma
flow directions; r WATER FILM ~ Z ^ " rUTCZ '. —
no heat transfer to the surroundings occurs;
/ W A L L ^ / / /
the passage walls are impervious to mass transfer;
pressures and mass flow rates are constant and uniform for ^777Z
l-HEAT FLOW
both streams;
tf. Cf •<
c/, c„, hc, ho, hg, hfg and U are constant and uniform;
the passage geometry is uniform throughout the heat ex-
changer. A ^ 7 V A A A A / T T y y Y
The model also satisfies the following condition, used previously * PERFECT INSULATOR A / y \ A A
for wet surface heat exchangers [6,7]: ./\/\/\/'\/\/\/\AAAA/<
7 the Lewis relation is satisfied, that is I-f
ho = hca/Ca (1)
Fig. 1 Structure of mathematical model of the wet surface heat ex-
Mickley's assumptions [6] (Section 1) are replaced by the following changer
conditions:
8 the specific enthalpy of moist air ha is a linear function of ta For conservation of energy in the fluid stream
and wa, thus

ha = ha0 + cata + hgWa (2) cfriifLf UA (tu - tf) (5)

dx
where hao is a constant; For conservation of energy and water at the air-water interface
9 the evaporating water film is stationary and continuously re-
plenished at its surface with water at the same temperature; hcaAa(ta - tw) + hDAahfg (wa - ww) + UA (tf -tw) =0 (6)
10 the moisture content ww of the air in equilibrium with the
For conservation of water vapor in the moist air stream
water surface is a linear function of the water surface tem-
perature t,„, so that the model saturation line is given by
maLa = hoAa (ww - wa) (7)
ww = d + etw (3)
For conservation of energy in the moist air stream
The linear equations (4-8) below may be derived from conditions 1
to 10. The approximate psychrometer equation is aAa (tw ~ta) + hDhsAa (ww - Wa) (8)
ay
Wa = Wa' + (ta' - ta)ca/hfg (4)
The boundary conditions satisfied by the model and to be used in
where ' indicates the adiabatic saturation state defined using the solving equations (1-8) are that the inlet fluid temperature and the
linearized or model saturation line of condition 10, equation (3). inlet air temperature and moisture content are known and constant.
.Nomenclature.
= thickness of plate or of water film (J/kg) Vf> V£, Vu = temperature efficiency, r)z =
(m) hg = specific enthalpy of water vapour (J/ fe out ~ tf m)l(ta'm - tf in), where z=f,£
A, Aa = surface area associated with U or kg) OTU (dimensionless)
with hc and ho (m 2 ) hwb = (1 + ehfg/ca)hca,wet bulb heat transfer r\wc = (fyout - tfin)/(ttt*m - f/in), water
Ac = minimum free flow area of passage coefficient (W/m 2 K) cooling efficiency (dimensionless)
(m 2 ) kp,kw = thermal conductivity of plate or of ix = dynamic viscosity (Pa s)
ba,bf = plate spacing for air or for fluid pas- water film (W/m K)
sage (m) L = passage length in flow direction (m) Superscripts
c = specific heat capacity (J/kg K) rha,rhf = mass flow rate of moist air (kg dry * = at adiabatic saturation relative to true or
Curb = (ca + ehfg), wet bulb specific heat ca- air/s) or of fluid (kg/s) actual saturation line
pacity (J/kg dry air K) Nu = Nusselt number (dimensionless) ' = at adiabatic saturation relative to linear-
d, e = constants in equation (3) (kg/kg dry Pr = Prandtl number (dimensionless) ized or model saturation line
air, kg/kg dry air K) Re = 2 m (1 + w)b/(Acn), parallel plate
ET* = effective temperature [16] (°C) Reynolds number (dimensionless) Subscripts
fs = 1.005, correction factor for presence of air St = hcAc/(c m), Stanton number (dimen- a = moist air stream
(dimensionless) sionless) / = fluid stream
ha = specific enthalpy of moist air (J/kg dry t = temperature (°C) in = inlet
air) U = overall heat transfer coefficient between £ = lower fluid stream
hao = constant in equation (2) (J/kg dry water surface and fluid (W/m 2 K) max = maximum value
air) w = humidity ratio (kg/kg dry air) mean = mean value
hc = heat transfer coefficient (W/m 2 K) W = overall width of plate heat exchanger min = minimum value
ho = water vapor transfer coefficient of moist (m) out = outlet
air stream (kg dry air/m 2 s) x,y = distance from fluid inlet, or from moist u = upper fluid stream
h/g - latent heat of vaporization of water air inlet, in flow direction (m) w = at water surface

580 / VOL. 103, AUGUST 1981 Transactions of the ASME

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 For a parallel flow exchanger y = xLJLf, for a counterflow exchanger 0-04
y = (1 — x/Lf)La and for a crossflow exchanger x and y are not re-
lated. w max
Conditions 1, 2 and 3 are approximately satisfied by many actual
wet surface heat exchangers. The model will approximate actual
performance if appropriately chosen mean values are used for the
constants in conditions 4, 5, 6 and 8. Condition 7 will be approximately
satisfied if the surface is wet by a continuous film of water.
Condition 9 is closely approximated in evaporative coolers and
finned coils with condensation. It is also closely approximated in w(kg/kg)
sprayed coils and wet plate heat exchangers if the wetting water mass
flow rate is negligible compared with the air mass flow rate. For cooling
*w mean
towers with high inlet water temperature and low inlet air wet bulb
temperature, a significant fraction of the water mass flow may be
evaporated. In this case, it is assumed here that the model water mass
flow rate is constant and equal to the average of the actual inlet and
outlet water flow rates.
Condition 10 is not satisfied in an actual exchanger. The model w mm . " w mm
represents actual performance if constants d and e in equation (3) are
chosen to give an approximate least squares fit to the true or actual
saturation line over the range of water surface temperatures. The
recommended values are

\Ww max Wn n)l(tu tu a) (9)

d = (2(WU + WW a) Ww m a x ) / " £ tw (10)
where tm m j n is an estimate of the minimum water surface tempera-
t( C)
ture, tw m a x of the maximum, and tw m e a n = (tw mm + tw max)/2. Also, Fig. 2 Graphical construction of the model saturation line
ww min, ww max and ww menn are the actual saturation moisture contents
at the corresponding temperatures. Equations (9) and (10) would be Since d is constant, substituting equations (1) and (3) gives
an exact least squares fit if the saturation line were a parabola. d l{d + eta — wa) h/g (d + etq - wa) hfg
CarflaLa " ~hcaJ\.(
This model saturation line may be constructed graphically on a Ca+e hfg
psychrometric chart as shown in Fig. 2. First, the estimated values dy \ ca + e hfg
of tw min, tw max and tw mean are used to plot the points w min, w max Substituting from equation (12) gives
and w mean on the actual saturation line. The straight line joining Hta ~ ta')
w min and w max is drawn, then, parallel to this line and two thirds Ca^a^-'a flca^a \CQ ta ) (13)
i>y
of the way from it towards point w mean, another line is drawn. This
is the model saturation line. The solution of this ordinary differential equation is

ta ~ ta' = (ta in ~ ta in) e x p ( - ftcaAay/(cam0La))

3 Solution of the Model Equations
In Section 3.1, the equations of Section 2 are rearranged to form which predicts that the wet bulb depression decays exponentially with
an ordinary differential equation for web bulb depression, and in distance y from the air inlet. At the outlet,
Section 3.2 an independent set of partial differential equations for
y ~ La ana ta — ta — ta out — ta out
wet bulb temperature. The equations are seen to be similar to those
for dry bulb temperature in dry surface exchangers, which have been so the solution becomes
solved elsewhere [8],
ta out — ta out —
\ta in —
ta in/ e x p \ flcaAa/\Caffla)) (14)
3.1 The Model Wet Bulb Depression, ta - ta'. Substituting
equation (2) in (8) and subtracting hg times (7) gives where hcaAa/(cama) is the number of heat transfer units for the
i>ta _ moist air stream and may be calculated using heat transfer literature
CafftaL/a ftca-A-a \ttv •ta) (ID [8]. The solution is analogous to that for the difference between dry
<>y bulb temperature and wall temperature in a dry surface exchanger
Now the wet bulb depression with constant wall temperature.
3.2 The Model Wet Bulb Temperature £„'. Rearranging
ta ~ ta' = (waJ — wa) h/g/ca (rearranging equation (4))
equation (12) gives
= (d + e ta' • Wa) hfg/Ca (by equation (3) since ta', wa'
on the saturation line) (1 + e hfg/Ca) ta' = ta + (Wa ~ d) hfg/ca (15)

= -(ta ta') hfge/ca + (d + e ta - Wa) hfg/ca Substituting equations (1, 3,15) in (6) and defining a wet bulb heat
(rearranging) transfer coefficient by hwb = (1 + e hfg/ca)h
ca §1V6S
KbAaifa' ~ tm) + UA {if -tw)=0 (16)

ta - ta' = (d + e ta ~ wa)hfe/(ca + e h/g) (12) Substituting equation (11) in (13), multiplying by (1 + hfg/ca) and
defining a wet bulb specific heat by cwo = ca + e hfg gives
Multiplying equation (11) by e hfg/(ca + e hfg) and subtracting
equation (7) multiplied by ca hfg/(ca + e hfg) gives dta' —_
Cwbrtla*-Ja ftwbAa\tw •ta') (17)
dy
d ((etq-Wq) hfg\
Ca?7l a ij a
dy\ ca + e hfg Equations (5,16,17) have been solved in the literature [8] for par-
allel flow, counterflow and crossflow heat transfer alone exchangers.
Aahfg(hcae (tw - ta) - cahp(ww - wa)) The literature solutions allow the outlet wet bulb temperature £„'• ou t
ca + e hfg to be calculated by replacing dry bulb temperature ta in the dimen-

Journal of Heat Transfer AUGUST 1981, VOL. 103 / 581

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 sionless parameters of the solutions by wet bulb temperature £„', in Section 2 and Baker and Shryock's [7] model are caused by condi-
moist air heat transfer coefficient hca by wet bulb heat transfer tions 4, 5, 8, and 9. Also, Baker and Shryock [7] did not report their
coefficient hwb and moist air specific heat ca by wet bulb specific heat vapor pressure and thermodynamic data which may have differed
Cwb- from those used here (Section 3.2). Furthermore, for the calculations
The values of cwb and (1 + e hfg/ca) for calculating hwb may be es- in their Table 2, they gave only the wet bulb temperature of the inlet
timated using Table 1 at tw meBn, or using equation (9) for e with c0 air. However, for the calculations in their Tables 1(6) and 1(c), satu-
and hfg evaluated at state w mean. Table 1 was calculated using rated inlet air was used; therefore the inlet air was assumed to be
simplified psychrometric equations from [14] with the correction saturated for the calculations in their Table 2.
factor fs = 1.005 and the latent heat of water The major difference between the two models is the linear satura-
tion line of condition 10. As defined by equations (3,9,10), condition
hfg = 2 501 000 - 2 412 tw J/kg 10 is accurate for actual exchangers with only a small water surface
where 0°C < t«, < 90° C. temperature range because a straight line can accurately approximate
a smooth curve only over a small distance. The comparison with Baker
4 * Comparison of Model with Finite Difference and Shryock's finite difference solutions suggests that the model and
Solution condition 10 are satisfactory for appreciable water surface tempera-
Table 2 compares some crossflow cooling tower finite difference ture ranges.
solutions of Baker and Shryock [7] with the model (Section 2) and its
solution (Section 3). The model shown in Fig. 1 represents a cooling 5 Prediction of Wet Plate Heat Exchanger
tower when the flowing fluid is water and the wall is removed. Performance from that Measured When Dry
Measurements on plastic plate heat exchangers with two cross flow
The first four columns of Table 2 are the first four columns of Baker
air streams, one with dry passages and the other wet by water sprays,
and Shryock's Table 2, and the fifth column is Baker and Shryock's
have been reported by Pescod [5]. The passages were otherwise
third last column which is for finite water film heat transfer resistance
identical and the dry air mass flow rates of the two streams were equal,
and water mass flow decreasing with evaporation. The mean mass flow
rha = rhf. Pescod gave equations for the wet bulb temperature
ratio rhf mean/rha in column 6 and the water cooling efficiency T\WC
efficiency of the dry stream t]f, which he called wet effectiveness, as
in column 8 were calculated using the model. The water cooling effi-
a function of passage velocity for three pin fin surfaces and one cor-
ciency T]Wc in column 7 was calculated from the first three columns
rugated surface ([5], Table 3). The full lines on Fig. 3 are these ef-
using the definition
ficiencies plotted against passage Reynolds number.
??tuc = (tf out - tf in)/(ta* in ~ tf ; n ) . Predictions of the wet bulb temperature efficiency t\f were made
using the model in Section 2 and its solution in Section 3. The heat
The percentage difference in water cooling efficiencies predicted using transfer coefficients used were those calculated by Pescod [5] from
the two methods is shown in the last column. The difference increases measurements made with all heat exchanger surfaces dry. For eval-
with the water temperature range but is less than 1.5 percent in all uating properties, a mean air temperature of 27°C and a mean water
cases. surface temperature of 22°C were assumed. The thickness of the
Minor differences between the predictions of the model presented polyvinyl chloride plates was 0.2 mm [4] and the water film thickness
was assumed as 0.2 mm. The long dashed lines on Fig. 3 are the pre-
Table 1 Wet bulb specific heat c w j, and its ratio to dictions, which are all over 10 percent higher than the measurements
moist air specific heat, with some over 20 percent higher.
cwb/ca = hwb/hca = 1 + e hfg/ea
The difference between the measurements and predictions con-
t 1 + Cwb t 1 + Cwb siderably exceeds the claimed experimental error. One or more of the
°c e hfgha J/gK °C e h/g/ca J/gK model conditions must not apply to the measurements. Pescod [5]
does not report the spray water mass flow rates. If they are excessive,
0 1.686 1.707 30 4.758 5.034
2 1.779 1.803 32 5.159 5.491
the water film would have a significant velocity and condition 9 would
4 1.882 1.910 34 5.597 5.999 not apply. If the water film does not cover the wet plates completely,
6 1.996 2.029 36 6.077 6.562 the effective value of ho may be significantly less than that of hca/ca
8 2.123 2.162 38 6.601 7.188 and condition 7 will not apply.
10 2.264 2.310 40 7.175 7.885
12 2.420 2.474 42 7.802 8.662 The effect of excessive spray water massflowrate may be examined
14 2.592 2.656 44 8.489 9.531 by replacing condition 9 by the extreme case of constant water surface
16 2.782 2.859 46 9.241 10.50 temperature throughout the exchanger. The short dashed line in Fig.
18 2.992 3.083 48 10.06 11.60 3 shows this prediction for the first pin fin surface. At low Reynolds
20 3.222 3.331 50 10.97 12.83
22 3.475 3.606 52 11.96 14.23 numbers the measurements are higher than these predictions but at
24 3.753 3.911 54 13.06 15.81 high Reynolds numbers they are lower. Assuming that constant water
26 4.058 4.248 56 14.26 17.62 mass flow rate was maintained while varying the air flow rates, the
28 4.392 4.621 58 15.59 19.70 water flow rates would be most excessive and measurements lowest
Table 2 Comparison of finite difference solution and model for a cooling tower with finite water film
resistance
Data from Baker and Shryock ([7], Table 2) Calculated values
Temperature °F rhf ^ hDAa mf mean Vwc Vwc percent
tf in f * . rhf in rha model diff.
t/out [71
69.2 62.3 47.7 1.088 1.049 1.086 0.3209 0.3225 0.48
85.1 71.7 47.7 1.09 1.032 1.085 0.3583 0.3570 -0.35
99.6 78.5 46.6 1.09 1.061 1.082 0.3981 0.3951 -0.76
119.1 86.8 46.8 1.12 1.049 1.108 0.4467 0.4404 -1.43
78.9 71.4 47.5 1.96 0.763 1.955 0.2389 0.2405 0.71
90.1 79.1 48.9 1.99 0.766 1.983 0.2670 0.2664 -0.21
102.0 86.8 51.2 2.04 0.761 2.029 0.2992 0.2966 -0.87
108.5 89.8 46.1 2.04 0.742 2.027 0.2997 0.2972 -0.84
88.0 71.6 44.3 0.954 1.073 0.9492 0.3753 0.3743 -0.26
99.1 77.3 47.4 0.928 1.099 0.9214 0.4217 0.4190 -0.64
111.6 84.6 54.6 0.964 1.071 0.9550 0.4737 0.4668 -1.45

582 / VOL. 103, AUGUST 1981 Transactions of the ASME

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 1-0

PROJECTION
0-8-

PROJECTION
b Imm) SPACING (mm)
0-6 2-82 6-7

T)f 3-56 9-5 PIN FIN SURFACES

PLASTIC
3-49 134 PLATE
0-4 3-56 CORRUGATED PLATES

PASSAGE CROSS SECTION

MEASURED
0-2-
MODEL PREDICTIONS
-EVAPORATIVE COOLER PAD
CONSTANT WATER
TEMPERATURE
J_i_ _l_ J I I i i i

0-8 1 3 4 5 6 8 10 20*100
Re
Fig. 3 Comparison of measured and predicted wet bulb temperature ef- ROOM _
ficiencies i)t for plate heat exchangers as functions of passage Reynolds INLET "
number, Re

relative to the predictions at low Reynolds numbers. Hence, excessive

water flow rates, and thus condition 9, are not the sole cause of effi-
ciency measurements lower than predicted by the model of Section
2. Poor wetting and thus condition 7 must be a significant cause of low
measured i)f efficiencies.

6 A Proposal for an Improved Regenerative

Evaporative Cooler
The precursors of the regenerative evaporative cooler analyzed in
Section 7 were a cycle attributed by Watt ([2], p. 210, Fig. 29A] to Carl PLATE HEAT EXCHANGER
Boester and a related regenerative evaporative cooler due to Pescod Fig. 4 Proposed regenerative evaporative cooler using wet plate heat ex-
([5], Fig. 10). Boester's cycle consisted of a dry finned coil precooling changer
air to a water cooling tower. The water flow in the coil was taken from
the tower sump and returned to the tower water inlet. Chilled tower
passage detail. To insure wettability, the walls of these passages are
water flow not required for the coil was used for air conditioning. A
unsized woven nylon, rayon or fiberglass cloth fastened horizontally
simple modification of Boester's cycle is to increase the size of the
by adhesive at approximately 30 mm vertical intervals to leave a
finned coil so it uses all the water circulating in the cooling tower and
capillary space with a strong horizontal wicking action between the
to use the excess cooled air directly for air conditioning. Pescod's re-
cloth and plate. The third modification, intended to minimize the
generative evaporative cooler uses this cycle but the finned coil and
overall volume of the cooler, is the irregular pentagonal shape for the
water cooling tower are replaced by a single wet plate heat ex-
plates, shown in Fig. 4, instead of the rectangular shape used by
changer.
Pescod. The shape would be designed and optimized using a com-
For the three related cycles mentioned in the last paragraph, if puter.
appropriate mass flows are chosen and the number of transfer units
increased, the cooled air or water supplied to the load approaches the
7 P r e d i c t e d P e r f o r m a n c e of t h e P r o p o s e d
dew point temperature of the ambient air even if the temperature rise Regenerative Evaporative Cooler
across the load is constant and finite. The cycle analyzed was chosen
The rectangular model of the proposed cooler shown in Fig. 5 is used
because it satisfies this condition. The minimum temperature does
below. The moist air flowing horizontally on the dry side of the plates
not approach the dew point with a constant and finite load for the
is referred to as the fluid stream with subscript /, and that flowing
other indirect evaporative cooling cycles described by Watt [2]. The
vertically on the wet side of the plates as the air stream with subscript
minimum temperature for simple evaporative coolers is the often
a. The upper part of the fluid stream which enters the evaporative
much higher wet bulb temperature. The analyzed cycle however re-
cooler pad and eventually the room is denoted by the subscript u and
tains the advantage of supplying all fresh air to the room like a simple
the lower part which becomes the air stream by £.
evaporative cooler.
The data used for the calculations detailed in Section 8 are tabu-
Figure 4 shows schematically the cooler to be analyzed. There are lated below, except for the external dimensions of the plate heat ex-
three principal modifications to Pescod's cooler ([5], Fig. 10). The first changer model which are shown in Fig. 5:
is that a simple woodwool evaporative cooler pad has been placed in
the room inlet stream to improve performance and balance the width of fluid passages bf = 2 mm
pressure drops of the two streams. The wet results for Pescod's pin overall plate thickness ap = 0.2 mm
fin surfaces discussed in Section 5 are inconsistent with theory and water film thickness aw - 0.1 mm
their pressure drop is excessive for the wet passages in Fig. 4. There- width of air passages ba = 2 mm
fore, the second modification is that the parallel plate passages tested fluid stream mass flow rate rhf = 0.500 kg dry air/s
in evaporative coolers [9,10] will be used here as shown in the Fig. 4 air stream mass flow rate rha = 0.167 kg dry air/s

Journal of Heat Transfer AUGUST 1981, VOL. 103 / 583

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 absolute pressure pt = 101.325 kPa 600 mm
£
inlet fluid dry bulb temperature
inlet fluid wet bulb temperature
tf in = 35°C
t,*m = 24°C
• 4^' »
,
The results of the calculations in Section 8 and others not given in
rhu
detail are summarized here. The room air state was chosen to be an
effective temperature ET* = 27°C. The ET* = 27°C line is shown
on Fig. 6 and passes through 25°C at 100 percent relative humidity
H
o o
and 27°C at 50 percent relative humidity [16]. The pressure drop in
IN _J
UJ
the fluid passages was 93 Pa, air passages 60 Pa and evaporative cooler m.
• '•'•
pad 50 Pa. The fan efficiency was assumed as 0.50 and the fan motor
efficiency 0.70 to estimate the fan power consumption at 189 W. The
pump power consumption was assumed to be 60 W giving a total
electric power consumption of 249 W. Inlet fluid state A, 35°C DB,
( <
J )
*

24°C WB, gave a room air inlet temperature of 21.2°C DB, and room
state of 26.2°C DB, 71 percent RH, equivalent to a cooling capacity
of 1.73 kW. Inlet fluid state B, 46°C DB, 24°C WB gave a room air - L f = 300mm
inlet temperature of 18.1°C DB, 17.0°C WB and room state of 26.9°C
Fig. 5 Model of proposed regenerative evaporative cooler using rectangular
DB, 53 percent RH, equivalent to a cooling capacity of 3.03 kW. wet plate heat exchanger
Simple evaporative coolers with humidifying efficiencies typically
0.80 are not capable of producing the design room condition ET* =
003
27° C. The volume of a commercial vapour compression unit with
coefficient of performance approximately 2 would be smaller than the
regenerative evaporative cooler for state A but larger for state B. The
electric power consumption of the proposed regenerative evaporative
cooler is less than 30 percent of vapor compression power consumption
even for inlet state A. For state B, its manufacturing cost could be
considerably less than that of a vapor compression unit if manufac-
tured in comparable numbers.

8 Sample Calculation Details for the Performance

Prediction
Details of the calculation of the mean outlet temperature of the
upper fluid stream tu out for inlet state A are given in subsections 8.1,
8.2, 8.3 and 8.4 below to illustrate the use of Sections 2 and 3. The
calculated outlet temperature tu out = 23.1°C is below the inlet wet
bulb temperature tf*\n = 24°C. On passing through the evaporative
cooler pad with humidifying efficiency 0.80, the temperature is further
reduced to 21.2°C, Fig. 6, which is close to the inlet fluid dew point
ofl9.7°C.
8.1 Evaluation of Thermodynamic and Transport Properties.
Initially the minimum water surface temperature will be assumed
tw min = 21°C and the maximum water surface temperature tw m a x
= 31°C. The mean water surface temperature
Fig. 6 Cycle of proposed regenerative evaporative cooler on psychrometric
••{tu; , + tw m a x )/2 = (21 + 31)/2 = 26°C, chart
and this temperature will be used for evaluating all fluid and air
stream properties. = 0.3 X 0.6 X 2/(2 + 2(0.2 + 0.1) + 2) = 0.078 26 m 2
From a psychrometric chart [17], ww mea n = 21.3 g/kg and the inlet
Re a = 2m a (1 + Ww mean
fluid humidity ratio Wf m = 14.2 g/kg. From Table 1, cwb = 4 248 J/kg
K and (1 + e hfs/ca) = 4.058. From equations (65, 66, 67) in [14], cf = = 2 X 0.167 (1 + 0.0213)2 X 10- 3 /(0.078 26 X 18.46 X 10~6)
1 033 J/kg K. From dry air property tables [18], Pr = 0.707 and /xa = = 472.2
Hf = 18.46 juPa s. From water property tables .[18], kw = 0.606 W/m
K, and from plastics properties tables [19], kp = 0.178 W/m K. From ([8], p. 103) Nu„ = 8.235
8.2 Calculation of Stanton Numbers and Heat Transfer so St a = N u a / ( P r Rea) = 8.235/(0.707 X 472.2)
Coefficients. Neglecting pin surface and cross-sectional areas,
following Pescod [4, 5], = 0.024 67

Ac, = LaWb,/(b, + 2(ap + aw) + ba) and hu,b = (1 + e h,g/ca) hca

= (1 + e hfgha) Stacama/ACa
2
= 0.6 X 0.6 X 2/(2 + 2(0.2 + 0.1) + 2) = 0.1565 m
— &tacwc,7na/Aca
Re/ = 2rhf (1 + wf jn ) bf/(Acf pf)
= 0.024 67 X 4 248 X 0.167/0.078 26
= 2 X 0.500 (1 + 0.0142)2 X 10- 3 /(0.1565 = 223.6 W/m 2 K.
X 18.46 X 10- 6 ) = 702.1
8.3 Calculation of Dimensionless Parameters and Temper-
From ([5], Fig. 3) St/ Pr 2 ' 3 = 0.027 so St/ = 0.027 X 0.707~-2/3 = 0.034 ature Efficiencies. The overall number of heat transfer units for
02 and hcf = St/c/m//A c/ = 0.034 02 X 1 033 X 0.500/0.1565 = using literature crossflow heat exchanger solutions as described in
112.3 W/m 2 K Section 3.2 is
Similarly
A _ hcfA
Aca = LfW bj(bf + 2(ap + aw) + ba) Cfriif (1/U + \/hwb) Cfihf (1 + hcf(ap/kp + aw/km + l/hwb))

584 / VOL. 103, AUGUST 1981 Transactions of the ASME

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 formance for a regenerative evaporative cooler using a wet plate heat
\U hcf kp kj exchanger (Section 7).
The measured performance of wet plate heat exchangers reported
= 2 StfLf
by Pescod [5] is substantially less than that predicted (Section 5). This
bf (1 + hcf (ctp/kp + aw/kw + l/hwb)) deviation may be reduced if the water film were made continuous and
( s t ^ ^ a n d ^ ^ ) the water mass flow rate small in these exchangers. Water mass flow
\ Cfriif Acf bf I rate and wet passage outlet state should be measured simultaneously
with other flows and states. Improved understanding of heat transfer
2 X 0.034 02 X 0.3
alone and pressure drop in Pescod's pin fin surfaces (Section 1) is also
~ 2 X 1 0 - 3 (1 + 112.3 (0.2 X 10" 3 /0.178 desirable. Predicted regenerative evaporative cooler performance can
+ 0.1 X 10" 3 /0.606 + 1/223.6)) be obtained only if predicted wet plate heat exchanger performance
is achieved.
= 6.197
The capacity rate ratio for the lower half of the exchanger is Acknowledgments
Cfrhe/(cwbma) = 1 033 X 0.167/(4 248 X 0.167) = 0.2432. The authors are grateful to Mr. D. Pescod and Dr. J. G. van Leer-
The lower half temperature efficiency sum, of the CSIRO Division of Mechanical Engineering, for helpful
suggestions.
V£ = (t£ out _ tf m)l(ta' in — tf ; n ) The first author is grateful to the Department of Mechanical En-
gineering, Monash University, Australia for the use of their facilities
may be calculated from the above two dimensionless parameters as
while this paper was being written. Financial support was received
described towards the end of Section 1 as y\e - 0.9768.
from the Australian Government under grant number F74/15117
The capacity rate ratio for the whole exchanger is Cfmf/(cwtma) =
awarded by the Australian Research Grants Committee.
1 033 X 0.500/(4 248 X 0.167) = 0.7281.
The fluid stream temperature efficiency t\f may be calculated
References
similarly to t\e as r\f = 0.8637.
1 ASHRAE Handbook and Product Directory, Equipment volume,
By conservation of dry air me = ma and rhf = mu + me, and by American Society of Heating, Refrigerating and Air-Conditioning Engineers,
conservation of energy mfx)f = rhuT)u + rherje, so solving simulta- New York, 1979.
neously 2 Watt, J. R., "Indirect Evaporative Cooling," Evaporative Air Condi-
tioning, The Industrial Press, New York, 1963, pp. 204-224.
Vu = (mfOf - maT)e)l(mf - ma) 3 Pescod, D., "Unit Air Cooler Using Plastic Heat Exchanger with Evap-
oratively Cooled Plates," Australian Refrigeration, Air Conditioning and
= (0.500 X 0.8637 - 0.167 X 0.9768)/(0.500 - 0.167) = 0.8070 Heating, Vol. 22, No. 9, Sept. 1968, pp. 22-26.
4 Pescod, D., "Effects of Turbulence Promoters on the Performance of
Plate Heat Exchangers," Heat Exchangers: Design and Theory Sourcebook,
8.4 Calculation of Upper Stream Outlet Temperature tu o u t . McGraw-Hill, New York, 1974, pp. 601-615.
From the approximate saturation line in Fig. 6, t/'i n = 23.8°C. 5 Pescod, D., "A Heat Exchanger for Energy Saving in an Air-Conditioning
Subtracting equation (12) for the air inlet from equation (12) for Plant," Transactions of the American Society of Heating, Refrigerating and
the fluid inlet and wa m = u>e out = u>f i„ gives Air-Conditioning Engineers, Vol. 85, Part 2,1979, pp. 238-251.
6 Mickley, H. S., "Design of Forced Draft Air Conditioning Equipment,"
tf in — tf' i n — (ta in — ta' in) = —
Ua in — tf j n ) e hfe/(ca + e hfg) Chemical Engineering Progress, Vol. 45, No. 12, Dec. 1949, pp. 739-745.
7 Baker, D. R. and Shryock, H. A:, "A Comprehensive Approach to the
which on rearranging becomes Analysis of Cooling Tower Performance," ASME JOURNAL OF HEAT
TRANSFER, Vol. 83, No. 3,1961, pp. 339-350.
ta in — tf i n = - {tf i n — tf' i n ) + (ta in ~ tf in)/(l + e hfg/ca) 8 Kays, W. M., and London, A. L., Compact Heat Exchangers, 2nd ed.,
McGraw-Hill, New York, 1964.
= ~ (tf i„ - tf' i n ) + (ta' in - tf in) IJ*/(1 + e hfg/ca) 9 Hollands, K. G. T., and Pott, P., "Development and Testing of Evapo-
rative Cooler Pads," Mechanical and Chemical Engineering Transactions of
(from definition of r/e) The Institution of Engineers, Australia, Vol. MC7, No. 2, Nov. 1971, pp.
66-72.
= ~ (tf in - tf' 10 Deutscher, K. O., Dunkle, R. V., and Megler, V. R., "A Fundamental
i n )/(l - W ( l + e hfg/Ca))
Study of Heat, Mass and Momentum Transfer in a Parallel Passage Evaporative
Cooler Pad," First Australasian Conference on Heat and Mass Transfer,
From definition of T\U Monash University, Melbourne, 1973, Section 5.2, pp. 7-14.
11 Baclic, B. S., "A Simplified Formula for Cross-Flow Heat Exchanger
Effectiveness," ASME JOURNAL OF HEAT TRANSFER, Vol. 100, No. 4, Nov.
tu out — tf in "r \ta in ~ tf i n ) l)u
1978, pp. 746-747.
12 HP-67/HP-97 MATH PAC 1, Hewlett-Packard Company, Corvallis,
= tf in - (tf i n - tf' i„))7u/(l - r\el(\ + e hfe/ca)) 1976, pp. 18-01-18-02.
13 Maclaine-cross, I. L. and Banks, P. J., "Coupled Heat and Mass Transfer
(substituting ta' in — tf j n from equation above) in Regenerators—Prediction Using an Analogy with Heat Transfer," Inter-
national Journal of Heat and Mass Transfer, Vol. 15, No. 6, June 1972, pp.
= 35 - (35 - 23.8) 0.8070/(1 - 0.9768/4.058) 1225-1242.
= 23.1°C 14 Maclaine-cross, I. L., "A Theory of Combined Heat and Mass Transfer
in Regenerators," Ph.D. Thesis, Department of Mechanical Engineering, Mo-
nash University, Australia, 1974.
9 Conclusion 15 Banks, P. J., "Prediction of Heat and Water Vapour Exchanger Per-
formance from That of a Similar Heat Exchanger," Compact Heat Exchang-
The linearized approximate theory presented in Sections 2 and 3 ers—History, Technological Advancement and Mechanical Design Problems,
extends the "enthalpy potential" theory [6, 7] of wet surface heat Conference Proceedings, HTD-Vol. 10, The American Society of Mechanical
exchangers, showing that wet bulb temperature and wet bulb de- Engineers, New York, 1980, pp. 57-64.
pression may be considered as independent driving potentials in the 16 ASHRAE Handbook and Product Directory, Fundamentals volume,
American Society of Heating, Refrigerating and Air-Conditioning Engineers,
operation of such exchangers, enabling performance prediction by New York, 1977, p. 8.21.
analogy from that of dry surface heat exchangers. 17 "What's in a Chart?," Australian Refrigeration, Air Conditioning and
The theory presented may be expected to predict accurately wet Heating, Vol. 28, No. 9, Sept. 1974, pp. 17-18.
surface heat exchanger performance provided that the water surface 18 Mayhew, Y. R and Rogers, G. F. C, Thermodynamic and Transport
Properties of Fluids SI Units, 2nd ed., Blackwell, Oxford, 1974.
temperature range is less than 20 K and the Lewis relation (condition 19 "Resins and Molding Compounds," Modern Plastics Encyclopedia
7) is satisfied (Sections 4 and 5). This theory predicts excellent per- edition, Vol. 54, No. 10A, Oct. 1977, p. 483.

Journal of Heat Transfer AUGUST 1981, VOL. 103 / 585

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 A Parametric Study of a Particulate
Ruth Letan
Associate Professor,
Direct Contact Heat Exchanger
Department of Mechanical Engineering,
Ben-Gurion University of the Negev,
Beer-Sheva, Israel The function performed by a direct contact heat exchanger is by definition a heat interac-
tion between two media without an interfering wall between them. The direct contact heat
exchanger analyzed herein refers to the restricted definition of systems in which tempera-
tures of the contacted media vary without change of phase. This class involves the partic-
ulate systems of immiscible liquids, solid particles in fluids, gas bubbles in liquids, and
droplets in gases. The applicability of direct contact heat exchangers has been considered
for use in fouling and crystallizing systems, sea-water heating, geothermal brines, as well
as in nuclear and MHD power plants. An analytic way to evaluate the performance of a
direct contact heat exchanger is applied to examine the effects of operational variables
upon length and diameter of a heat exchanger. The heat exchanger becomes shorter as
the particle size decreases, and heat capacity ratio, flow rate ratio and approach temper-
ature increase. The diameter of the column decreases with the particle size, and the den-
sity ratio of the contacted media. It increases with the flow rate ratio. All the evaluations
made, correspond to the laminar bulk flow in the heat exchanger, and to particle Reyn-
olds numbers in the range of 20-500.

Introduction size. The assumption of uniform size, or a representative average size,

Direct contact heat exchangers have attracted attention mainly applies to a variety of systems in which change of phase does not take
because of their applicability to fouling and corrosive fluids, clogging place. Systems with widely distributed sizes across or along the ex-
suspensions, and solid media of granular character. Development of changer may be analyzed accordingly, by applying the general prin-
the heat exchangers has been initiated for desalination processes [1] ciples underlined herein.
and carried on to geothermal systems [2], to crystallizing brines [3, The question to be considered relates to the size of the heat ex-
4], to nuclear reactors [5], and considered for application in MHD changer, namely the diameter and length of the exchanger, needed
combustion systems [6]. to affect a temperature change, AT, in the process medium supplied
In this paper, a direct contact heat exchanger is parametrically at a flow rate, Q. The effects imposed on the exchanger geometry by
analyzed, namely the effects of operational parameters upon the ge- the variation of the operational conditions and the physical properties
ometry design variables are examined. These parameters relate to flow of the contacted media are investigated.
rates, temperatures, physical properties of the contacted media, and
size of the dispersed medium. Geometry Variables
By definition, in a direct contact system heat is transferred between The main contact volume of the two media is the column proper.
media without an interfering wall between them. Presently, the study Therefore, the elemental geometry variables are the diameter of the
is concerned with a direct contact heat exchanger within its conven- column proper and the length of the column proper. Other geometry
tional, restricted definition which applies to a heat interaction be- variables such as inlets and outlets vary in design with the character
tween two media of different temperatures, taking place without of the contacted media and are not presently considered.
change of phase, mass transfer, or reaction in the contacted media. Diameter. In the analysis, the process medium is considered to
However, the principle to be demonstrated, is applicable to direct be a continuous fluid in the direct contact heat exchanger.
contact heat exchangers in their broader definition [4], The diameter, D, of the column proper of the exchanger may be
Operationally, a direct contact heat exchanger is a container of any calculated for a specified flow rate of the process fluid, Qc [7],
shape or configuration which provides good contact between the
process and working media. The conventional shape, configuration, D=2 (1)
and operation relate to a tubular vertical column in which the process
and working media move countercurrently to each other, with one of where, uc is the superficial velocity. The evaluation of uc is to be
them being particulately dispersed. The dispersion may consist of achieved through the following relations:
solid particles, liquid droplets or gas bubbles. The character of the (a) The superficial velocities in a counterflow are related to the
dispersion is usually determined by the process medium, although slip velocity [8],
in some processes, an optimization may dictate the choice. For ex- uc IR 1 \
ample, coal preheating requires the process medium of solid particles ud
H = uc\ (2)
to be contacted with gases. On the other hand, a geothermal brine may 1 -0 U 1-0/
be contacted with a working medium of liquid droplets or solid peb- (b) Slip velocity is related to holdup. The semi-empirical ex-
bles. pression [9],
The present parametric analysis is concerned with a vertical tubular
Us = UT(1 - 0 ) " (3)
heat exchanger operated in a counterflow with particles uniformly
dispersed in a fluid. The physical properties of the contacted media fits with reasonable accuracy a wide range of experimental results
remain the same along, and across the column, and so does the particle found in the literature [10]. For the presently considered range of 20
< Re a ^ 500, we obtain the exponent, n, [9] as,
Contributed by the Heat Transfer Division and presented at the Winter
Annual Meeting, December 10-15,1978, San Francisco, Calif, of THE AMER- n = 4.45Re0-°-1-l (4)
ICAN SOCIETY OP MECHANICAL ENGINEERS. Revised manuscript received
by the Heat Transfer Division June 3,1980. Paper No. 78-WA/HT-16. The terminal velocity, UT, may be expressed as,

586 / VOL. 103, AUGUST 1981 Transactions of the ASME

Copyright © 1981 by ASME
Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 Pd - 1 Tdo •
uT = 0.268 R e 0 ° - 3 \g (5)
TV
by using the drag coefficient relationship for smooth spheres [7]. mS
(c) The operation holdup, <t>, at any flow rate ratio, R(= Ud/uc), 1- « 2 + 1 exp (a-sL) + 1 exp (ID
is usually chosen as a fraction of the holdup at flooding, <j>f [7]. The rai
state of flooding corresponds to the onset of disruption of stable flow
conditions, and therefore the holdup at flooding represents the where, L » 0 and.
maximum achievable holdup at any specified flow rate ratio. P = M+ l/R (12)
The relationship between 0/ and R, has been previously derived
[7] in an implicit form. In an explicit form, the relationship is ex- with
pressed as, 1 1 1 1 1 1\2 J_ 1/2'
«1,2 = —+ ± (13)
2 1/2 M r P M r P
-{n + 2) ± [(n + 2) + 4(ra + l)(l/fl - 1)] Mr
(hi = (6) and
' 2(n + l)(l/fl - 1)
where 0 < 0/ < 1 within OX R < °°. m mr
ai -\ exp - 1
ForR = a>(uc = 0) we o b t a i n r PM
(14)
4>f\ (7) Ct2 - «1
n +1
Specifying the inlet and outlet temperatures of the process fluid, T c j,
whereas, for R 1, the original expression [7] reduces to, and T c o , respectively, and choosing as a parameter the approach
1 temperature between the two media at one end of the column, for
<t>f (8) example A.Ta = TM — Tco, determines also Tdo by the relationship,
n+2
Thus, equations (6-8) yield, Tci = Rr(Tdi - Td0) (15)

<l>f = <t>f(n,R) = <t>f(Re„,R) (9) The parameters involved in equation (11) may be classified as free
parameters and wake parameters.
with, n, defined by equation (4). The free parameters relate to r, R, and T&. Each of those may serve
(d) Combining equations (1-5) leads to, as a free variable in the design of an exchanger.
„ 2.18Q c 1 / 2 [fl(l/tf>-l) + l]i/2 Re„ The wake parameters, M and m, vary with the flow regime in the
D= — : .,( n +...—
1>
x- (10) particulate system and have to be evaluated or measured. The method
>1/4 (1 - 0 ) 1/4
Pd of evaluation adopted herein, relates to the general flow characteristics
Pc of a particle and their extension to a particulate system.
where, 0, as a fraction of 0/ is obtainable from equations (6-8). For evaluation of wake parameters, let us define, Re s ,
In a study of the relationship expressed by equation (10), 0/ may
dD
be substituted for 0. Re s •• Re„(l - <f)n (16)
Equation (10) provides us with the tools to follow the variation of
the exchanger diameter by varying the flow rates and properties of Obviously, Re s , is always smaller than, Re D .
the working medium for any specified flow rate of the process Within the presently considered flow range of 20 < Re 0 < 500 let
fluid. us restrict our analysis to
Length. The methods of approach to direct contact heat ex-
changers based upon an area heat transfer coefficient or volumetric 24 < Re s < 500(1 - 0) 1 - 4
overall heat transfer coefficient were previously discussed in detail
where, the upper limit corresponds to Re 0 < 500 [4].
[7,11]. Their applicability to specific cases was outlined [7]. A more
Thus, the relative wake volume, M, is evaluated as follows (equation
generalized relationship and thereupon a general method of evalua-
(14) of [4]). For 24 « Re s « 150,
tion of an exchanger length necessitated an analysis of the flow regime
imposed upon the moving particles, and knowledge of the mechanism M = 0.25(1 — cos a s )(cos as — cos 2as) + (1.05 Re s — 1.45) sin 2 a s
of heat transfer between the particles and the fluid. In a particulate
uniform system, of solid particles, liquid droplets or gas bubbles in (17)
a fluid, where the particle flow regime corresponds to 20 < Re 0 < 500, For 150 < Re s *S 500(1 - 0) - 1 4

and the bulk flow in the column is laminar, the mechanism of transfer
by wake shedding [7,11] dominates. M = 0.25(1 - cos a s )(cos as - cos 2as) + 0.835 sin 2 a s (18)
The relationship between temperatures and length was formulated as, the point of boundary layer separation on a sphere, was measured
for the wake dominated system [7] as, and graphically presented [12] as a unique function of Re s namely

.Nomenclature.
cp = specific heat capacity, J/kg °C li = dynamic viscosity, kg/m-s
D = diameter of column proper, m r = (cpp)d/cpp)c v = kinematic viscosity, m 2 /s
dp = particle diameter, m Re 0 = single particle Reynolds number, p = density, kg/m 3
g = 9.8 m/s 2 d„UT/vc 0 = holdup, volumetric fraction of particles
L = length of column proper, m Re s : particle Reynolds number, dpus/vc in a column
M = relative volume of wake to particle S = defined by equation (14)
m = volume of wake elements shed per vol- T = temperature, °C Subscripts
ume of particle and unit length of column, u = superficial velocity, Q/(TTD2/4), m/s b = bottom of the column
m-1 us = slip velocity, equation (2), m/s c = continuous
n = defined by equation (4) UT = terminal velocity, m/s d = dispersed
Q = volumetric flow rate, m 3 /s «i,2 = defined by equation (13) / = flooding
R = volumetric flow rate ratio, Qd/Qc = as = angle of separation from rear stagnation i = inlet
UdlUc point o = outlet

Journal of Heat Transfer AUGUST 1981, VOL. 103 / 587

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 as=a8{Res) (19) in turn brings upon a smaller slip velocity between the phases. Again,
a larger cross-section is required for a given flow rate. Namely, di-
The other wake parameter, the relative wake shedding m, is eval- ameter increases with flow rate ratio, as presented in Fig. 2.
uated as (equation (25) of [4]) Variables which govern heat transfer rates between the phases,
2 determine the exchanger length. Particle size has a complex effect
, K l + j>/fl(l-0) sin ag
P
upon these rates. The larger a particle is, the larger is the amount of
Re 0 i/ 2 ( l - * ) » « (cosa,)1'2 U
°' heat to be transfered to it. That calls upon a longer contact time with
if, is an experimental constant. It may be obtained by a single ex- the continuous medium, namely a longer heat exchanger.
periment in any particulate system in which, ra, is measurable [11] Additionally to that, a spherical particle increases in velocity with
at specified flow conditions (if, <p) of the particular process fluid (jic, its size, and needs a longer column for a given contact time. The wake
pc) and the working medium (dp, pd). parameters, M, and, m, also affect the rates of heat transfer, although
The relationship obtainable combining equations (11-15, 20), and implicitly. The overall effect, as observed in Fig. 3, makes the heat
(17-18) will show us that the length of an exchanger for a specified exchanger longer, as the particle size becomes larger.
temperature change in the process fluid, may be designed by varying Figure 3 shows also that increasing the heat capacity ratio, r, makes
inlet temperatures, flow rates, and properties of the working me- the column shorter. This effect may be physically analyzed as follows:
dium. For a specified temperature change in the continuous phase, and a
Volume. The two elemental geometry variables, the diameter, D, higher capacity ratio, the particle temperature change is smaller. A
and the length, L, determine the volume of the heat exchanger re- higher temperature difference is preserved along the exchanger, the
quired to affect a temperature change, AT C , in the process fluid rate of heat transfer is higher, and for a given amount of heat to be
supplied at a volumetric flow rate, Qc. The volume of the exchanger, transferred, a shorter column is needed.
(TTLD 2 /4), may be followed through the term, LD2, where, L, is cal- The effect of approach temperature, AT;,, between the contacted
culated by equations (11-20), and D, by equations (10) and (6). media is exhibited in Fig. 4. It is unidirectional and related to length
All the equations derived so far, relate solely to a particulate system only. The larger the difference, the higher is the heat transfer rate,
of constant physical properties and uniform particle size in the ex- and the shorter is the exchanger needed for a specified operation to
changer. The analysis is also limited to a particle flow regime within take place.
the range of 20 ^ Re 0 ^ 500, and a laminar bulk flow in the ex- With regard to flow rate ratio, R, the same physical phenomena,
changer. which bring upon a larger diameter of the column, make it shorter.
With flow rate ratio increases the holdup of particles, and their slip
E f f e c t of O p e r a t i o n a l V a r i a b l e s velocity, decreases. The slower particles require a shorter column to
The operational variables presently considered relate to a single transfer an amount of heat to their surrounding medium. The actual
heat exchanger designed for a process fluid of a specified flow rate and picture is much more complex, as the mechanism of transfer between
temperature change. In the analysis of an industrial plant, which in- the phases is governed by the flow characteristics. However, the
volves several heat exchangers and other types of equipment, addi- overall trend, as illustrated in Fig. 4, follows the above indicated di-
tional variables have to be accounted for. In an optimization of a single rection.
heat exchanger, the geometry considerations play a major role. The effect of other variables upon length and diameter can be ex-
However, the cost function may dictate a length to diameter ratio amined by physical reasoning in a similar way. However, an overall
away from the geometrical optimum. effect of the non-unidirectional variables, upon the heat exchanger
The operational variables may be divided into four groups: physical volume, cannot be qualitatively predicted by such interpretation.
properties, particle size, flow rates, and temperatures. Their effect The volume of the heat exchanger is determined by the combination
upon the column geometry varies, and may be followed by assigning LD2, which may be investigated with respect to any physical or op-
numerical values to all variables involved in the governing equations erational variation in the exchanger. As an example, the particle size
(equations (10) and (11)). The physical fundamentals of those equa- and density ratio variations have been chosen for the illustration. As
tions were outlined in their original formulations [7, 11]. Presently may be observed in Fig. 5, the role of density ratio, and thereupon of
of interest to us are the trends exhibited by the geometry variables, the exchanger diameter is decisive in the vicinity of, palpc — 1. Away
which result from those formulations. Some numerical examples are from that value the effect of the particle size upon exchanger length
presented in Figs. 1-5. is reflected in the volume of the exchanger.
The physical phenomena responsible for diameter variation relate The graphical presentations numerically provide an approximation
to the relative velocity between the phases, while those responsible of the effects to be expected. The inaccuracy is mainly due to the
for length variation relate to the rate of heat transfer between the uncertainty involved in the evaluation of the wake parameters, M,
phases. and, m, and to a smaller extent in the flow characteristics. The con-
With respect to column diameter the physical picture, in a sim- stant, K, of equation (20) was evaluated as, K = 0.1, for the present
plified way, may be illustrated as follows. The faster particles move computational program of the column length. The numerical evalu-
in a column, the smaller is the cross section required for a specified ation was based on experimental data obtained in a kerosene-water
_3
flow rate. The larger a spherical particle is, the higher is its velocity system [11], with 3 X 10 m, sized droplets. The accuracy of if and
in a medium. Hence, for any flow rate, the column diameter is reduced its general applicability has to be experimentally tested in systems
as the particle size is increased. Figure 1 shows us that effect clearly. of different physical properties, and long exchangers. K, obtained in
The effect of density ratio is explicit within the terminal velocity a specific system for which the exchanger is designed, may consider-
(equation (5)). From equation (10) it is apparent that the column ably improve the estimates of length.
diameter increases infinitely as the density ratio approaches unity. With respect to the column diameter, a closer evaluation may be
By physical reasoning we may expect that effect. In counterflow, achieved through the information of the slip velocity-holdup rela-
vertical movement between phases is due to their difference in den- tionship in the actual system.
sity. The closer their densities are, the slower the phases move, and The analysis carried so far, corresponds solely to a laminar bulk
a larger cross section is needed for a given flow rate. This trend is flow, and a particle flow regime, where phenomena of wake formation
obvious in Fig. 1. and shedding behind particles dominate. The study was mainly in-
The flow rate ratio, R, is another operational variable which affects tended to gain some more insight into the nature of particulate direct
the column geometry. Increasing the flow rate ratio of the two phases, contact heat exchangers.
for a specified flow rate of one phase, means a higher total flow rate.
That explicitly requires a larger cross section. However, the implicit C o n c l u s i o n s
physical phenomenon related to flow rate ratio expresses itself in The geometry of a direct contact heat exchanger of particulate
holdup of particles. It increases with the flow rate ratio [7], and that character operated countercurrently, varies with the physical prop-

588 / VOL. 103, AUGUST 1981 Transactions of the ASME

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 0.8

Q C = 2M(5 3 m 3 /s
ATb = I0° C
R =3

L,m

I x ICTm
d p = 8» K)3m
5« I0°m
5 x I0 3 m
3 d D = 3« lO^m
10 x I0 m

Fig. 1 Effect of particle size and density ratio on diameter Fig. 3 Effect of particle size and heat capacity ratio on length

I.Oi

Q c = 2 * I0"3 m3/%
dp = I«l6 3 m
j>d'.?-=Q5
0.8

D, m
0.6

0.4

0.2
0 2 4 6 0 2 4 6 8

R — -
Fig. 2 Effect of particle size and flow rate ratio on diameter Fig. 4 Effect of approach temperature and flow rate ratio on length

erties of the contacted media, their flow rate ratio, the approach
temperature between them, and the particle size of the dispersed
phase.
For a process fluid of a specified flow rate, and temperature change,
the column length decreases with approach temperature, and with
flow rate and heat capacity of the working medium. It increases with LD>3
particle size.
The column diameter becomes larger with the flow rate of the
working medium and with the density ratio as it approaches unity.
The diameter becomes smaller as the particle size increases.
The volume of the heat exchanger is determined by the combined
effects of diameter and length. The volume increases with the particle
size, and with the density ratio of the contacted media as it approaches
unity.
All the evaluations made, correspond to a laminar bulk flow in the
exchanger and to particle Reynolds numbers in the range of 20-
500.

Acknowledgment
The computational work presented graphically was carried out by
Mr. E. Hoch. Fig. 5 Effect of particle size and density ratio on volume

Journal of Heat Transfer AUGUST 1981, VOL. 103 / 589

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 References 7 Letan, R., "Design of a Particulate Direct Contact Heat Exchanger:
1 Woodward, T., "Heat Transfer in a Spray Column," Chemical Engi- Uniform Countercurrent Flow," ASME Paper 76-HT-27,16th National Heat
neering Progress, Vol. 57,1961, pp. 52-57. Transfer Conference, St. Louis, Mo,, Aug. 9-11,1976.
2 Surrat, W. B., and Hart, G. K., "Testing of Direct Contact Heat Ex- 8 Lapidus, L., and Elgin, J. C, "Mechanics of Vertical-Moving Fluidized
changers for Geothermal Brines," ASME Paper 77-HT-4,17th National Heat Systems," Journal of American Institute of Chemical Engineers, Vol. 3, No.
Transfer Conference, Salt Lake City, Utah, Aug. 14-17,1977. 1,1957, pp. 63-68.
3 Letan, R., "A Direct Contact Cooled Crystallizer," Industrial and En- 9 Richardson, J. F., and Zaki, W. N., "Sedimentation and Fluidization,"
gineering Chemistry Process Design and Deuelopment, Vol. 12, No. 3,1973, Transactions of Institution of Chemical Engineers, Vol. 32,1954, pp. 35-53.
pp. 300-305. 10 Letan, R., "On Vertical Dispersed Two-Phase Flow," Chemical Engi-
4 Zmora, J., and Letan, R., "Direct Contact Cooling of a Crystallizing neering Science, Vol. 29,1974, pp. 621-624.
Brine," Proceedings of the International Heat Transfer Conference, Vol. 4, 11 Letan, R., and Kehat, E., "The Mechanism of Heat Transfer in a Spray
1978, pp. 61-65. Column Heat Exchanger," Journal of American Institute of Chemical Engi-
5 Pierce, R. D., Dwyer, O. E., and Martin J. J., Journal of American In- neers, Vol. 14, No. 3,1968, pp. 398-405.
stitute of Chemical Engineers, Vol. 5,1959, pp. 257-261. 12 Taneda, S., "Experimental Investigation of the Wake behind a Sphere
6 Eustis, R. H., Private Communication, Mechanical Engineering De- at Low Reynolds Numbers," Reports of Research Institute of Applied Me-
partment, Stanford University, Stanford, Calif. chanics, Kyushu University, Vol. 4,1956, pp. 99-105.

Change of Address Form for the Journal of Heat Transfer

Present Address-Affix Label or Copy Information from Label

If You Are Planning
To Move, Please
Notify The
ASME-Order Dep't
345 East 47th St.
N.Y., N.Y. 10017
Don't Wait! Print New Address Below
Don't Miss An Issue!
Allow Ample Time To
Effect Change. Name -

Attention.
Address

City/State or Country/Zip .

590 / VOL. 103, AUGUST 1981 Transactions of the ASME

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 Conduction of Heat across
J. Eftekhar
G. Darkazalli Rectangular Cellular Enclosures
Mem. ASME

A simplified analytical model for the computation of thermal conduction across rectangu-
A. Haji-Sheikh lar-celled enclosures based on the assumption of quasi-one-dimensional conduction in
Mem. ASME the cell partitions is presented. The rectangular enclosures may contain solid or liquid
for which the conduction is two or three-dimensional depending on the geometrical con-
Mechanical Engineering Department, figuration. Additional assumptions concerning radiation interchange between participat-
University of Texas at Arlington, ing surfaces are necessary when the enclosure contains a stagnant gas. This analytical
Arlington, Tex. model leads to a closed form solution for temperature distribution in the partitions and
the multidimensional conductive region. A parametric study of heat flux is presented.
The numerical data define a range of parameters for which a one-dimensional conduction
model is satisfactory.

Introduction Analysis
Numerous experimental and numerical computational studies The analytical model consists of a two-dimensional region bounded
dealing with free convection within enclosures have been reported between thin partitions, a constant temperature plate, and a thin plate
in the literature. As early as 1954, Batchelor [1] presented the results (plate 2, Fig. 1(a)) parallel to the constant temperature plate. The
of the experimental investigations of free convection across a cavity partitions and plate 2 are assumed to be sufficiently thin (see Ap-
with opposite vertical boundaries maintained at two different tem- pendix) so that the conduction of heat within them is quasi-one-
peratures. Additional research includes the study of free convection dimensional. The geometrical configuration for this model is shown
within air filled enclosed rectangular cavities by Davis [2] and Yin, in Fig. 1. The differential equation which governs the steady state
et al [3]. Recent studies concerning free convection heat transfer across conduction of heat across a two-dimensional air layer is the Laplace
rectangular-celled honeycombs are reported in references [4] through equation, which may be obtained from any standard heat transfer
[9]. It is reported in reference [9] that when the Rayleigh number is text, e.g. reference [10].
less than 104 to 106 (or more) depending on various parameters, heat
dzr„ | d2Ta
transfer is by combined conduction and radiation. This condition is :0 (1)
2
highly desirable if rectangular-celled cavities are used in order to re- dx dy 2
duce the rate of heat loss from a wall. Hexagonal or rectangular-celled where Ta is the temperature of the two-dimensional region; e.g., air.
honeycombs are used to reduce the transfer of heat across skylights
and doubleglazed windows by suppressing the free convection of heat
PARTITION
in the trapped air layer.
The experimental data in reference [9] are obtained for carefully Jit
controlled laboratory conditions where heat transfers across cavities \ m f
located between two isothermal plates. At low Rayleigh number this ,»
situation leads to an identical linear temperature distribution in the
fluid layer and partitions. However, in many situations, one surface,
e.g., plate 1, Fig. 1, can be considered isothermal, while the other _. 2H
X
surface, e.g., plate 2, Fig. 1, transfers heat to the inside of a conditioned
space by free convection. The temperature distribution in the fluid
layer is no longer one-dimensional. The rate of heat transfer depends
PLATE 1
on the geometrical dimensions, thermal conductivity of all partici- V
pating surfaces, and the heat transfer coefficient between the exterior \ \ PARTITION \
of plate 2 and an external fluid. This paper provides reference values PLATE 2
(low Rayleigh number) for heat flux across rectangular-celled cavities L
when plate 2 is not isothermal, and will enhance the range of useful-
ness of available information on this subject. An analytical model for Fig. 1(a)
the calculation of heat loss is presented and an analytical solution is
derived for conduction of heat in a composite medium which consists
of two homogeneous regions; a quasi-one-dimensional and a multi-
dimensional. This solution is valid when both regions are solid as well
as when the rectangular cell cavities are filled with a fluid, provided
the corresponding Rayleigh number for the latter case is below the
?t
+ J L
onset of free convection. A parametric study describing the effect of t
V
two-dimensional conduction in the fluid layer is discussed. A closed
form two-dimensional solution is obtained when the enclosures are
separated by horizontal partitions, Fig. 1(a). The generalization is X
_. J'.. 2H

made to include the conduction of heat across rectangular celled re-

gions, Fig. 1(6).

L_ L
,—— g>
1 2W
ir
Fig. 1 ( 6 )
Contributed by the Heat Transfer Division for publication in the JOURNAL
OP HEAT TRANSFER. Manuscript received by the Heat Transfer Division July Fig. 1 Schematic of a two and three-dimensional rectangular-celled
10,1980. cavity

Journal of Heat Transfer AUGUST 1981, VOL. 103 / 591

Copyright © 1981 by ASME
Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 The boundary conditions which should be satisfied are where

T = T 0 at x = 0, rh 1 - Bf/Ba
L cos(X„L) (10a)
t 1+Bf
•• 0 a t y = 0, (2) Cognizance should be taken of the continuity of heat flux at the in-
dy
terface used in the derivation of equation (6). The continuity of
dTa
-ktt | - ^ | = Ua(T-T„)atx=L temperature at the interface was utilized for evaluation ot the constant
i>y i Cn in equations (4) and (10); that is,
including continuity of temperature and heat flux at the interface of
:
air and partition which will be discussed later. The thermal conduc- Y, A„in + R;m • 1,2. (11)
tivity of air is ka and the overall heat transfer coefficient is Ua which n=l
includes the contribution of the thermal resistance of plate 2 and in which
convection coefficient of the cooling fluid on the external surface of
Am = Cm sinh (XmH) (Ha)
plate 2. A considerable simplification is realized by introducing the
transformation - ( l + B a )cos(X m L)
(Ufa)
BaXmL
= (xlL)BJ(l + Ba) + I (3) sin (2X m L)
To-T. (He)
y m = r -
The solution of the function 8a using standard separation of vari 4X,„L
ables solution technique is ImT
Vm = 7m c o t h (\mH) + (Hd)
6a = £ Cn sin (Xnx) cosh (X„y) (4)
n=l and
In which A„ is the eigenvalue obtained from the equation Bf Ba
R = (He)
1+Bf 1 + Ba
tan (XL) = -XL/Ba (5)
It should be mentioned that the coordinate x was eliminated from
The quasi-one-dimensional conduction of heat in the partitions is equation (10) using orthogonal properties of the sine function.
derived by applying the energy conservation law to the material ele- Equation (11) appears to require a simultaneous solution of infinite
ment in Fig. 1. equations, but the solution is accomplished analytically. It is common
2 practice to obtain a series solution, and then study the convergence
kit ld T, of the solution. However, here it is assumed that equation (11) con-
[d \ dy / a t y = H
(6)
tains a convergent series. Based on this assumption, the right hand
The transformation side of equation (11) is constant and independent of m.
T0-Tf Ami A2T)2
= (x/L)Bf/(l + Bf) + 0f (7) T.An£n +R (12)
To - T„ 7T" /?2 _ ' " _ /9m " fin
in conjunction with equation (3) and (4) reduces equation (6) to the which leads to the equation
following form
An = (AmVm/Pm)(BJVn) (13)
d 2 0, r -
—l - = - £ CnX„ sin (\nx) sinh (X„ff) (8) The substitution of An from equation (13) in equation (11) results
dx U=i
in the evaluation of Am (which is related to Cm, equation (11a)).
wherein r = ka/kf. The solution for this equation after using the ap-
propriate boundary conditions Rfim/Vm
Am ~ " (14)
0/ = 0 at x = 0 1 - L Pntn/Vn
and The final solution hence becomes

-kf UfBf at * = L (9) :

£ Am sin (\mx) cosh (X m y)/sinh (XmH) (14a)
dx m=l
takes the following form and

£ — sin (X„%) sinh (X„ff)

Bf = - - £ ^ sin (Xmx) + j £ AmU (146)
t n=\ Xn t m=\ A m u m=l

+ ~ £ C„f„smh(A„#) (10) Now the examination of the convergence becomes necessary. As Xm

L n=i goes to infinity, /3 m goes to zero since 1/Xm goes to zero. All other
quantities converge to finite limits.

^Nomenclature-
An, Bn,Cn = Fourier coefficients Q = heat transfer rate
Ba = UaL/ka Qr = Qa + Qf X = eigenvalue, equation (15)
Bf = UfL/kf t = i thickness of partition, Figs. 1 8 = temperature function, equation (3)
H = cell height, Fig. 1 U = conductance atx = L, equation (2)
k = thermal conductivity W = cell width, Fig. 1 Subscripts
L = cell length, Fig. 1 x,y,z = coordinate, Fig. 1 a = air
Nu = heat flux ratio, QT,2/QT,I /J = equation ( l i b ) / = fin-type partition
r = kjkf y = equation (lie) m,n = indices
R = constant, equation ( l i e ) f = equation (10a) 1 = one-dimensional solution
qr = radiant heat flux r] = equation ( l i d ) 2 = two-dimensional solution

592 / VOL. 103, AUGUST 1981 Transactions of the ASME

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 The examination of equation (14) reveals that these series converge JD (2)'
uniformly [11] over the entire domain. Numerically, it was observed - £ (A,„lB(«-AmiBW)f„
n=l
that 300 terms are sufficient, in most cases, to achieve an accuracy of ,(2)
,(D
greater than four significant figures. 2—, A-m,n (1) U y A w)f
Rectangular Cells. The temperature distribution in a rectan- 11 = 1 ra = l

gular-celled cavity, Fig. 1(6), can be derived following the transfor- „<» (2)

mation given by equation (2). The function 8a can be decomposed into • - £ Am,„< - V A <2)f (206)
n=l
two functions, 8a\ and 8a2.
where
6a(x,y,z) = 8ai(x,y) + 8a2(x,z) (15)
A„ (21a)
The substitution of 8a into the three-dimensional Laplace equation
results in two distinct two-dimensional Laplace equations. Applying An, ,(1) = rm,n(V2>u(1>-£n(1W2)) (216)
the appropriate boundary conditions similar to derivation of equation (2) ! 1
A„ , = r B ,fe»'5 l ,f i-f,\ 1 ( )) (21c)
(4) results in
^m,n (21d)
Oai(x,y)= E Am sin (\mx) cosh (X,„v)/sinh (\mH) (16a) /3„ 1
jn = \ r, (21e)
and
(1) : coth (Xmff) + (21/)
ea2(x,z)= £ Bm sin (Xmx) cosh (A m z)/sinh (AmW) (16&)
m=l
r
(2) = . coth(X m W) +
The solution for quasi-one-dimensional temperature distributions (21g)
t\r.
in the partitions remains unchanged; that is, at y = H partition
fr (1) : — and£?m<2> = ^ - (21h)
9/i = - - t T 2 sin (Xmx) + j £ Am fm (17a) ,W XmH
t m=\ Am Li m = l
Parametric Example: The total heat flux leaving the constant
and at z = W partition, temperature plate, at x = 0, per unit width of the plate is calculated
r °° B x m by the application of the Fourier equation to the temperature solu-
- - T, r ^ s i n (Xmx) + - £ B m f m (176) tions, equations (14), (14a) and (146). The result is
t 771 = 1 A m Li
LQa Ba L_ -
The Fourier coefficients Am and Bm in equations (16) and (17) can (22a)
kaH{Ta - T„) 1 + Ba + Hn^i "
be determined by satisfying the condition of continuity of temperature
at the interface of air and partitions. Unlike the two-dimensional and
solution, the three-dimensional solution requires an additional as- Bf rh »
LQf
sumption in order to satisfy the basic assumption of quasi-one-di- (226)
mensional conduction in the partitions. This assumption is included . V ( T 0 - T„ i + is/ n=l t rc = l

in the following relations The total heat transfer, QT, is the sum of Q„ and Q/ which may be
written in the following form
ai(x,H) + 8a2{x, W) = Rx/L + 0fi(x); at y = H partition
LQT QaL Q/i t
(18a) (23)
/e0tf (To - T„) kaH(T0 - T„) - +/ ^- ( T o - T J rff
ai(x,H) + 8a2(x, W) =Rx/L + 8f2(x);aty = W partition
A simpler solution which disregards the transfer of heat from the slats
(186) to the trapped air may be derived by the equivalent parallel electrical
where
network method
8al(x,H)=±- CH8aldyimd8a2(x,W)=l- f*' 8a2dz LQT Ba B t
H Jo W Jo (24)
kaH(T0-T„) 1 + Ba l + BfrH
Substitution of 8ai, 8a2, 0/i, and 8f2 from equations (16) and (17) in
The ratio of Nu = QT/QT,U using equations (23) and (24), is computed
the above equations and utilization of the orthogonality properties
and presented in Table 1. The entries in Table 1 are prepared for a
of the sin (X m x) function results in
range of r between 0.025 and 1, fluid layer aspect ratio H/L between
ym coth (\mH) rym ym 0.05 and 2, and partition aspect ratio L/t between 50 and 5000. It is
+ B„ •R + Z A n f„
fimt\n KW/3n. n=l
reasonable to assume that the external heat transfer coefficient on
plate 2 is constant, at least within a small region —(H + t)<y<(H
(19a)
+ t). Since Ua = U; is a reasonable selection for most of the practical
ym coth (\mW) ry„ applications, it is selected for this parametric study. For the sake of
- + Bn = R+ D Bn$n brevity, data for only Ba = 1,2,5, and 10 are reported. When Ba —•
XmHj3„, fimt^n n=l
°°, the analytical solution degenerates into (Xn —*• nir/L and f„ —>- 0)
(196) a linear one-dimensional solution.
The method used for calculating the Fourier coefficient Am from An in-depth study of the influence of thermal radiation requires
equation (11), for the two-dimensional problem, now can be utilized a solution of an integro-differential equation and is beyond the scope
for determination of Am and Bm for the three-dimensional case. The of this paper. However, the radiation interchange among internal
result is surfaces of the cavity and exchange of radiation between plate 2 and
(2) (1)' surroundings will not influence the entries in Table 1 if the following
R & - L (Am,„<2> - Am,,
(1
>)f„ conditions are satisfied. The gas (air) does not participate in radiation
«=i exchange, the partitions are reradiating surfaces, and the net energy,
„(« (2) qr, leaving surface 2 is constant (independent of y). Accordingly, the
11 m
- V A ©f
n=i Pm n=l
boundary condition at x = L, equation (2), must be replaced by
fc (1) (2)
* V A (2) t (20a) -k • — Ua(T-T„) + qr
n=i . 2-i J*-m,7i J7 dy

Journal of Meat Transfer AUGUST 1981, VOL. 103 / 593

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 Table 1 Nu = Qr,2/Qr,i lor Ba = 1 Table 1 Nu = QT,ZIQT,\ for B„ = 2

H/L r L/t»50 100 200 500 1000 5000 H/L r L/t=50 100 200 500 1000 5000
0.0 5 0.025 1.131 1.500 1.1189 1.367 1. 218 1.0t>7
0.05 0.050 1.358 1.386 1.311 1.220 1. 1J5 1.033 0.05 0.025 1.637 1.760 1.713 1.536 1.31(1 1.090
0.05 0. 100 1.25b 1.219 1 . 1 % 1.111 1.063 1.011 0.05 0.030 1.199 1.511 1.176 1.297 1.178 1.012
0.05 0.200 1.110 1.122 1.0B6 1.1)1(11 1 . 0 2 1 1.005 0.05 0. 100 1.328 1.319 1.219 1.138 1.078 1.017
0.05 0. 250 1.106 1,089 1.061 1.030 1.016 1 . 0 0 1 0.05 0.200 1.162 1.110 1.099 1.051 1.028 1.006
0.05 0.500 1.026 1.020 1.013 1.006 1.003 1.061 0.05 0.250 1.118 1.099 1.068 1.031 1.018 1.001
0.05 1.000 1.000 1.000 1.000 1.000 1.000 1.000 0.05 0.500 1.026 1.020 1.013 1.006 1.003 1.001
0.05 1.000 1.000 1.000 1.000 1.000 1.000 1.000
0 . 10 0.025 1.177 1.169 1.391 ' 1.212 " I . " 116 1.034
0 . 10 0.050 1.370 1.329 1.215 1.133 1 . 0 7 5 1.017 0.10 b.655 1.713 1.706 1.5vll 1.339 1.199 1.016
0 . 10 0. 100 1.210 1. m i 1.127 1.0u3 1 . 0 3 1 1.007 0 . 10 0.050 1.517 1.157 1.332 1.175 1.097 1.021
0 . 10 0.<:00 1.118 1.001 1.052 1 . 0 2 1 1 . 0 1 3 1.003 0 . 10 0. 100 1.306 1.212 1.160 1.077 1.012 1.009
0. 10 0.250 1.000 1.060 1.036 1.016 1 . 0 0 9 1.002 0.10 0.200 1.136 1.097 1.060 1.027 1.011 1.003
0 . 10 0 . 5 00 1.019 1.012 1.007 1.003 1 . 0 0 2 1.000 0 . 10 0.250 1.096 1.066 1.010 1.018 1.010 1.002
0. 10 1.000 1.000 1.000 1.000 1.000 1 . 0 0 0 1.000 0. 10 0.500 1.019 1.012 1.007 1.003 1.002 1.000
0 . 10 1.000 1.000 1.000 1.000 1.000 1.000 1.000
0.20 0.025 1.133 1.365 1 . 2 6 1 1.112 1.080 1.010
0.20 0.050 0.20 0.025 1.639 1.531 1.373 1.193 1.106 1.023
1.308 1.232 1 . 1 5 1 1.073 1 . 0 3 9 1.00U
0.20 0. 100 0.20 0.050 1.122 1.313 1.200 1.095 1.051 1.011
1.1(11 1.123 1.071 1.033 1.017 1 . 0 0 1
0.20 0.200 0.20 0. 100 1.227 1.153 1.091 1.011 1.021 1.001
1.001 1.051 1.029 1.012 1.006 1 . 0 0 1
0.20 0.250 0.20 0.200 1.092 1.058 1.033 1.011 1.007 1.002
1.058 1.035 1.020 1.000 1.001 1.001
0.20 0.500 0.20 0. 250 1.063 1.039 1.022 1.009 1.005 1.001
1.012 1.007 1 . 0 0 1 1.002 1 . 0 0 1 1.000
0.2 0 1.000 0.20 0.500 1.012 1.007 1.001 1.002 1.001 1.000
1.000 1.000 1.000 1.000 1.000 1.000
0.20 1.000 1.000 1.000 1.000 1.000 1.000 1.000
0.25 0 . 0 25 1.101 1.321 1.226 1.117 1.065 1 . 0 1 1
0.25 0.050 1.270 1.200 1.126 1.060 1 . 0 3 2 1.007 0.25 0.025 1.587 1.165 1.315 1.158 1.0U6 1.019
0.25 0 . 100 1.159 1.103 1 . 0 6 1 1.027 1 . 0 1 1 1.003 0.25 0.050 1.377 1.268 1.166 1.077 1.011 1.009
0.2 5 0.200 .1.069 1.012 1.023 1.010 1.005 1 . 0 0 1 0.25 0.100 1.198 1.128 1.075 1.033 1.017 1.001
0.25 0.250 1.019 1.029 1.016 1.007 1.001 1.001 0.25 0.200 1.079 1.018 1.027 1.011 1.006 1.001
0.25 0.500 1.010 1.006 1.003 1.001 1 . 0 0 1 1.000 0.25 0.250 1.051 1.032 1.018 1.008 1.001 1.001
0.25 1.000 1.000 1.000 1.000 1.000 1.000 1.000 0.25 0.500 1.010 1.006 1.003 1.001 1.001 1.000
0.25 1.000 1.000 1.000 1.000 1.000 1.000 1.000
0.50 0.025 1.277 1.201 1.129 1.062 1.033 1.007
0.50 0.050 "6.50 0.025 1.392 1.279 1.175 1.082 1.011 1.009
1.178 1.116 1 . 0 6 9 1.031 1.016 1.003
0.50 0.100 0.50 0.050 1.236 1.152 1.089 1.010 1.021 1.001
1.095 1.057 1.032 1 . 0 1 1 1.007 1 . 0 0 1
0.50 0.200 0.50 0 . 100 1.117 1.070 1.039 1.017 1.009 1.002
1.010 1.023 1.012 1.005 1.003 1 . 0 0 1
0.50 0.250 0.50 0.200 1.015 1.026 1.011 1.006 1.003 1.001
1.028 1.015 1.008 1.003 1 . 0 0 2 1.000
0.50 0.500 0.50 0.250 1.030 1.017 1.009 1.001 1.002 1.000
1.005 1.003 1.002 1.001 1.000 1.000
0.50 1.000 0.50 0.500 1.005 1.003 1.002 1.001 1.000 1.000
1.000 1.000 1.000 1,000 1.000 1.000
0.50 1.000 1.000 1.000 1.000 1.000 1.000 1.000
1.00 0.025 1.159 1.109 1.067 1.032 1.017 1 . 0 0 1
1.00 0.050 1.099 1.062 1.035 1.016 1.008 1.002 1.00 0.025 1.220 1.119 1.091 1.012 1.022 1.005
1.00 0. 100 1.051 1.030 1.016 1.007 1 . 0 0 1 1.001 1.00 0.050 1.130 1.080 1.016 1.020 1.010 1.002
1.00 0.200 1.021 1.012 1.006 1.003 1 . 0 0 1 1.000 1.00 0. 100 1.0b3 1.037 1.020 1.008 1.001 1.001
1.00 0.250 1.011 1.008 1 . 0 0 1 1.002 1 . 0 0 1 1.000 1.00 0.200 1.021 1.013 1.007 1.003 1.001 1.000
1 .00 0.500 1.003 1.002 1.001 1.000 1.000 1.000 1.00 0.250 1.016 1.009 1.005 1.002 1.001 1.000
1.00 1.000 1.000 1.000 1.000 1.000 1.000 1 . 0 0 0 1.00 0.500 1.003 1.002 1.001 1.000 1.000 1.000
1.00 1.000 1.000 1.000 1.000 1.000 1.000 1.000
2.00 0.025 1.082 1.056 1 . 0 3 1 1.016 1. 00B 1.002
2.00 0.050 1.051 1.031 1.018 1.008 1.001 1.001 2.00 0.025 1.111 1.076 1.016 1.021 1.011 1.002
2.00 0.100 1.026 1.015 1.008 1.003 1 . 0 0 2 1.000 2.00 0.050 1.067 1.011 1.023 1.010 1.005 1.001
0.200 1.011 1.006 1.003 1 . 0 0 1 1 . 0 0 1 1.000 2.00 0.100 1.032 1.019. 1.010 1.001 1.002 1.000
2.00
0.250 1.007 1.001 1.002 1.001 1.000 1.000 2.00 0.200 1.012 1.007 1.001 1.001 1.001 1.000
2.00
0.500 1.001 1.001 1.000 1.000 1 . 0 0 0 1.000 2.00 0.250 1.008 1.001 1.002 1.001 1.000 1.000
2.00 2.00 1.001 1.001 1.000 1.000 1.000 1.000
1.000 1.000 1.000 1.000 1.000 1.000 1.000 0.500
2.00 2.00 1.000 1.000 1.000 1.000 1.000 1.000 1.000

= Ua(T-Tl) Cavity," International Journal of Heat and Mass Transfer, Vol. 11,1968, pp.
1675-16.
3 Yin, S. H., Wang, T. Y., and Chen, K., "Natural Convection in an Air
where T„* = T „ — qr/Ua- Therefore, throughout the calculations, T„* Layer Enclosed within Rectangular Cavities," International Journal of Heat
instead of T«, must be used. While QT and QT,I are affected by T„,*, and Mass Transfer, Vol. 21,1978, pp. 307-315.
the entries in Table 1 remain unchanged. 4 Arnold, J. N., Edwards, D. K., Catton, I., and Edward, D. K., "Experi-
mental Investigation of Natural Convection in a Finite Rectangular Region
The fin effect at the partition causes the heat flux at the base of the Inclined at Various Angles from 0° to 180"," Proceedings of the 1974 Heat
partition (x = 0) to increase and the heat flux at x = 0 in the two- Transfer and Fluid Mechanics Institute, edited by L. R. Davis and R. E. Wil-
dimensional layer to decrease. Therefore, the entries in Table 1 are son, Stanford University Press, 1974, pp. 321-329.
a measure of departure from one-dimensionality due to net fin effect. 5 Arnold, J. N., Edwards, D. K., and Catton, I., "Effect of Tilt and Hori-
According to Table 1, the fin effect is negligible when Ba -»• 0, Ba -"- zontal Aspect Ratio on Natural Convection in Rectangular Honeycombs,"
ASME JOURNAL OF HEAT TRANSFER, Vol. 99,1977, pp. 120-122.
°°, H/L —• c°, and r —>-1. The largest increase in heat flux is observed 6 Hollands, K. G. T., "Natural Convection in Horizontal Thin-Walled
for Ba in the neighborhood of 5. As H/L -* 0, N 8 approaches unity Honeycomb Panels," ASME JOURNAL OF HEAT TRANSFER, Vol. 95,1973, pp.
if H/t remains constant. The value of H/t = (L/t)-(H/L) for a few 439-444.
entries in Table 1 is small for some applications. Moreover, it should 7 Cane, R. L. D., Hollands, K. G. T., Raithby, G. D., Unny, T. E., "Free
Convection Heat Transfer Across Inclined Honeycomb Panels," ASME
be noted that even for a small H/t the quasi-one-dimensional criterion JOURNAL OF HEAT TRANSFER, Vol. 99,1977, pp. 86-91.
is satisfied (see Appendix). According to entries in Table 1, when the 8 Hollands, K. G. T., Unny, T. E., Raithby, G. D., Konicel, L., "Free
two-dimensional region is air and the partition is nonmetallic, the Convection Across Inclined Air Layers," ASME JOURNAL OF HEAT TRANS-
departure from one-dimensionality may exceed 25 percent for certain FER, Vol. 98,1976, pp. 189-193.
variables. The high thermal conductivity partitions cause a larger 9 Smart, D. R., Hollands, K. G. T., and Raithby, G. D., "Free Convection
Heat Transfer Across Rectangular-Celled Diathermanous Honeycombs,"
departure from one-dimensionality. Therefore, the one-dimensional ASME JOURNAL OF HEAT TRANSFER, Vol. 102,1980, pp. 75-80.
solution defined by equation (24) is generally satisfactory except for 10 Arpaci, Vedat, Conduction Heat Transfer, Addison-Wesley, 1966.
small values of r and H/L, and L/t < 1000. 11 Kaplan, Wilfred, Advanced Calculus, Addison-Wesley, 1956.
The computations were carried out on the IBM 370/155 digital
computer. The total computation time, using a double precision code,
APPENDIX
for one value of Ba and 500 eigenvalues, was 15 min. The partitions, Fig. 1, should be sufficiently thin so that the average
temperature variation AT across the partition in y-direction is small
References in comparison with the size of T 0 — To,. A bound on the value of | AT|
1 Batchelor, G. K., "Heat Transfer by Free Convection Across a Closed can be estimated if we consider 7/ not to be one-dimensional
Cavity between Vertical Boundaries at Different Temperatures," Quarterly
of Applied Mathametics, Vol. 12, No. 3,1954, pp. 209-233. dT/,2
2 Davis, G. D., "Laminar Natural Convection in an Enclosed Rectangular dx x=0 i>x x=L

594 / VOL. 103, AUGUST 1981 Transactions of the ASME

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 Table 1 Nu = Q T , 2 / Q T , I 'or Ba = 5 Table 1 Nu == 0 T ,2/Qr,i for Ba = 10
H/L r L/t=50 100 200 500 1000 5000 H/L r L/t=50 100 200 500 1000 5000
0.0 5 0.02!) 1. 790 1. 906 1.974 1. b 8 2 1 . 4 30 1 . 10b 0.05 1.65b 1.417 1. 104
0. 025 1.72b 1 . 9 14 1 . 9 18
0.0 5 0 . 0 50 1.558 1 . t> 1 9 1.543 .335 1 . 198 1. 04b 0. 05J 1. 458 1. 5 1 5 1. 176
0.05 1.461 1.293 1.042
0.0 5 0 . 100 1 . J 17 1 . J 1.J 1.24b .138 1. 078 1.01 / 0. U J 1 . 109 1. 063 1 . 0 14
0.05 1.234 1 .2ib 1 . 189
0.0 5 0 . 200 1. 115 1 . 11 U 1.005 .044 1. 0 2 4 1.005 1.091 1. 08 1
0.05 0. 230 1 . 059 1.031 1 . 0 17 1.004
0.0 5 0 . 250 1. 0 9 4 1.079 1 .055 1 . 0 28 1. 0 1 5 1.O03 0.05 0.250 1.Ool 1 . 05 3 1 . 0 19 1 . 0 11 1.002
1.037
0.0 5 0.500 1 . 0 111 1.014 1 .009 1.004 1.002 1.000 0.05 1.000
u.'jOO 1 . 0 11 1.008 1 .005 1.003 1. 0 0 1
0.05 1.000 1.000 1.000 1.000 .000 1. 000 1.000 1. 0 0 0 1.000
0.05 1 . JJ0 1.000 1.000 1 . 0 00 1 .000
0 . 10 0.02 5 1. 902 1.908 1 . 7 10 . 4 18 1.240 1.054 0 . 10 0.025 1.810 I . 84 1 1 .693 1.404 1. 233 1.053
0 . 10 0.05 0 1. 575 1. 5 1 5 1.373 1.19 5 1 . 107 1.023 0 . 10 0.050 1 . 4o7 1. 4 3 1 1.322 1 . 172 1. 09b 1.021
0 . 10 J . 100 1.215 1.23b 1.158 . 0 77 1. 0 4 1 1. 009 0.10 0 . 100 1.217 1.179 1 . 123 l.Obl 1.033 1 . 0 07
0 . 10 0 . 20 0 1. 1 1 J 1. 082 1.051 .024 1 . 0 13 1.003 0 . 10 0 . 2 00 1 . 0 75 1.056 1.03b 1.017 1. 0 0 9 1.002
0 . 10 0 . 250 1 . 0 70 1.053 1.033 .015 1. 0 0 8 1.002 0 . 10 0. 250 1 . 0 49 1.036 1.022 1.010 1. 005 1.0O1
0 . 10 J.500 1 . 0 1J 1 . OOi 1.005 1.002 1.00) 1.000 0 . 10 0.500 1.008 1. 0 0 5 1.003 1.001 1. 0 0 1 1.000
0.10 1 . 000 1.000 1. 000 1 .000 .000 1. 000 1.000 0. 1 0 1.000 l.OOU 1. 0 0 0 1.000 1.000 1. 0 0 0 1.000
0.2 0 0.025 1. 707 1. bb5 1.459 .231 1. 12b 1.027 0.20 0.025 1. 721 1. 624 1 .440 1.225 1. 123 1.027
0.2 J 0.050 1.405 1.349 1.222 . 105 1. 056 1.012 0.20 O.OJO 1.383 1. 2 9 8 1 . 195 1.094 1. 0 5 0 1.011
0.20 0 . 100 1. 219 1. 150 1.090 1.041 1.021 1.004 0.20 0 . 100 1. 103 1. 1 1 5 1.071 1.033 1 . 0 17 1.004
0.20 0.20O 1.077 1.049 1.028 1.012 1. 00b 1.001 0.20 0 . 200 1.052 1. 0 3 4 1.020 1.009 1. 005 1.001
0.20 0.2501 1.050 1.031 1.018 .008 1. 004 1.001 1.005 1.001
0.20 u . 250 1.033 1. 0 2 1 1.012 1. 003
0.20 J .500 1.008 1.005 1.003 1.001 1. 00 1 1.000 0.20 0. 5 0 0 1.005 1. 0 0 3 1.002 1.001 1. 000 1.000
0.20 1.000 1.000 1.000 1.000 .000 1. 000 1.000 1.000 1. 000 1.000
0.20 1.0J0 1.000 1. 000 1 .000
0.2 5 0.025 1. 727 1.577 1.385 . It)9 1. 102 1.022 0.25 0.025 1.603 1. 5 4 5 1.370 1 . 184 1 . 100 1.022
0.25 0.050 1 . 4 15 1.297 1 . 184 .005 1. 045 1.009 1.07o 1. 009
0.25 0.050 1.344 1. 2 5 6 1 . 1b2 1. 0 4 1
0.25 0 . 100 1 . 141 1 . 12b 1.074 .033 1. 0 17 1.004 1 . 0 26 1. 0 1 4 1.003
0.25 0 . 1 JJ 1. 143 1. 0 9 7 1 .058
0.25 0 . 20 0 1. Oob 1.041 1.023 .010 1. 005 1.001 1 .007 1. 004 1.001
0.25 0.2JO 1.044 1.02 8 1.01b
0.25 0.250 1. 013 1. 02b 1.014 .00b 1. 0 0 3 1. 001 0.25 0.250 1.0 28 1. 0 1 7 1.010 1.004 1 . 0U2 1.000
0.2 5 0.500 1.007 1.004 1.002 .001 1.000 1.000 1. 000 1.000
0.25 0.300 1.004 1. 0 0 2 1.001 1.001
0.2 5 1.000 1 . 000 1. 000 1 .000 .000 1. 000 1.000 0.25 1.000 1.000 1. 000 1.000 1 .000 1. 0 0 0 1.000
0.50 0.025 1.479 1.339 1.210 .098 1.652 • 1 . 6 1 1 ' 1.09b 1.051 1 . 0 11
0.5 0 0. J 2 J 1.449 1.32b 1.204
0.50 0.050 1.260 1. IbH 1.098 .044 1. 0 2 3 1.005 1.004
0.50 0.050 1.221 1. 148 1.087 1.039 1.021
0.5 0 0 . 100 1 . 1 14 1. 0 b 9 1.039 .017 1.009 1.002 1.001
0.50 0.100 1.087 1.054 1.031 1.014 1. 0 0 7
0.^0 0.200 1.0J7 1.022 1.012 .005 1.003 1.001 0.50 0.200 1.025 1. 0 1 5 1.008 1.004 1.002 1.000
0.50 0.250 1.024 1.014 1.007 .003 1. 0 0 2 i.ooa 0.50 0.250 1.01b 1. 0 0 9 1 .005 1.002 1. 0 0 1 1.000
0.5 0 0.500 1.004 1. 0 0 2 1.001 .000 1. 000 1.000 0.50 0.500 1.002 1. 0 0 1 1.001 1.000 1.000 1.000
0.50 1.000 1.000 1.000 1.000 .000 1.000 1.000 1.000 1. 000 1.000
0.50 1.03O 1.000 1.000 1.000
1.00 6.025 1.271 1.169 .656
r.182 1. 02b 1.006 1.00 0.025 1.263 • . 178 1.107 1.049 1. 026 1 . 0 05
1.00 0.050 1. 144 1. 0b9 1.050 I.022 1. 011 1.002 1.020 1. 0 1 0 1.002
1.00 0.050 1.12b 1.079 1.045
1.00 o . 100 1.002 1.03b 1.020 .008 1.004 1.001 1.007 1.004
1.00 0 . 130 1.040 1. 0 2 9 1.016 1.001
1.00 0 . 200 1.020 1. 01 1 1 .00b 1.003 1. 001 1.000 1.00 0.200 1.014 1. o o a 1.004 1.002 1. 0 0 1 1.000
1.00 0.250 1.013 1. 007 1.004 .002 1. 0 0 1 1.000 1.001 1. 0 0 1 1.000
1.00 0.250 1.008 1. 6 0 5 1.003
1.00 0.500 1.002 1 . OP 1 1.001 .000 1.000 1.000 1.00 0.500 1.001 1. 0 0 1 1.000 1.000 1.000 1.000
1.00 1.00 0 1 . 000 1. 000 1.000 .000 1.000 1. 000 1.000 1.000 1.000
1.00 1.000 1.000 1. 0 0 0 1.000
2.00 0.025 1 . 143 1.093 1.055 .025 '1.013 1.003 2.0o 0.025 1.142 . 093 1.055 1.025 1. 0 13 1.003
2.00 0.050 1 . 0 75 1. 04b 1.026 .011 1. 0 0 b 1.001 2.00 0. 050 1.067 1.041 1.023 1.010 1. 005 1.001
2.00 0 . 100 1. 032 1.018 1.010 .00 4 1. 002 1.000 1.002 1,000
2.00 0 . 130 1.025 1. 0 1 5 1.008 1.003
2.00 0.200 1.010 1. 00b 1.003 1.001 1. 0 0 1 1.000 2.00 0.200 1.007 1. 0 0 4 1.002 1.001 1. 0 0 0 1.000
2.00 u . 250 1.00b 1.004 1.002 .001 1.000 1.000 1.001 1. 0 0 0 1.000
2.00 0.250 1.004 1. 0 0 2 1.001
2.00 0.50 0 1. 001 1.001 1 . 0 00 .000 1.000 1.000 2.00 0.500 1.001 1. 0 0 0 1.000 1.000 1.000 1.000
2.00 1.000 1.000 1. 000 1.000 .000 1.000 1.000 1.000 1.000
2.00 1.000 1.006 1. 0 0 0 1.000 1.000

It can be noted that Ba Bf

aTM dT/,2 1-K 1 + Bf,
lx=0 < and >
dx dx dx x=L dx when Uf = Ua and kf ->• ka, the quantity in the square brackets will
vanish and since this quantity is always less than unity, the rela-
then
tion
AT I ATI
\T0-T„\ <
\L,
LQa.l
Hka(T0-T„)
LQf,i
tkf(T0 - T„) |r 0 -T„
<z
is used as a guideline for satisfying the quasi-one-dimensional con-
The terms in the square brackets are readily available by elementary dition in the partitions. The largest | A T | / | T 0 - T«,\ for entries in
methods, hence, the final result is Table 1 is less than 0.0004.

Journal of Heat Transfer AUGUST 1981, VOL. 103 / 595

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 /-

X * S

A Closed-Form Analytical A schematic diagram of the physical situation is pictured in Fig.

1. As shown there, a liquid phase-change medium at its saturation
Solution for Freezing Adjacent temperature T* is contained in a space which is relatively long in the
to a Plane Wall Cooled by direction into and out of the plane of the figure. The space is bounded
Forced Convection at the left by a parallel plate channel having an interplate spacing W.
For times t < 0, the entire system is at a uniform temperature T* and,
C. F. Hsu 1 and E. M. Sparrow 1 in addition, there is no flow in the channel. Then, at t = 0, a coolant
with an inlet temperature Tbi < T* is caused to flow in the channel,
Nomenclature and freezing takes place along the wall.
Bi = Biot number, hW/k Figure 1 shows an instantaneous view of the frozen layer. Owing
c p = specific heat to the rise of the bulk temperature in the streamwise direction, the
Fo = Fourier number, a t / W 2 thickness of the frozen layer will diminish in that direction, as por-
h = coolant heat transfer coefficient trayed in the figure. Note also that there will be no natural convection
k = thermal conductivity in the liquid phase-change medium since its temperature is uni-
St = Stanton number, h/u(pcp)c form.
Ste = Stefan number, cp{T* - T bi )/X The closed-form solution given here should be useful in its own right
Tb = coolant bulk temperature as a guide to design and, in addition, it can serve as a standard of
TM = inlet coolant temperature comparison against which to test approximate solution methods.
Tw = temperature at channel wall on which freezing occurs
T* = fusion temperature Analysis
t = time The first step in the analysis is to write an energy balance for a
u = mean velocity of coolant control volume which spans the channel cross section and is of length
W = channel width, Fig. 1 dx in the flow direction. The relevant energy quantities include the
X = dimensionless axial coordinate, x/W enthalpy convected into and out of the control volume, the heat stored
x,y = coordinates, Fig. 1 in the fluid, and the convective heat transfer from the wall to the fluid.
a = thermal diffusivity In mathematical terms,
A = dimensionless layer thickness, 5/W
5 = frozen layer thickness, Fig. 1 (pcp)cW(ui>Tb/dx + i>Tb/dt) = h(Tw - Tb) (1)
r) = transformed coordinate, equation (7) where the subscript c denotes properties of the coolant fluid. As in-
6 = dimensionless temperature, (T - T*)l(Tbi - T * ) dicated in Fig. 1, both Tm and Tb are variables and are unknown. The
A = latent heat of fusion coolant flow is assumed to be fully turbulent and hydrodynamieally
p = density developed, and h will be assumed to be constant. Note that h may also
T = dimensionless time, FoSte encompass the transverse conductance of the wall.
U = coolant heat-capacity parameter, equation (6) For the freezing problem, two highly plausible simplifications will
be made. The first is that for moderate values of the Stefan number
Subscripts that lie in the practical range (Ste < 0.5), sensible heat effects do not
c = properties of coolant materially affect the freezing-induced heat flow through the solid. The
properties of frozen layer are unsubscripted

Introduction
The analysis of multi-dimensional freezing on a wall that is con-
vectively cooled from the rear has evoked recent interest because of
its relevance to phase-change thermal energy storage (e.g., [1, 2]). The
published solutions have involved either finite difference calculations
or other numerical work. There is, however, a particular case for which
a closed-form analytical solution can be obtained, subject to certain
reasonable and widely accepted assumptions about the physical
model. The case in point is the two-dimensional freezing on a plane
wall wherein the rear face of the wall is one of the boundaries of a
parallel plate channel through which a coolant passes.

1
Department of Mechanical Engineering, University of Minnesota, Min-
neapolis, Minn. 55455
Contributed by the Heat Transfer Division for publication in the JOURNAL
OF HEAT TRANSFER. Manuscript received by the Heat Transfer Division
September 12,1980. Fig. 1 Freezing on a wall that is convectively cooled from the rear

596 / VOL. 103, AUGUST 1981 Transactions of the ASME

Copyright © 1981 by ASME
Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 second is that the axial conduction through the solid is negligible Once G has been found, the closed-form solution for A follows from
compared with the transverse conduction, a condition akin to the (12) as
well-established boundary layer model. This condition should be fully
satisfied for channels that are not too short. A + InA/Bi = - ( S t / B i ) X + A 0 + lnA 0 /Bi (14)
A consequence of the foregoing simplifications is that at any x Furthermore, considering equations (10), it is seen that the solutions
dT/dy = (T* - Tw)lb for 6W and 6b are obtained by differentiating equation (14), so that
(2)
and, in addition, : A(dA 0 /dj7)g(A 0 )/g(A), : A(dA0/d7))g(Ao) (15)

h(Tw - Tb) = k(dT/dy) = p\(d8/dt) where

(3)
The solution task involves the three coupled partial differential « W = 1 + 1/Bii//, dAo/djj = ((1/Bi) 2 + 2TJ) -1/2 (16)
equations (1) and (3) which contain Tb,Tw, and 5 as unknowns and Equations (14) and (15) convey the closed-form solutions for A, 8W,
x and t as independent variables. Dimensionless variables are first and 8b. The only visible parameters are the Stanton and Biot num-
introduced as bers, St and Bi, but the Q, parameter is embedded in the 7/ variable.
X = x/W, A = b/W, (T - T*)/(Tbi - T*), T = FoSte (4) If E denotes the energy extracted from the freezing solid during a
time interval t = 0 to t = t and over a length of channel from x = 0 to
where x = x, then
Fo = at/W2, Ste = c p ( T * - T w ) / X (5) E/pXW2 = P AdX (17)
Jo
Unsubscripted properties correspond to the solid. The dimensionless
parameters which are needed to characterize the solution are which requires numerical integration for its evaluation.
St = h/u(pcp)c, Bi = hW/k, fi = ((pc p ) c /pc p )(Ste/Bi) (6)
Results and Discussion
Care should be taken not to confuse the Stefan number Ste and the In view of the space limitations inherent in a Note, only illustrative
Stanton number St. results will be presented in order to identify trends. It should be noted,
In addition to the dimensionless variables and parameters, a major however, that the availability of the closed-form algebraic solution,
step forward toward the solution is accomplished by a change of as detailed in the foregoing, enables results to be readily obtained for
variables from X , T to X,T) such that any values of the three parameters St, Bi, and £2.
Figure 2 deals with parameter ranges representative of turbulent
X = X, 7] = T- (StQ)X (7) gas flows (e.g., air) as coolants. The figure is subdivided into two
Application of the change of variables to equations (1) and (3) panels, respectively for Bi = 2 and 20. In each panel, for each of a
yields succession of parametric times T, the frozen layer thickness A is
plotted as a function of the distance X along the channel (T, A, and
(d0 6 /dX)/St = ew-8b (8) X are dimensionless quantities defined in equation (4)). The solid and
dashed curves correspond respectively to Stanton numbers St = 0.002
Bi(9b -Bw) = 6JA = dA/di; (9)
and 0.004.
where, by dint of the — (StQ)X part of r), the storage term has been It is relevant to note that the parameter fi does not appear in Fig.
eliminated from equation (8). Then, upon rearrangement of equation 2. The absence of fi can be traced back to its role as a measure of the
(9) heat capacity of the coolant. On physical grounds, it can be expected
that the heat capacity of a gaseous coolant will play a negligible role.
eio=i(dAVdrl), db = i(dA2/dV) + (l/Bi)(dA/d7)) (10)
To verify this expectation numerically, results were obtained for fi
By employing equations (9) and (10), 8W and 8b can be eliminated values representative of the conditions for the left-hand and right-
from equation (8) in favor of A. The result of these eliminations yields hand panels, fi = 0.0002 and 0.00002, respectively, and for Q = 0 as
an equation of the form d/dr](. . .) = 0, which, after integration, be- a comparison case. Within the scale of the figure, the A versus X re-
comes sults for fi ^ 0 and Q = 0 are coincident.
By inspecting the figure, it is seen that for the conditions of the
1 A2 A
d
— + — + | = F(X) (ID left-hand panel, the frozen layer thickness decreases in the flow di-
StdX 2 Bi rection at any fixed time. Whereas this same trend is in evidence in
where F{X) is a function of integration. Thus, although A and dA/dX the right-hand panel, the extent of the decrease is much smaller. The
are functions of r\ as well as of X, equation (11) states that the left- decrease of the layer thickness is a reflection of the increase of the
hand side is independent of r\. Thus, it can be evaluated at any con-
venient value of 7], i.e., where A and dA/dX are known. SI
|Bi=?| 0.002 |Bi = 20|
It will now be shown that A = dA/dX = 0 at r\ = 0, so that F(X) = 0.004
T -- 1.0
0. It can easily be verified that r) = 0 corresponds to an axial station T=0.50
x = ut. If account is taken of the slug-flow model implied by equation
0.75
(1), then x = ut corresponds to the penetration distance of the coolant ^~— = -
into the channel at time t. Since there is no freezing for x > ut in the
0.25
absence of axial conduction, the foregoing assertion is substan- 0.50
tiated.
^~ -
With F(X) = 0, equation (11) can be integrated to give 0.10
„„
A/St + InA/BiSt + X/Bi = G(?j) (12) 0.05

where G(?j), the function of integration, states that the left-hand side
- O.IO 0.025

is independent of X. It is especially convenient to evaluate the left-

1 i i i i i i i
hand side at X = 0 and, thereby, to find G. The solution for A at X 20 40 60 80 100
= 0 is denoted by A0. It is found by setting 6b = 1 in equation (10) and O 20 40 60 80 100
X
integrating, with the end result X
Fig. 2 Distributions of frozen layer thickness for conditions representative
A 0 = - 1 / B i + ((1/Bi) 2 + 2T,) 1/2 (13) of turbulent gas flows as coolants

Journal of Heat Transfer AUGUST 1981, VOL. 103 / 597

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 St x I03 fl case in which the coolant is a liquid such as water and the flow is
St x 10s fl turbulent. The structure of the figure is similar to that of Fig. 2. For
1.2 - 0.75 - 1.5 0 the conditions depicted in the left-hand panel, the coolant heat ca-
1.0 0.02 0.75 - 1.5 0.004
1.5 0.02 pacity parameter Q is ~0.02, while an appropriate ft value for the
right-hand panel is ~0.004. Results for these 0 values are shown in
T=0.5 T = 0.5 the respective panels along with those for which the coolant heat ca-
0.8 -1 3i = 4 0 | |Bi = 2 0 0 | pacity is neglected (0 = 0).
0.25 0.25 The figure shows that only for very short times does the coolant heat
capacity play a role, and that role is heightened at higher Stanton
0.6 - numbers (i.e., lower Reynolds numbers). However, as the freezing
0.10 0.10 process proceeds (i.e., at larger times), the heat capacity effect wanes.
0.4
" 0.05
Note also that at higher Biot numbers, the dominance of the solid-
0.05 side resistance blunts the heat capacity effect, even at early times.
0.025 0.025
0.2 It may also be noted from the figure that aside from the afore-
0.005 0.005 mentioned early-time heat capacity effect, the frozen layer thickness
_i- — 1 l - i ~~" i i i i is nearly independent of x. This behavior is due to the relatively slight
20 40 60 80 100
0 20 40 60 80 streamwise temperature rise experienced by the liquid coolant and
X
X to the dominance of the solid- side thermal resistance. With increasing
Fig. 3 Distributions of frozen layer thickness for conditions representative Biot number, the results tend to approach those of the one-dimen-
of turbulent liquid flows as coolants sional Stefan solution, which gives A = (2r) 1 / 2 .

coolant temperature in the flow direction. The coolant temperature Acknowledgment

rise becomes larger at lower Reynolds numbers (i.e., larger Stanton This research was performed under the auspices of the U.S. De-
numbers), and this is why the dashed curves drop off to a greater ex- partment of Energy (DOE/DE-ACO2-79ER10343).
tent than do the solid curves.
At larger Biot numbers (right-hand panel), the thermal resistance References
of the frozen layer controls the rate of heat transfer, and this dimin- 1 Green, T. F., and Vliet, G. C, "Thermal Response of a Latent Heat
ishes the impact of the coolant temperature rise on the freezing rate. Storage Unit: An Analytical and Experimental Investigation," ASME Paper
Consequently, there is only a slight dropoff in the frozen layer No. 79-HT-36,1979.
2 Shamsundar, N., and Srinivasan, R., "Effectiveness-NTU Charts for
thickness in the streamwise direction. Latent Heat Storage Unite," ASME Journal of Solar Energy Engineering, Vol.
Attention is now turned to Fig. 3, where results are plotted for the 102, No. 4,1980, pp. 263-271.

Phase Change around a Circular a = 8/R

Cylinder p = density
ait
V. J. Lunardini 1 T = — ST, dimensionless time
n?
„2
Nomenclature Subscripts
A = area of phase change interface 0 = initial
c = mass specific heat 1 = region 1
k = thermal conductivity 2 = region 2
£ = latent heat of fusion, mass basis / = fusion of frozen
q = surface heat transfer rate s = surface
qci , dimensionless heat transfer t = thawed
2TTM
r = radial coordinate Introduction
ro- = radius of cylinder The question of freezing and thawing of soil systems is important
;
R = radius of phase change location for engineering design in permafrost regions. Thus solutions to con-
= ci(7> - Ts)/£ freeze duction problems with moving boundaries are sought for application
ST to analysis and synthesis. Interest in these problems has also been
ci(T s - Tj)/e thaw
stimulated by the requirements of latent heat storage in solar energy
time systems.
r = temperature The homogeneous, semi-infinite medium, initially at a temperature
X£ = =volumetric water content of soil different than the fusion value, has a general solution for a step change
a = thermal diffusivity of surface temperature, Neumann [1], generalized in Carslaw and
«12 = a\lai Jaeger [2]. Recently, Lunardini and Varotta [3] derived an accurate,
R/ro dimensionless phase change position but approximate, solution to this problem by using the heat balance
temperature penetration radius integral technique. Yuen [14] also used the heat balance integral with
T0-T/ collocation to generate a less accurate approximation.
— — superheat or subcooling parameter Unfortunately, there are no general solutions for phase change in
cylindrical coordinates. Since this is an important geometry for
practical applications a significant effort has been expended upon
1 approximate and numerical solutions. The geometry of the system,
U.S. Army Cold Regions Research and Engineering Laboratory, Hanover,
N.H. Fig. 1, is that of a circular cylinder surrounded by an infinite, homo-
Contributed by the Heat Transfer Division and presented at the 19th geneous, medium. Tien and Churchill [4] numerically solved the
AIChE/ASME Heat Transfer Conference, Orlando, Fla., July 27-30, 1980.
Revised manuscript received by the Heat Transfer Division December 15, freezing problem using interpolation near the phase change nodes.
1980. More recently, Sparrow, et al. [5] examined the same problem for

598 / VOL 103, AUGUST 1981 Transactions of the ASME

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 St x I03 fl case in which the coolant is a liquid such as water and the flow is
St x 10s fl turbulent. The structure of the figure is similar to that of Fig. 2. For
1.2 - 0.75 - 1.5 0 the conditions depicted in the left-hand panel, the coolant heat ca-
1.0 0.02 0.75 - 1.5 0.004
1.5 0.02 pacity parameter Q is ~0.02, while an appropriate ft value for the
right-hand panel is ~0.004. Results for these 0 values are shown in
T=0.5 T = 0.5 the respective panels along with those for which the coolant heat ca-
0.8 -1 3i = 4 0 | |Bi = 2 0 0 | pacity is neglected (0 = 0).
0.25 0.25 The figure shows that only for very short times does the coolant heat
capacity play a role, and that role is heightened at higher Stanton
0.6 - numbers (i.e., lower Reynolds numbers). However, as the freezing
0.10 0.10 process proceeds (i.e., at larger times), the heat capacity effect wanes.
0.4
" 0.05
Note also that at higher Biot numbers, the dominance of the solid-
0.05 side resistance blunts the heat capacity effect, even at early times.
0.025 0.025
0.2 It may also be noted from the figure that aside from the afore-
0.005 0.005 mentioned early-time heat capacity effect, the frozen layer thickness
_i- — 1 l - i ~~" i i i i is nearly independent of x. This behavior is due to the relatively slight
20 40 60 80 100
0 20 40 60 80 streamwise temperature rise experienced by the liquid coolant and
X
X to the dominance of the solid- side thermal resistance. With increasing
Fig. 3 Distributions of frozen layer thickness for conditions representative Biot number, the results tend to approach those of the one-dimen-
of turbulent liquid flows as coolants sional Stefan solution, which gives A = (2r) 1 / 2 .

coolant temperature in the flow direction. The coolant temperature Acknowledgment

rise becomes larger at lower Reynolds numbers (i.e., larger Stanton This research was performed under the auspices of the U.S. De-
numbers), and this is why the dashed curves drop off to a greater ex- partment of Energy (DOE/DE-ACO2-79ER10343).
tent than do the solid curves.
At larger Biot numbers (right-hand panel), the thermal resistance References
of the frozen layer controls the rate of heat transfer, and this dimin- 1 Green, T. F., and Vliet, G. C, "Thermal Response of a Latent Heat
ishes the impact of the coolant temperature rise on the freezing rate. Storage Unit: An Analytical and Experimental Investigation," ASME Paper
Consequently, there is only a slight dropoff in the frozen layer No. 79-HT-36,1979.
2 Shamsundar, N., and Srinivasan, R., "Effectiveness-NTU Charts for
thickness in the streamwise direction. Latent Heat Storage Unite," ASME Journal of Solar Energy Engineering, Vol.
Attention is now turned to Fig. 3, where results are plotted for the 102, No. 4,1980, pp. 263-271.

Phase Change around a Circular a = 8/R

Cylinder p = density
ait
V. J. Lunardini 1 T = — ST, dimensionless time
n?
„2
Nomenclature Subscripts
A = area of phase change interface 0 = initial
c = mass specific heat 1 = region 1
k = thermal conductivity 2 = region 2
£ = latent heat of fusion, mass basis / = fusion of frozen
q = surface heat transfer rate s = surface
qci , dimensionless heat transfer t = thawed
2TTM
r = radial coordinate Introduction
ro- = radius of cylinder The question of freezing and thawing of soil systems is important
;
R = radius of phase change location for engineering design in permafrost regions. Thus solutions to con-
= ci(7> - Ts)/£ freeze duction problems with moving boundaries are sought for application
ST to analysis and synthesis. Interest in these problems has also been
ci(T s - Tj)/e thaw
stimulated by the requirements of latent heat storage in solar energy
time systems.
r = temperature The homogeneous, semi-infinite medium, initially at a temperature
X£ = =volumetric water content of soil different than the fusion value, has a general solution for a step change
a = thermal diffusivity of surface temperature, Neumann [1], generalized in Carslaw and
«12 = a\lai Jaeger [2]. Recently, Lunardini and Varotta [3] derived an accurate,
R/ro dimensionless phase change position but approximate, solution to this problem by using the heat balance
temperature penetration radius integral technique. Yuen [14] also used the heat balance integral with
T0-T/ collocation to generate a less accurate approximation.
— — superheat or subcooling parameter Unfortunately, there are no general solutions for phase change in
cylindrical coordinates. Since this is an important geometry for
practical applications a significant effort has been expended upon
1 approximate and numerical solutions. The geometry of the system,
U.S. Army Cold Regions Research and Engineering Laboratory, Hanover,
N.H. Fig. 1, is that of a circular cylinder surrounded by an infinite, homo-
Contributed by the Heat Transfer Division and presented at the 19th geneous, medium. Tien and Churchill [4] numerically solved the
AIChE/ASME Heat Transfer Conference, Orlando, Fla., July 27-30, 1980.
Revised manuscript received by the Heat Transfer Division December 15, freezing problem using interpolation near the phase change nodes.
1980. More recently, Sparrow, et al. [5] examined the same problem for

598 / VOL 103, AUGUST 1981 Transactions of the ASME

Copyright © 1981 by ASME
Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 £n\
8-r \S,
T2 = T0 + (Tf - To) (5)

'8
Equation (4) exhibits the correct analytical form for the quasi-steady
solution of equation (1). This is valid as ST -»• 0, contributes to the
accuracy of the results, and suggests a method to choose the tem-
perature approximation.
Equations (1-5) lead to the following two coupled equations,
written in nondimensional form.

c dF 1
!
(6)
«i2 ST — = 1
dr Q- 1 into
Fig. 1 Geometry of cylindrical system ST 1_P
\2£nB
t
4B(£nB)2,
dB
+8 — =
1 ,
k2i<t>
1
+
1
dr £nB \fi - 1 £nQ
melting via a finite difference solution with phase boundary immo-
bilization although the properties of the two phases were assumed (7)
equal. Both used the plane (Neumann) solution to start the calcula- B2 B2
tion and both illustrate the usual problem of elaborate programming F(P, fi) = ^ - + ^ --(fl*-l)
2 (1 - Q)£nQ 4
necessary for numerical solutions to Stefan problems. Bell [6] used
the heat balance integral method, with an arbitrary number of regions, /fl 1\ 1
+ \£nQ, + - (fi 3 - 1) (8)
to solve the solidification problem when the initial temperature was \2 3/ 9
at the fusion value. The technique simplifies the choice of the assumed The instantaneous heat flow from the cylinder is given by
temperature profile, but introduces a set of simultaneous, first order
differential equations to solve. This tends to negate the basic sim- dTi(r 0 , t)
q = —k\A - (9)
plicity of the heat balance integral technique and the single phase dr
problem is not a good test of the computational complexity. Using equation (4), the nondimensional heat flow is
The heat balance integral method is used here to reduce the nu-
qCl ST
merical computation to an elementary form. Solutions can easily be * _ (10)
Q
generated for a range of parameters, when freezing or thawing oc- 2-Kk^e £nB
curs.
Solution and Application to Soil Systems
Approximate Solution Equations (6-8) were solved with a fourth order Runge-Kutta,
The derivation of the heat balance integral method and applications predictor-corrector technique. Since the problem is initially singular
have been described previously, Goodman [7, 8]. The equations for at the origin, the Neumann solution was used to start the solution.
a melting system are, Sparrow, et al. [5] solved this problem numerically, with ai2 = ki2 =
1, for a range of ST and (j>. The results of the much simpler method
dTA_ r dTj presented here are within 5 percent of these values. The completely
(1)
dr\ dr j oti dt different numerical solution of Tien and Churchill [4] is also in close
(la) agreement as can be seen in Fig. 2. The comparison of the heat flow
TiCB, t) = T,
from the cylinder can be seen from Fig. 3. The calculations have been
Ti(r 0 , t) = Ts (16) generalized for a range of ct\2, k\2t appropriate for soil systems, and
are presented in Lunardini [13]. The property ratios for soils can be
(2) expressed as a function of X£, the volumetric fraction of water in the
dr \ dr I a2 dt soil. These property ratios can be obtained using the method of Lu-
T2(R, t) = T, (2a) nardini and Varotta [3] and references [10-12]. F o r ^ = 0, aw = ki2
= 1.0. The curves are valid for any medium with the same property
T2(8, t) = To (26) ratios. .
dT2(8, t) For a soil the value of %e = 0 denotes zero latent heat and no phase
0 (2c)
dr change. However, in general, the X£ = Q curves could characterize any
The depth of penetration of the temperature wave is 8(t) and beyond material with constant thermal properties and arbitrary latent heat.
this radius the temperature is To. The energy balance at the phase At the other extreme, when %i = 1, the case of pure water occurs and
change interface is the effects of convection should be considered. For a soil, even if xe
is quite high, the nature of the soil solid skeleton minimizes free
dT^R, t) , , dT2(R,t) „dR convection distortions.
-k\ r R2- Pl£- (3)
dr dr ' ' dt
The solution method consists of choosing approximations for the
Discussion
temperatures which will satisfy all of the conditions in terms of the The heat balance integral method reduces the partial differential
unknown parameters 8(t) and i?(£)-LardnerandPohle [9] noted that equations of heat flow to ordinary differential equations. The nu-
a logarithmic temperature approximation is most appropriate when merical solution is then greatly simplified and can be carried out with
the area is varying with r. An approximation for T\ that satisfies elementary computer systems and algorithms. The accuracy is related
equations (la,b) is to the choice of the temperature approximations. For the cylindrical
system it has been shown that the results are essentially identical to
two different, complete, numerical solutions. The heat balance inte-
(T s - T,)£n
fol gral method is practical for two phase, Stefan problems.
Ti = Ts — (4) The effect of the property ratios is quite significant. For the same
eu- values of <f>, S T , as a2\, k2\ increase, the thaw rate increases and the
ro freeze rate decreases. For the same values of %t, 4>, ST, the freeze rate
A similar approximation for T2 that satisfies equations (2a-c) is will exceed the thaw rate at any time, excepting the case for xe = 0.

Journal of Heat Transfer AUGUST 1981, VOL. 103 / 599

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 Heat Conduction through "Yin-
Yang" Bodies
John H. Lienhard 1
Consider steady two-dimensionai heat conduction through bod-
0 5- ies whose boundaries are either maintained a t one of two tempera-
tures, or are insulated. If such a body has a geometrical axis of sym-
metry; if one piece of the left side is insulated and the other piece of
it is held at a temperature, 7 \ ; if the right side is either insulated or
isothermal at T2, with the positions of the insulated and isothermal
segments inverted from those on the left; then the shape factor, S, is
unity.
To see why this is true, consider two such cases in Figs. 1(a) and
1(6). The symmetry specified above obviously requires that the iso-
thermal lines are mirror images of the adiabatic lines. The number
of isothermal increments will therefore match the number of adiabatic
T
increments so S = 1. We shall refer to such shapes as Yin-Yang2 fig-
Fig. 2 Accuracy of phase change radius, S T = 1.0, a-n = * r « = 1.0
ures. Actually any Yin-Yang figure must be the result of conformally
T n T T "WT1"! I'l'l' I ' I I'l'l' I ' 1 I'l'l' I ' I mapping Fig. 1(a). Accordingly it is sufficient that it be simply-con-
Xt=0, * = 4
nected with only one piece of its boundary at each temperature. It can
(•) Heat Balance Integral also be doubly-connected (since Fig. 1(a) can be mapped into an in-
— Sparrow (5) finite region with a single hole in it) as long as the point at infinity is
qe/qp-Cylindrical to plane heat flow ratio
not singular. Two other restrictions are that the thermal conductivity,
k, must be constant, and there should be no heat sources.
Nine more Yin-Yang figures are presented in Fig. 2: (a) is a semi-
infinite strip; (b), (c), and (g) have double axes of Yin-Yang symmetry
although (b) and (g) are infinite in extent and (c) is finite; (d), (/), and
(£) are semi-infinite bodies, (e) and (h) are finite; however, (i) can be
viewed as the inverted limit of case (h) as the piece of boundary at T"i
is shrunk. Each of these situations would be hard to deal with ana-
lytically or numerically. Since S = 1 in each case we can express the
heat transfer, Q (W/m), as simply

ibid. Q = k(Ti - T2)

T The number of possibilities is, of course, unlimited, and they include

Fig. 3 Accuracy of cylinder heat flow, a 1 2 = /t 1 2 = 1.0, (j> = 4.0 many potentially useful configurations. Anyone doing a two-dimen-
sional heat conduction calculation should first look at it to see if it
References resolves into a Yin-Yang figure. If it does, Q can be obtained without
1 Neumann, P., Lectures given in the 1860's C. F. Riemann-Weber, Die calculation.
partiellen Differential-gleichungen Physik, 5th ed., Vol. 2, p. 121.
2 Carslaw, H. S., and Jaeger, J. C, Conduction of Heat in Solids, 2nd ed.,
Clarendon Press, Oxford, 1959.
3 Lunardini, V., and Varotta, R., "Approximate Solution to Neumann
Problem for Soil Systems,"ASME Journal of Energy Resources Technology,
Vol. 103, No. 1,1981, pp. 76-81. 1
Professor, Mechanical Engineering Department, University of Houston,
4 Tien, L. C, and Churchill, S. W., "Freezing Front Motion and Heat Houston, Tex. 77004, Fellow ASME.
Transfer Outside an Infinite, Isothermal Cylinder," AIChE Journal, Vol. 11, 2
No. 5,1965, pp. 790-793. We use the Chinese term to evoke the sort of inverted symmetry involved,
although the actual Yin-Yang figure does not really have the required mirror-
5 Sparrow, E. M., Ramadhyani, S., and Patankar, S. V., "Effect of Sub- image geometrical symmetry.
cooling on Cylindrical Melting," ASME JOURNAL OF HEAT TRANSFER, Vol.
Contributed by the Heat Transfer Division for publication in the JOURNAL
100, No. 3,1978, pp. 395-402. OF HEAT TRANSFER. Manuscript received by the Heat Transfer Division
6 Bell, G. E., "Solidification of a Liquid About a Cylindrical Pipe," In- November 24,1980.
ternational Journal of Heat Mass Transfer, Vol. 22,1979, pp. 1681-1686.
7 Goodman, T. R., "The Heat-Balance Integral and Its Application to
Problems Involving a Change of Phase," Trans. ASME, Vol. 80, 1958, pp.
335-342.
8 Goodman, T. R., "Application of Integral Methods to Transient Non-
linear Heat Transfer," Advances in Heat Transfer, Vol. 1, T. F. Irvine and J.
P. Hartnett, Eds., Academic Press, New York, 1964, pp. 52-122.
9 Lardner, T. J„ and Pohle, F. V., "Application of the Heat Balance In-
tegral to Problems of Cylindrical Geometry," Trans. ASME, Vol.'83, No. 2,1961,
pp. 310-312.
10 Gold, L. W., and Lachenbruch, A. H., "Thermal Conditions in Perma-
frost—A Review of North American Literature," Proceedings of Second In-
ternational Permafrost Conference, National Academy of Science, 1973, pp.
3-25.
11 Kersten, M. S., "Thermal Properties of Soils," University of Minnesota
Experiment Station, Bulletin No. 28,1949, 227 pages.
12 Lunardini, V. J., "Presentation of Some Thermal Properties of Soil
Systems," Canadian Society of Mechanical Engineers, paper 71-CSME-38,
Engineering Institute of Canada Meeting, 1971, Quebec City. boundary of T,
13 Lunardini, V. J., "Phase Change Around a Circular Pipe," CRREL Rent. boundary at T2
80-27, Dec. 1980. boundary insulated
14 Yuen, W. W., "Application of the Heat Balance Integral to Melting
Problems with Initial Subcooling," International Journal of Heat Mass ndiobotic owl isothermal lines
Transfer, Vol. 23,1980, pp. 1157-1160. Fig. 1 Two Yin-Yang figures

600 / VOL. 103, AUGUST 1981 Transactions of the ASME

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 Heat Conduction through "Yin-
Yang" Bodies
John H. Lienhard 1
Consider steady two-dimensionai heat conduction through bod-
0 5- ies whose boundaries are either maintained a t one of two tempera-
tures, or are insulated. If such a body has a geometrical axis of sym-
metry; if one piece of the left side is insulated and the other piece of
it is held at a temperature, 7 \ ; if the right side is either insulated or
isothermal at T2, with the positions of the insulated and isothermal
segments inverted from those on the left; then the shape factor, S, is
unity.
To see why this is true, consider two such cases in Figs. 1(a) and
1(6). The symmetry specified above obviously requires that the iso-
thermal lines are mirror images of the adiabatic lines. The number
of isothermal increments will therefore match the number of adiabatic
T
increments so S = 1. We shall refer to such shapes as Yin-Yang2 fig-
Fig. 2 Accuracy of phase change radius, S T = 1.0, a-n = * r « = 1.0
ures. Actually any Yin-Yang figure must be the result of conformally
T n T T "WT1"! I'l'l' I ' I I'l'l' I ' 1 I'l'l' I ' I mapping Fig. 1(a). Accordingly it is sufficient that it be simply-con-
Xt=0, * = 4
nected with only one piece of its boundary at each temperature. It can
(•) Heat Balance Integral also be doubly-connected (since Fig. 1(a) can be mapped into an in-
— Sparrow (5) finite region with a single hole in it) as long as the point at infinity is
qe/qp-Cylindrical to plane heat flow ratio
not singular. Two other restrictions are that the thermal conductivity,
k, must be constant, and there should be no heat sources.
Nine more Yin-Yang figures are presented in Fig. 2: (a) is a semi-
infinite strip; (b), (c), and (g) have double axes of Yin-Yang symmetry
although (b) and (g) are infinite in extent and (c) is finite; (d), (/), and
(£) are semi-infinite bodies, (e) and (h) are finite; however, (i) can be
viewed as the inverted limit of case (h) as the piece of boundary at T"i
is shrunk. Each of these situations would be hard to deal with ana-
lytically or numerically. Since S = 1 in each case we can express the
heat transfer, Q (W/m), as simply

ibid. Q = k(Ti - T2)

T The number of possibilities is, of course, unlimited, and they include

Fig. 3 Accuracy of cylinder heat flow, a 1 2 = /t 1 2 = 1.0, (j> = 4.0 many potentially useful configurations. Anyone doing a two-dimen-
sional heat conduction calculation should first look at it to see if it
References resolves into a Yin-Yang figure. If it does, Q can be obtained without
1 Neumann, P., Lectures given in the 1860's C. F. Riemann-Weber, Die calculation.
partiellen Differential-gleichungen Physik, 5th ed., Vol. 2, p. 121.
2 Carslaw, H. S., and Jaeger, J. C, Conduction of Heat in Solids, 2nd ed.,
Clarendon Press, Oxford, 1959.
3 Lunardini, V., and Varotta, R., "Approximate Solution to Neumann
Problem for Soil Systems,"ASME Journal of Energy Resources Technology,
Vol. 103, No. 1,1981, pp. 76-81. 1
Professor, Mechanical Engineering Department, University of Houston,
4 Tien, L. C, and Churchill, S. W., "Freezing Front Motion and Heat Houston, Tex. 77004, Fellow ASME.
Transfer Outside an Infinite, Isothermal Cylinder," AIChE Journal, Vol. 11, 2
No. 5,1965, pp. 790-793. We use the Chinese term to evoke the sort of inverted symmetry involved,
although the actual Yin-Yang figure does not really have the required mirror-
5 Sparrow, E. M., Ramadhyani, S., and Patankar, S. V., "Effect of Sub- image geometrical symmetry.
cooling on Cylindrical Melting," ASME JOURNAL OF HEAT TRANSFER, Vol.
Contributed by the Heat Transfer Division for publication in the JOURNAL
100, No. 3,1978, pp. 395-402. OF HEAT TRANSFER. Manuscript received by the Heat Transfer Division
6 Bell, G. E., "Solidification of a Liquid About a Cylindrical Pipe," In- November 24,1980.
ternational Journal of Heat Mass Transfer, Vol. 22,1979, pp. 1681-1686.
7 Goodman, T. R., "The Heat-Balance Integral and Its Application to
Problems Involving a Change of Phase," Trans. ASME, Vol. 80, 1958, pp.
335-342.
8 Goodman, T. R., "Application of Integral Methods to Transient Non-
linear Heat Transfer," Advances in Heat Transfer, Vol. 1, T. F. Irvine and J.
P. Hartnett, Eds., Academic Press, New York, 1964, pp. 52-122.
9 Lardner, T. J„ and Pohle, F. V., "Application of the Heat Balance In-
tegral to Problems of Cylindrical Geometry," Trans. ASME, Vol.'83, No. 2,1961,
pp. 310-312.
10 Gold, L. W., and Lachenbruch, A. H., "Thermal Conditions in Perma-
frost—A Review of North American Literature," Proceedings of Second In-
ternational Permafrost Conference, National Academy of Science, 1973, pp.
3-25.
11 Kersten, M. S., "Thermal Properties of Soils," University of Minnesota
Experiment Station, Bulletin No. 28,1949, 227 pages.
12 Lunardini, V. J., "Presentation of Some Thermal Properties of Soil
Systems," Canadian Society of Mechanical Engineers, paper 71-CSME-38,
Engineering Institute of Canada Meeting, 1971, Quebec City. boundary of T,
13 Lunardini, V. J., "Phase Change Around a Circular Pipe," CRREL Rent. boundary at T2
80-27, Dec. 1980. boundary insulated
14 Yuen, W. W., "Application of the Heat Balance Integral to Melting
Problems with Initial Subcooling," International Journal of Heat Mass ndiobotic owl isothermal lines
Transfer, Vol. 23,1980, pp. 1157-1160. Fig. 1 Two Yin-Yang figures

600 / VOL. 103, AUGUST 1981 Transactions of the ASME

Copyright © 1981 by ASME
Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 (initially at x = X0) is heated uniformly. We assume that the solid,
initially at temperature T = T 0 , undergoes inert expansion until the
exposed surface reaches a phase change temperature, Tpc. Then a new
solid phase forms and an interface begins to move toward x = 0.
(Another possibility that may be treated just as well is the melting
of a horizontal slab, under uniform heating from above to avoid
buoyant convection currents.) The new solid phase warms up further
until T = Tpyr say, when gasification begins by surface pyrolysis or
sublimation. We neglect transient internal stresses, and assume that
the kinetic energy of the solid is small. The pressure is assumed to be
uniform.
This model should not be applied when the heating rate is too high.
Super-heating could occur, resulting in a variable composition zone
between the two solid phases and subsurface gas flows near the ex-
posed face. With a high radiant influx, subsurface absorption would
be expected. If the expansion speed approaches the speed of sound,
then the kinetic energy of the solid may become important, and in-
ternal pressure variations and transient stresses should be consid-
ered.
In the interior of each phase, the temperature and velocity vary in
accordance with heat conservation
dT dT
• boundary at T( — + v— = (1)
dt dx pC
>Cdx\ dxj
= boundary at Tg
- boundary insulated and mass conservation,
Fig. 2 Nine more Yin-Yang figures
:
— + — (pv) 0 , (2)
Acknowledgment dt dx
I am grateful to the Lloyds Trefethen and to Lewis Wheeler for where C = C(T) is the specific heat at constant pressure, k = k(T) is
useful criticism and discussion of this problem. Lloyd N. Trefethen the thermal conductivity, and p = p(T) is the density. We now define
pointed out the simply-connected restriction. a quantity

£= £ dx'p/p(T0), (3)
Transient Thermal Expansion of
Solids During Inert Heating, reminiscent of the Dorotnitzyn-Stewartson transformation [7]. In
Phase Change, and Surface equation (3), we integrate across the phase interface if necessary, so
Gasification that £ is continuous even when p changes abruptly at the interface.
We make £ a new coordinate and write T =.T(j-, t). It is then found
that equation (1) simplifies to
L. W. Hunter 1
dt p(T 0 )C d{:(**§)• (4)

Introduction where k* = k*(T) is an effective thermal conductivity defined by

A difficulty in calculating heat conduction in an expanding solid
k*(T) = k(T)p(T)/p(T0). (5)
is that the boundary locations are unknown in advance. However, by
recognizing that some energy is carried by the internal motion in the The free boundary difficulty that arises in the usual formulation
solid, we show here that the governing equations (in one dimension) of inert thermal expansion [1] is avoided in ^-coordinates. The value
may be reduced to an ordinary fixed boundary heat conduction of £ at the exposed surface, £ = H(t) say, is simply the initial thickness
problem in a new coordinate. When phase changes and surface gas- of the slab, a(t) = XQ. The value remains XQ during the subsurface
ification occur with thermal expansion, the same approach reduces phase change.
the equations to constant density form. We now list the remaining auxiliary conditions necessary to com-
Transient thermal expansion has a number of important effects. pletely determine the temperature. In these conditions, it may be
A substantial resistance to heat flow can develop between two solids noted for later use that the density p appears explicitly only in the
initially in contact, when one of the solids cools and recedes [1]. When effective thermal conductivity, k*, as in equation (4).
a solid is constrained from expanding, a large stress can result.
Thermal expansion in ramjet insulators [2, 3] and intumescent
coatings [4] significantly reduces their heat throughput. As a last
example, the velocity of a subsurface phase change front can be af-
fected by thermal expansion [5,6].

Governing Equations
To combine a number of possibilities in one example, Pig. 1, we Radiant heating (F)
Convective
consider a solid slab which insulates a metal from heating on one side. cooling (H) " - Reradiation
The back face of the slab (at x = 0) is held fixed while the front face

change Pyrolysis
1
Applied Physics Laboratory, The Johns Hopkins University, Laurel, MD
20810.
Contributed by the Heat Transfer Division for publication in the JOURNAL
OF HEAT TRANSFER. Manuscript received by the Heat Transfer Division, Fig. 1 A swelling insulator which protects a metal from heating on one
October 20,1980. side

Journal of Heat Transfer AUGUST 1981, VOL. 103 / 601

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 (initially at x = X0) is heated uniformly. We assume that the solid,
initially at temperature T = T 0 , undergoes inert expansion until the
exposed surface reaches a phase change temperature, Tpc. Then a new
solid phase forms and an interface begins to move toward x = 0.
(Another possibility that may be treated just as well is the melting
of a horizontal slab, under uniform heating from above to avoid
buoyant convection currents.) The new solid phase warms up further
until T = Tpyr say, when gasification begins by surface pyrolysis or
sublimation. We neglect transient internal stresses, and assume that
the kinetic energy of the solid is small. The pressure is assumed to be
uniform.
This model should not be applied when the heating rate is too high.
Super-heating could occur, resulting in a variable composition zone
between the two solid phases and subsurface gas flows near the ex-
posed face. With a high radiant influx, subsurface absorption would
be expected. If the expansion speed approaches the speed of sound,
then the kinetic energy of the solid may become important, and in-
ternal pressure variations and transient stresses should be consid-
ered.
In the interior of each phase, the temperature and velocity vary in
accordance with heat conservation
dT dT
• boundary at T( — + v— = (1)
dt dx pC
>Cdx\ dxj
= boundary at Tg
- boundary insulated and mass conservation,
Fig. 2 Nine more Yin-Yang figures
:
— + — (pv) 0 , (2)
Acknowledgment dt dx
I am grateful to the Lloyds Trefethen and to Lewis Wheeler for where C = C(T) is the specific heat at constant pressure, k = k(T) is
useful criticism and discussion of this problem. Lloyd N. Trefethen the thermal conductivity, and p = p(T) is the density. We now define
pointed out the simply-connected restriction. a quantity

£= £ dx'p/p(T0), (3)
Transient Thermal Expansion of
Solids During Inert Heating, reminiscent of the Dorotnitzyn-Stewartson transformation [7]. In
Phase Change, and Surface equation (3), we integrate across the phase interface if necessary, so
Gasification that £ is continuous even when p changes abruptly at the interface.
We make £ a new coordinate and write T =.T(j-, t). It is then found
that equation (1) simplifies to
L. W. Hunter 1
dt p(T 0 )C d{:(**§)• (4)

Introduction where k* = k*(T) is an effective thermal conductivity defined by

A difficulty in calculating heat conduction in an expanding solid
k*(T) = k(T)p(T)/p(T0). (5)
is that the boundary locations are unknown in advance. However, by
recognizing that some energy is carried by the internal motion in the The free boundary difficulty that arises in the usual formulation
solid, we show here that the governing equations (in one dimension) of inert thermal expansion [1] is avoided in ^-coordinates. The value
may be reduced to an ordinary fixed boundary heat conduction of £ at the exposed surface, £ = H(t) say, is simply the initial thickness
problem in a new coordinate. When phase changes and surface gas- of the slab, a(t) = XQ. The value remains XQ during the subsurface
ification occur with thermal expansion, the same approach reduces phase change.
the equations to constant density form. We now list the remaining auxiliary conditions necessary to com-
Transient thermal expansion has a number of important effects. pletely determine the temperature. In these conditions, it may be
A substantial resistance to heat flow can develop between two solids noted for later use that the density p appears explicitly only in the
initially in contact, when one of the solids cools and recedes [1]. When effective thermal conductivity, k*, as in equation (4).
a solid is constrained from expanding, a large stress can result.
Thermal expansion in ramjet insulators [2, 3] and intumescent
coatings [4] significantly reduces their heat throughput. As a last
example, the velocity of a subsurface phase change front can be af-
fected by thermal expansion [5,6].

Governing Equations
To combine a number of possibilities in one example, Pig. 1, we Radiant heating (F)
Convective
consider a solid slab which insulates a metal from heating on one side. cooling (H) " - Reradiation
The back face of the slab (at x = 0) is held fixed while the front face

change Pyrolysis
1
Applied Physics Laboratory, The Johns Hopkins University, Laurel, MD
20810.
Contributed by the Heat Transfer Division for publication in the JOURNAL
OF HEAT TRANSFER. Manuscript received by the Heat Transfer Division, Fig. 1 A swelling insulator which protects a metal from heating on one
October 20,1980. side

Journal of Heat Transfer AUGUST 1981, VOL. 103 / 601

Copyright © 1981 by ASME
Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 During the pyrolysis, S(t) satisfies Acknowledgment
dT d"E(t) The author thanks J. C. W. Rogers, L. L. Perini, and R. W. Hart for
F-aTpyr*-k*-~ = -Bpyrp(T0)-~ (6) helpful discussions. Support for this work was provided by the De-
d£ at
partment of the Navy, Naval Sea Systems Command, under Contract
at £ = &(t), where F is the radiant influx, a is the Stefan-Boltzmann N00024-78-C-5384.
constant, and ftpyr is the specific heat of pyrolysis. The location of the
phase change interface £ = (t) say, satisfies
Extension of the Adams-Welty
lk*^\ -fk*^) =-UpeP(T0)7-Epc(t), (7) Fluid Bed Heat Transfer Model
where fipc is the specific heat of phase change.
to the Packed Bed Case
The boundary condition at the metal (£ = 0) is
Ronald L. Adams1
dT dT
CmPml— = k*—- H(T- T0) at f = 0, (8)
dt d£
Introduction
where C m is the (constant) specific heat, pm the (constant) density,
The purpose of this note is to propose a packed bed heat transfer
and I the thickness of the metal. In addition, H is the heat transfer
model based upon the gas convection dominant fluid bed heat transfer
coefficient for convective losses to air at temperature To. The
model of Adams and Welty [1]. The assumption of large particle
boundary condition at the phase change front is T = Tpc. At the ex-
Reynolds number is retained and the resulting packed bed model
posed surface, the condition is
seems to be valid for particle Reynolds number greater than about
dT 50. The new packed bed heat transfer model can be combined with
F- aT^-k* — = 0 if T < Tpyr (9) the original fluid bed heat transfer model such that analytical studies
of heat transfer during the turndown phase of combustor operation
and T = Tpyr otherwise. can be carried out.
The solution of the last equations determines the x-location of the
exposed face,
Model Description
H(t)
X d£p(T0)/p (10)
The fundamental element of the gas convection dominant heat
transfer model of Adams and Welty is analysis of flow within inter-
stitial channels adjacent to the heat transfer surface. The model is
and of the phase change front similarly. The mass loss flux is
based upon a boundary layer analysis of flow within these channels
with separate treatment of the two-dimensional flow near the channel
m» = -p(T0)£-m). (ID center and three-dimensional, but Stokes' flow in the cusped corners
at formed at particle contact points. For the fluidized regime, the ther-
mal state of the average particle is assumed to be such that the tem-
perature distribution along the particle surface (from the wall to the
Discussion edge of the Stokes' region) is linear and this assumption yields rea-
In the thermal expansion problem in ^-coordinates, equations (4-9), sonable agreement with experiment (see Catipovic, et al, [2]). The size
the density appeared only in the effective thermal conductivity, k*. of the Stokes' region is established by requiring the heat transfer be
Now suppose that the density is set constant without changing the continuous across the channel and therefore depends upon the
other thermal properties. Then i- becomes x and k* becomes k and thickness of the two-dimensional boundary layer near the channel
we can denote the temperature by T^ix, t; k), in which the depen- center.
dence on k = h(T) is indicated. It is seen that the variable density When the bed is in the packed state, solid particles will remain at
solution is just the heat transfer surface indefinitely and consideration must be given
to the conditions under which the model of Adams and Welty remains
T = T<°>(£,t;fe*), (12)
valid. In particular, the model assumes that the influence of the heat
in which £ replaces x and k* replaces k. It follows that variable density transfer surface is confined to a single particle layer. This condition
problems for a variety of boundary conditions including those in Fig. is satisfied as long as the major contribution to the heat transfer is
1 may be reduced to equivalent constant density problems and solved convection due to the two-dimensional gas boundary layer forming
by the usual methods. on the tube wall. Thus the Reynolds number must be large enough
such that boundary layer thickness within the interstitial channel is
less than half a particle diameter and so that the Stokes' region does
not encompass the entire width of the channel. Furthermore, the
References thermal state of the average particle will be different than under
1 Meyer, G. H., "A Numerical Method for Heat Transfer in an Expanding fluidized conditions and this difference is approximated by assuming
Rod," International Journal of Heat and Mass Transfer, Vol. 21, 1978, pp. that the portion of the particle surface near the wall and bounding
824-826. the Stokes' region is at the wall temperature. The major difference
2 Roberts, W., Laub, B., Suchsland, K., Shimizu, A., and Chambers, J.,
"Ramburner Internal Insulation Investigation. Modeling and Computer Pro- between this model and those of Gabor [3] and Botterill and Denloye
grams," Marquardt Company Report AFAPL-TR-75-109-Vol.-l, 1975. [4] is its fundamental approach to the details of the heat transfer
3 Chambers, J., Harkins, W., Roberts, W., and Meyers, G., "Ramburner processes rather than the use of empirical, effective conductivities.
Internal Insulation Investigation. Characterization and Tailoring of Silicone Also, heat transfer beyond the particle layer adjacent to the wall has
Elastomers," Marquardt Company Report AFAPL-TR-75-199-Vol.-2,1975.
4 Cagliostro, D. E., Riccitiello, S. R., Clark, K. J., and Shimizu, A. B., been neglected and this confines the model to large particle Reynolds
"Intumescent Coating Modeling," Journal of Fire and Flammability, Vol. 6, number.
1975, pp. 205-221.
5 Danckwerts, P. V., "Unsteady-State Diffusion or Heat-Conduction with
Moving Boundary," Transactions of the Faraday Society, Vol. 46,1950, pp.
1
701-712. Assistant Professor, Department of Mechanical Engineering, Oregon State
6 Carslaw, H. S., and Jaeger, J. C, Conduction of Heat in Solids, 2nd ed. University, Corvallis, Ore. 97331.
Oxford at the Clarendon Press, 1959, pp. 290-291. Contributed by the Heat Transfer Division for publication in the JOURNAL
7 Eckert, E. R. G., and Drake, R. M., Jr., Analysis of Heat and Mass OF HEAT TRANSFER. Manuscript received by the Heat Transfer Division,
Transfer, McGraw-Hill, 1972, p. 424. January 12,1981.

602 / VOL. 103, AUGUST 1981 Transactions of the ASME

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 During the pyrolysis, S(t) satisfies Acknowledgment
dT d"E(t) The author thanks J. C. W. Rogers, L. L. Perini, and R. W. Hart for
F-aTpyr*-k*-~ = -Bpyrp(T0)-~ (6) helpful discussions. Support for this work was provided by the De-
d£ at
partment of the Navy, Naval Sea Systems Command, under Contract
at £ = &(t), where F is the radiant influx, a is the Stefan-Boltzmann N00024-78-C-5384.
constant, and ftpyr is the specific heat of pyrolysis. The location of the
phase change interface £ = (t) say, satisfies
Extension of the Adams-Welty
lk*^\ -fk*^) =-UpeP(T0)7-Epc(t), (7) Fluid Bed Heat Transfer Model
where fipc is the specific heat of phase change.
to the Packed Bed Case
The boundary condition at the metal (£ = 0) is
Ronald L. Adams1
dT dT
CmPml— = k*—- H(T- T0) at f = 0, (8)
dt d£
Introduction
where C m is the (constant) specific heat, pm the (constant) density,
The purpose of this note is to propose a packed bed heat transfer
and I the thickness of the metal. In addition, H is the heat transfer
model based upon the gas convection dominant fluid bed heat transfer
coefficient for convective losses to air at temperature To. The
model of Adams and Welty [1]. The assumption of large particle
boundary condition at the phase change front is T = Tpc. At the ex-
Reynolds number is retained and the resulting packed bed model
posed surface, the condition is
seems to be valid for particle Reynolds number greater than about
dT 50. The new packed bed heat transfer model can be combined with
F- aT^-k* — = 0 if T < Tpyr (9) the original fluid bed heat transfer model such that analytical studies
of heat transfer during the turndown phase of combustor operation
and T = Tpyr otherwise. can be carried out.
The solution of the last equations determines the x-location of the
exposed face,
Model Description
H(t)
X d£p(T0)/p (10)
The fundamental element of the gas convection dominant heat
transfer model of Adams and Welty is analysis of flow within inter-
stitial channels adjacent to the heat transfer surface. The model is
and of the phase change front similarly. The mass loss flux is
based upon a boundary layer analysis of flow within these channels
with separate treatment of the two-dimensional flow near the channel
m» = -p(T0)£-m). (ID center and three-dimensional, but Stokes' flow in the cusped corners
at formed at particle contact points. For the fluidized regime, the ther-
mal state of the average particle is assumed to be such that the tem-
perature distribution along the particle surface (from the wall to the
Discussion edge of the Stokes' region) is linear and this assumption yields rea-
In the thermal expansion problem in ^-coordinates, equations (4-9), sonable agreement with experiment (see Catipovic, et al, [2]). The size
the density appeared only in the effective thermal conductivity, k*. of the Stokes' region is established by requiring the heat transfer be
Now suppose that the density is set constant without changing the continuous across the channel and therefore depends upon the
other thermal properties. Then i- becomes x and k* becomes k and thickness of the two-dimensional boundary layer near the channel
we can denote the temperature by T^ix, t; k), in which the depen- center.
dence on k = h(T) is indicated. It is seen that the variable density When the bed is in the packed state, solid particles will remain at
solution is just the heat transfer surface indefinitely and consideration must be given
to the conditions under which the model of Adams and Welty remains
T = T<°>(£,t;fe*), (12)
valid. In particular, the model assumes that the influence of the heat
in which £ replaces x and k* replaces k. It follows that variable density transfer surface is confined to a single particle layer. This condition
problems for a variety of boundary conditions including those in Fig. is satisfied as long as the major contribution to the heat transfer is
1 may be reduced to equivalent constant density problems and solved convection due to the two-dimensional gas boundary layer forming
by the usual methods. on the tube wall. Thus the Reynolds number must be large enough
such that boundary layer thickness within the interstitial channel is
less than half a particle diameter and so that the Stokes' region does
not encompass the entire width of the channel. Furthermore, the
References thermal state of the average particle will be different than under
1 Meyer, G. H., "A Numerical Method for Heat Transfer in an Expanding fluidized conditions and this difference is approximated by assuming
Rod," International Journal of Heat and Mass Transfer, Vol. 21, 1978, pp. that the portion of the particle surface near the wall and bounding
824-826. the Stokes' region is at the wall temperature. The major difference
2 Roberts, W., Laub, B., Suchsland, K., Shimizu, A., and Chambers, J.,
"Ramburner Internal Insulation Investigation. Modeling and Computer Pro- between this model and those of Gabor [3] and Botterill and Denloye
grams," Marquardt Company Report AFAPL-TR-75-109-Vol.-l, 1975. [4] is its fundamental approach to the details of the heat transfer
3 Chambers, J., Harkins, W., Roberts, W., and Meyers, G., "Ramburner processes rather than the use of empirical, effective conductivities.
Internal Insulation Investigation. Characterization and Tailoring of Silicone Also, heat transfer beyond the particle layer adjacent to the wall has
Elastomers," Marquardt Company Report AFAPL-TR-75-199-Vol.-2,1975.
4 Cagliostro, D. E., Riccitiello, S. R., Clark, K. J., and Shimizu, A. B., been neglected and this confines the model to large particle Reynolds
"Intumescent Coating Modeling," Journal of Fire and Flammability, Vol. 6, number.
1975, pp. 205-221.
5 Danckwerts, P. V., "Unsteady-State Diffusion or Heat-Conduction with
Moving Boundary," Transactions of the Faraday Society, Vol. 46,1950, pp.
1
701-712. Assistant Professor, Department of Mechanical Engineering, Oregon State
6 Carslaw, H. S., and Jaeger, J. C, Conduction of Heat in Solids, 2nd ed. University, Corvallis, Ore. 97331.
Oxford at the Clarendon Press, 1959, pp. 290-291. Contributed by the Heat Transfer Division for publication in the JOURNAL
7 Eckert, E. R. G., and Drake, R. M., Jr., Analysis of Heat and Mass OF HEAT TRANSFER. Manuscript received by the Heat Transfer Division,
Transfer, McGraw-Hill, 1972, p. 424. January 12,1981.

602 / VOL. 103, AUGUST 1981 Transactions of the ASME

Copyright © 1981 by ASME
Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 The alteration of the boundary conditions on the Stokes' region Comparison with Experiment
described above will modify both the Stokes' region size and average Heat transfer data for a horizontal tube immersed in an atmo-
Nusselt number. Following the analysis of Adams [5], the average spheric pressure packed bed reported by Catipovic [8] were selected
Nusselt number for the Stokes' region when the particle surface is at to test the validity of the model. Other data for horizontal tubes im-
the wall temperature is mersed in large particle/high pressure beds have been reported by
2S 0
Chandron, et al. [9] but packed bed data are not included. Also,
Nu D (1) Botterill and Denloye [4] report data for a vertical cylinder immersed
(sP/rp) in a large particle bed under both packed and fluidized conditions.
with Local and total Nusselt number calculations were made for particle
sizes in the range 1.3-6.6 mm in a low temperature bed at atmospheric
pressure. The parameters used in the calculations are listed in Table
2\ TB)\TB) XTJTB + AJTB)
1 and comparison of calculated results with Catipovic's data is made
where TB and Tw are bed and wall temperatures, respectively, A\ is in Figs. 1 and 2. Figure 1 illustrates the accuracy of calculated local
the Sutherland constant for the gas, Sp is half the particle spacing at values for 4 mm dia particles. It appears from the figure that local
the wall and r p is the particle radius. While the Stokes' region fluidization and bubbling may be occurring near the side of the tube
matching criterion is modified such that the parameters k\ and hi
appearing in the approximations for the Stokes' region boundary are
(see Adams [6]) Table 1 Calculation parameters
Bed Temperature 308 K
(Wpacked = (fel)fluidized/l-4 (3)
Wall Temperature 376 K
(4)
Bed Voidage 0.40
( ^ p a c k e d = [(Wfluidized] X 1.4 Surface Voidage
Lower Stagnation 0.8 (d = -0.452 to 0.452 rad)
Therefore, Remaining Surface 0.44
Particle Sphericity 0.97
ki = 5.26 - 7.72 (5) Turbulence Intensity 0.20
Prandtl Number 0.70

•• 0.0127 + 0.0222
H
Then, the channel Nusselt number (based upon particle diameter)
(6) 40
OD A CATIPOVIC'S DATA

follows from the previous analysis, i.e., THEORY

/Hep + b + Nu p o Re p > Re P m (7a) 30
cRe p + Nu p o (76)

with
a.
a = 0.76a (8) 5 20
b = —kiSo (9)
= k2a2
S0 (10)
10
1.58 1/2
' Pr o.4 ,(s /r )(1.405 + 5.6e-°- 0 8 4 9 "VRe QJ) (ID
p p p

where Pr is the gas Prandtl number, Qg is the interstitial gas velocity

(nondimensionalized with superficial velocity), u' is the interstitial _L _L
turbulence intensity and Re p is the Reynolds number based upon 0 200 400 600
particle diameter and supericial velocity. The maximum Reynolds Re D
number, Re P m , for quasi-linear behavior satisfies the nonlinear
equation Fig. 1 Comparison of Catipovic's packed bed data for local Nusselt number
with theory for 4 mm dia particles fluidized with air at 1 ATM
c
Re Pm Q+ = A + B < r V R e p „ , Q ; (12)
where bU " I I I I I
. HORIZONTAL TUBE dP =
A = 1.405/3 (13) - 6.6mm
40
B = 5.6/3 (14)
y— -°
C = 0.0849u' (15)
30 r s

with 4mm, / -
s\ s '
, _ /' 1.79PrO' 044
1.79Pr i feiSoVl/2a p /r p |2 jfr* '
(16) 20 ygi THEORY
V>.76 - (0. 578 - 4fe1fe2)1« CATIPOVIC'S DATA
X' o — o d P = 6.6mm
Equation (7) together with a potential model of the interstitial flow X' D—D 4 mm
IO
with corrections for local voidage variation (to obtain Qg) allows es- A---A 2 mm
timation of local and total heat transfer to immersed surfaces. In the O—O 1.3mm
case of a horizontal tube, the lower stagnation region exhibits a rela- y i I I I l
tively high voidage such that the potential model is inaccurate; 200 400 600 800 I000 1200
however, the heating there can be obtained by using the finite dif- Re P
ference solutions for interstitial velocity obtained by George, et al. Fig. 2 Comparison of Catipovic's packed bed data for total Nusselt number
[7] and neglecting the Stokes' region contribution there. with theory for particles fluidized with air at 1 ATM

Journal of Heat Transfer AUGUST 1981, VOL. 103 / 603

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 (8 = 7I-/2), resulting in poor agreement at the higher Reynolds number. support was also provided by the Battelle Memorial Institute Dis-
The top and lower stagnation values agree reasonably well with the tributions to Universities Program.
measurements though it is noted that the analytical value for the lower
stagnation point is sensitive to the voidage assumed. Figure 2 is a References
comparison of calculated and measured total Nusselt number for the 1 Adams, R. L., and Welty, J. R.," A Gas Convection Model of Heat Transfer
four particle sizes considered. Note that the theoretical curves are in Large Particle Fluidized Beds," AIChE Journal, Vol. 25, No. 3, May 1979,
nearly coincident and linear with Reynolds number. The best agree- pp. 395-405.
ment was obtained in the mid-range of Reynolds Number and worst 2 Catipovic, N. M., Fitzgerald, T. J., George, A. H., and Welty, J. R., "Ex-
perimental Validation of the Adams-Welty Model for Heat Transfer in
agreement when local bubbling and fluidization effects seem to be Large-Particle Fluidized Beds," submitted to AIChE Journal, 1980.
present. Over the range of Reynolds number from 50 to 1000, the 3 Gabor, J. D., "Heat Transfer to Particle Beds with Gas Flows Less Than
maximum difference between theory and experiment is about 20 or Equal to That Required for Incipient Fluidization," Chemical Engineering
percent. Science, Vol. 25,1970, p. 979.
4 Botterill, J. S. M.and Denloye, A. O. 0., "Gas Convective Heat Transfer
to Packed and Fluidized Beds," AIChE Symposium Series, Vol. 74, No. 176,
Summary 1978.
The gas convection dominant heat transfer model of Adams and 5 Adams, R. L., "An Analytical Model of Heat Transfer to a Horizontal
Welty has been extended to the case of a horizontal tube immersed Cylinder Immersed in a Gas Fluidized Bed," Ph.D. Thesis, Oregon State Uni-
versity, Corvallis, Ore., 1977.
in a large particle packed bed. This extension will allow analytical 6 Adams, R. L., "An Approximate Formula for Gas Convection Dominant
studies of heat transfer during turndown of a fluid bed combustor Heat Transfer in Large-Particle Fluidized Beds," ASME JOURNAL OF HEAT
operating in the gas convection dominant regime. Comparison of the TRANSFER, Vol. 103, No. 2,1981, pp. 395-397.
model with measurements made by Catipovic [8] shows reasonable 7 George, A. H., Catipovic, N. M., and Welty, J. R., "An Analytical Study
of Heat Transfer to a Horizontal Cylinder in a Large Particle Fluidized Bed,"
agreement with both local and total Nusselt number for an immersed ASME Paper No. 79-HT-78, Aug. 1979.
tube for the particle Reynolds number range of 50 to 1000. 8 Catipovic, N. M., "Heat Transfer to Horizontal Tubes in Fluidized Beds:
Experiment and Theory," Ph.D. Thesis, Oregon State University, Corvallis,
Acknowledgments Ore., 1979.
9 Chandron, R., Chen, J. C, and Staub, F. W., "Local Heat Transfer
The work reported herein was prepared, in part, for the U. S. De- Coefficients Around Horizontal Tubes in Fluidized Beds," ASME JOURNAL
partment of Energy under Contract No. EF-77-S-01-2714. Some OF HEAT TRANSFER, Vol. 102, No. 1,1980. pp. 152-157.

fluid. (The conductances and capacity rates for the two fluids and the
Thermal Storage Regenerator flow periods, ta and tb, are in general different.)
Parameters 2 No axial transfer of heat occurs in the matrix.
for Almost Constant Gas 3 The thermal conductivity of the solid is infinite in the direction
Discharge Temperature normal to fluid flow or, equivalently, the temperature of the matrix
solid is uniform in any plane normal to fluid flow.
4 The ratio of the thermal capacity of the fluid contained in the
F. E. Romie 1 matrix (at any instant) to the thermal capacity of the matrix is neg-
ligibly small (zero).
The fourth idealization will be satisfied for thermal storage re-
Introduction generators when gases are used as the heat transfer agents [9]. The
Regenerator parameters can be selected to satisfy two important third idealization can be relaxed by use of approximate treatments
criteria for thermal energy storage; first, a large fraction of the energy of transverse conduction [10,11]. In many cases, however, regenerators
made available during the charge process should be returned during will be designed to have a very large heat transfer area per unit volume
regenerator discharge and, second, the temperature of the discharge and the resultant fine subdivision of the matrix composing the re-
gas leaving the regenerator should be close to the constant entrance generator implies that this idealization will be closely realized. The
temperature of the gas used to charge the regenerator. The latter absence of axial transfer of heat required by the second idealization
criterion has particular importance if the stored energy is to be used appears to be satisfied for axial conduction in pebble bed and similar
as the heat source for a heat engine because the loss in thermodynamic constructions which offer no continuous path for conduction. How-
availability due to a decreased discharge gas temperature reduces the ever, with operation at sufficiently high temperature the axial transfer
work that can be realized from the stored thermal energy. of heat by radiation can become significant with a consequent re-
The purpose of this note is to present regenerator parameters which duction of the thermal effectiveness and the gas discharge tempera-
satisfy both criteria. In particular, parameters are presented which ture. For constructions with continuous axial conduction paths [12]
give a normalized gas discharge temperature equal to 0.95 at the end can be consulted.
of the discharge period. (T a (l,l) = 0.95). Both the thermal effectiveness, f„, and the normalized final gas
discharge temperature, r a ( l , l ) , ( T „ ( 1 , 1 ) = ( T „ ' ( 1 , 1 ) - T 0 I ' ) / ( T 6 I ' -
Method and Results T„I')) are expressible in terms of four parameters: (l), the ratio of the
The data presented are obtained by iteration (T„(1,1) is a dependent thermal capacity of the mass of gas passed through the regenerator
variable) of the solution method described in [1]. The solution is based during discharge to the thermal capacity of the mass of gas passed
on idealizations common to most treatments [2-8] of the steady-state during charge, (wet) a/(wet) b, (2), the ratio of the thermal capacity
periodic operation of the counterflow regenerator. The idealizations of the mass of gas passed during discharge to the thermal capacity of
are: the matrix solid, (wct)a/WC, (3), the ratio Na/Nb which is equal to
1 The entrance temperatures, Tai and Tbi, of the discharge and the ratio of the Stanton numbers for the two flows and, (4), the value
charge gases are constant. (Actual temperatures are denoted by a of Ntu, the number of transfer units for the regenerator.
prime.) The conductance, hA, for transfer of heat between the gas and
f 1 , (wct)a l - i M,
solid is constant and uniform throughout the matrix as are the thermal Ntu= — + (l)
capacity, WC, of the matrix solid and the capacity rate, roc, of the Na (wct)b Nb
The parameter N with suscripts a or 6 is the ratio of the thermal
1
Assoc. Mem. ASME. conductance, hA, to the capacity rate, wc, during the discharge and
Contributed by the Heat Transfer Division for publication in the JOURNAL charge periods respectively.
OF HEAT TRANSFER. Manuscript received by the Heat Transfer Division Three regenerator properties are expressed in terms of the thermal
November 14,1980.
effectiveness. The energy stored during charge and delivered during

604 / VOL. 103, AUGUST 1981 Transactions o! the ASME

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 (8 = 7I-/2), resulting in poor agreement at the higher Reynolds number. support was also provided by the Battelle Memorial Institute Dis-
The top and lower stagnation values agree reasonably well with the tributions to Universities Program.
measurements though it is noted that the analytical value for the lower
stagnation point is sensitive to the voidage assumed. Figure 2 is a References
comparison of calculated and measured total Nusselt number for the 1 Adams, R. L., and Welty, J. R.," A Gas Convection Model of Heat Transfer
four particle sizes considered. Note that the theoretical curves are in Large Particle Fluidized Beds," AIChE Journal, Vol. 25, No. 3, May 1979,
nearly coincident and linear with Reynolds number. The best agree- pp. 395-405.
ment was obtained in the mid-range of Reynolds Number and worst 2 Catipovic, N. M., Fitzgerald, T. J., George, A. H., and Welty, J. R., "Ex-
perimental Validation of the Adams-Welty Model for Heat Transfer in
agreement when local bubbling and fluidization effects seem to be Large-Particle Fluidized Beds," submitted to AIChE Journal, 1980.
present. Over the range of Reynolds number from 50 to 1000, the 3 Gabor, J. D., "Heat Transfer to Particle Beds with Gas Flows Less Than
maximum difference between theory and experiment is about 20 or Equal to That Required for Incipient Fluidization," Chemical Engineering
percent. Science, Vol. 25,1970, p. 979.
4 Botterill, J. S. M.and Denloye, A. O. 0., "Gas Convective Heat Transfer
to Packed and Fluidized Beds," AIChE Symposium Series, Vol. 74, No. 176,
Summary 1978.
The gas convection dominant heat transfer model of Adams and 5 Adams, R. L., "An Analytical Model of Heat Transfer to a Horizontal
Welty has been extended to the case of a horizontal tube immersed Cylinder Immersed in a Gas Fluidized Bed," Ph.D. Thesis, Oregon State Uni-
versity, Corvallis, Ore., 1977.
in a large particle packed bed. This extension will allow analytical 6 Adams, R. L., "An Approximate Formula for Gas Convection Dominant
studies of heat transfer during turndown of a fluid bed combustor Heat Transfer in Large-Particle Fluidized Beds," ASME JOURNAL OF HEAT
operating in the gas convection dominant regime. Comparison of the TRANSFER, Vol. 103, No. 2,1981, pp. 395-397.
model with measurements made by Catipovic [8] shows reasonable 7 George, A. H., Catipovic, N. M., and Welty, J. R., "An Analytical Study
of Heat Transfer to a Horizontal Cylinder in a Large Particle Fluidized Bed,"
agreement with both local and total Nusselt number for an immersed ASME Paper No. 79-HT-78, Aug. 1979.
tube for the particle Reynolds number range of 50 to 1000. 8 Catipovic, N. M., "Heat Transfer to Horizontal Tubes in Fluidized Beds:
Experiment and Theory," Ph.D. Thesis, Oregon State University, Corvallis,
Acknowledgments Ore., 1979.
9 Chandron, R., Chen, J. C, and Staub, F. W., "Local Heat Transfer
The work reported herein was prepared, in part, for the U. S. De- Coefficients Around Horizontal Tubes in Fluidized Beds," ASME JOURNAL
partment of Energy under Contract No. EF-77-S-01-2714. Some OF HEAT TRANSFER, Vol. 102, No. 1,1980. pp. 152-157.

fluid. (The conductances and capacity rates for the two fluids and the
Thermal Storage Regenerator flow periods, ta and tb, are in general different.)
Parameters 2 No axial transfer of heat occurs in the matrix.
for Almost Constant Gas 3 The thermal conductivity of the solid is infinite in the direction
Discharge Temperature normal to fluid flow or, equivalently, the temperature of the matrix
solid is uniform in any plane normal to fluid flow.
4 The ratio of the thermal capacity of the fluid contained in the
F. E. Romie 1 matrix (at any instant) to the thermal capacity of the matrix is neg-
ligibly small (zero).
The fourth idealization will be satisfied for thermal storage re-
Introduction generators when gases are used as the heat transfer agents [9]. The
Regenerator parameters can be selected to satisfy two important third idealization can be relaxed by use of approximate treatments
criteria for thermal energy storage; first, a large fraction of the energy of transverse conduction [10,11]. In many cases, however, regenerators
made available during the charge process should be returned during will be designed to have a very large heat transfer area per unit volume
regenerator discharge and, second, the temperature of the discharge and the resultant fine subdivision of the matrix composing the re-
gas leaving the regenerator should be close to the constant entrance generator implies that this idealization will be closely realized. The
temperature of the gas used to charge the regenerator. The latter absence of axial transfer of heat required by the second idealization
criterion has particular importance if the stored energy is to be used appears to be satisfied for axial conduction in pebble bed and similar
as the heat source for a heat engine because the loss in thermodynamic constructions which offer no continuous path for conduction. How-
availability due to a decreased discharge gas temperature reduces the ever, with operation at sufficiently high temperature the axial transfer
work that can be realized from the stored thermal energy. of heat by radiation can become significant with a consequent re-
The purpose of this note is to present regenerator parameters which duction of the thermal effectiveness and the gas discharge tempera-
satisfy both criteria. In particular, parameters are presented which ture. For constructions with continuous axial conduction paths [12]
give a normalized gas discharge temperature equal to 0.95 at the end can be consulted.
of the discharge period. (T a (l,l) = 0.95). Both the thermal effectiveness, f„, and the normalized final gas
discharge temperature, r a ( l , l ) , ( T „ ( 1 , 1 ) = ( T „ ' ( 1 , 1 ) - T 0 I ' ) / ( T 6 I ' -
Method and Results T„I')) are expressible in terms of four parameters: (l), the ratio of the
The data presented are obtained by iteration (T„(1,1) is a dependent thermal capacity of the mass of gas passed through the regenerator
variable) of the solution method described in [1]. The solution is based during discharge to the thermal capacity of the mass of gas passed
on idealizations common to most treatments [2-8] of the steady-state during charge, (wet) a/(wet) b, (2), the ratio of the thermal capacity
periodic operation of the counterflow regenerator. The idealizations of the mass of gas passed during discharge to the thermal capacity of
are: the matrix solid, (wct)a/WC, (3), the ratio Na/Nb which is equal to
1 The entrance temperatures, Tai and Tbi, of the discharge and the ratio of the Stanton numbers for the two flows and, (4), the value
charge gases are constant. (Actual temperatures are denoted by a of Ntu, the number of transfer units for the regenerator.
prime.) The conductance, hA, for transfer of heat between the gas and
f 1 , (wct)a l - i M,
solid is constant and uniform throughout the matrix as are the thermal Ntu= — + (l)
capacity, WC, of the matrix solid and the capacity rate, roc, of the Na (wct)b Nb
The parameter N with suscripts a or 6 is the ratio of the thermal
1
Assoc. Mem. ASME. conductance, hA, to the capacity rate, wc, during the discharge and
Contributed by the Heat Transfer Division for publication in the JOURNAL charge periods respectively.
OF HEAT TRANSFER. Manuscript received by the Heat Transfer Division Three regenerator properties are expressed in terms of the thermal
November 14,1980.
effectiveness. The energy stored during charge and delivered during

604 / VOL. 103, AUGUST 1981 Transactions o! the ASME

Copyright © 1981 by ASME
Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 T a b l e 1 S i n g l e p a r a m e t e r v a r i a t i o n of r a a n d r a (1,1)
a b o u t t h e p o i n t N t u = 128, ( w e t ) J (wet) b = 0.95,
(wct)JWC = 0.7 a n d Na/Nb = 1.0
(Point)
Ntu 64 90 128 180 256
T„ 0.984 0.991 0.996 0.998 1.000
Ta(l,D 0.873 0.914 0.950 0.976 0.992
0.95
Na/Nb 0.5 0.7 1.0 1.4 2.0 (wct)q
Ta 0.996 0.996 0.996 0.996 0.996
Ta(l,D 0.948 0.949 0.950 0.951 0.952 (wct)D

(wct)J(wct)b 0.91 0.93 0.95 0.97 0.99 0.90 -

Ta 0.999 0.998 0.996 0.992 0.986
Ta(l,D 0.985 0.972 0.950 0.912 0.855
(lVCt)a/WC 0.5 0.6 0.7 0.8 0.9
Ta 0.999 0.998 0.996 0.992 0.983 0.8 5 -
Ta(l,D 0.998 0.976 0.950 0.894 0.754

discharge is Q.

= (wCt)aha{Tbl -Tal) (2)

0.80
200 400
The fraction of the energy made available during charge, (wct)b(Tbi Ntu
— Tai), that is stored and thus returned on discharge is the storage Fig. 1 Conditions for a normalized gas temperature equal to 0.95 at the end
effectiveness, R. of the discharge period

(wct)g
R (3)
(wCt)b(Tbl' -Tal) (wct)b I.O 1 1 1—-"=-=: 4 _ _ : > ^
The fraction of the matrix maximum energy storage capability, fa
WC(rbi' - Tai), that is used is the capacitance utilization, U.
{wct)a_ - Ntu 23 -) 2 5 6 -|
-
U-- (4)
WC(rbl' - Tal') WC (wct) a /(wct) b 0.802 0.961

For a given thermal storage, Q, the required capacity, WC, of the (wct) a /WC 0.50 0.80
matrix is inversely proportional to the capacitance utilization which _ Na/Nb I.O 1.0 i
cannot exceed unity. /
Figure 1 presents, for Na/Nb = 1, values of Ntu, (wct)a/(wct)b and /
/
(wct)a/WC which correspond to a final normalized gas discharge / / 1
temperature of 0.95. For all points on the curves of Fig. 1 the value / 1
' T 1
of the thermal effectiveness, f „, is greater than 0.99. Hence, with an 7
• ' b //
inaccuracy of less than 1 percent, the ordinate, (wet) „ /'(wet) b can be • /
• /
read as the storage effectiveness, R, and the curve parameter, • /
(wct)a/WC, as the capacitance utilization, U. The data are insensitive / /
to the value used for Na/Nb- Calculations using Na/Nb = 0.5 and 2.0 y / //
produce only slightly different results. The difference cannot be shown s/ //
s /
to the scale of Fig. 1.
i ~-
** i i _Z\
/
Table 1 shows the effect on f0 and T 0 (1,1) of changing, one at a time,
the values of the four parameters all of which, except the one being t/ta, t/tb
varied, remain at the point defined in the table heading. Inspection Fig. 2 Temperature histories of the leaving gases during discharge (T„) and
of the table shows, for example, that varying Na/Nb from 0.5 to 2.0 charge ( T b )
has no effect (i.e., less than 0.1 percent) on the thermal effectiveness
and produces only a 0.4 percent increase in the value of Ta (1,1). The
table also shows, for given values of (wct)a/(wct)b and (wct)a/WC,
that a value of Ntu greater than that given by Fig. 1 produces a value
of T „ ( 1 , 1 ) greater than 0.95, and conversely.
Figure 2 shows the temperature histories of the charge, T;,, and
discharge, T „ , gases leaving the regenerator for two cases corre-
sponding to Ta (1,1) = 0.95. It may be noted that the discharge gas
temperature decreases from 1.0 to 0.95 only towards the end of the
discharge period. In Fig. 2 the area below a Ta curve divided by the
total area of the figure is the thermal effectiveness and the area above
a T;, curve divided by the total area of the figure is the storage effec-
tiveness, R. Figure 3 shows the temperature distributions in the matrix
at the beginning of the discharge (Ta) and charge (71/,) periods for the.
two cases defined in Fig. 2. The ratio of the area included between a
pair of curves (both dashed or solid) to the total area of the figure is
the capacitance utilization, U.

References
1 Romie, F. E., "Periodic Thermal Storage: The Regenerator," ASME
JOURNAL OP HEAT TRANSFER, Vol. 101, Nov. 1979, pp. 726-731.
2 Nusselt, W., "Der Beharrungszustand im Winderhitzer," Zeitschrift
des Vereins Deutscher Ingenieiw, Vol. 72, July 1928, pp. 1052-1054.
3 Hausen, H., "Nahrungsverfahren zur Berechnung des Warmeaus- Fig. 3 Matrix temperature distributions at the beginning of discharge ( r „ )
tausches in Regeneratoren," ZAMM, Vol. 11,1931, pp. 105-114. and the beginning of charge ( T 0 ) . See Fig. 2 for key.

Journal of Heat Transfer AUGUST 1981, VOL. 103 / 605

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 4 Iliffe, M. A., "Thermal Analysis of the Contra-Flow Regenerative Heat transfer rate per unit base area of the drop. The characteristic drop
Exchanger," Proceedings of the Institute of Mechanical Engineers, Vol. 159, size in the actual dimension r i = 2X;//i;, which corresponds to fl* =
1948, pp. 363-372.
5 Lambertson, T. J., "Performance Factors of a Periodic-Flow Heat Ex- 2, is listed in the fourth column of Table 1.
changer," Trans. ASME, Vol. 80, Apr. 1958, pp. 586-592. The so-called steady dropwise condensation on an inclined surface
6 Nahavandi, A. N., and Weinstein, A. S., "A Solution to the Periodic-Flow is in reality composed of transient dropwise condensation which is
Regenerative Heat Exchanger Problem," Applied Scientific Research, Sec. taking place repeatedly on tracks left by departing drops. The in-
A, Vol. 10,1961, pp. 335-348.
7 Hausen, H., Warmeiibertragung im Gegenstrom, Gleichstrom und stantaneous drop-size distribution during this transient dropwise
Kreuzstrom, Springer-Verlag, Berlin, 1950, pp. 262-444. condensation occurring locally on a newly exposed surface follows a
8 Jakob, M., Heat Transfer, Vol. 2, John Wiley & Sons, New York, 1957, form [2]:
pp. 261-341.
9 Maclaine-cross, I. L., "Effect of Interstitial Fluid Heat Capacity on (fi*/fl> (2)
Regenerator Performance," ASME JOURNAL OF HEAT TRANSFER, Vol. 102,
Aug. 1980, pp. 572-574. where 5 denotes the fraction of the area covered by drops with sizes
10 Hausen, H., "Survey of the Heat Transfer Theories in Regenerators," greater than R* and exponent p has a value around 0.3. JS is the in-
Heat Exchangers: Design and Theory Sourcebook, McGraw-Hill, New York,
1974, pp. 207-222. stantaneous, effective maximum drop size in the dimensionless form
11 Romie, F. E., "ApproximatejTreatment of Transverse Conduction in and it grows with time t elapsed from the beginning of the transient
Regenerators," ASME JOURNAL OF HEAT TRANSFER, Vol. 102, Aug. 1980, dropwise condensation as [1]
pp. 574-576.
12 Bahnke, G. D., and Howard, C. P., "The Effect of Longitudinal Heat
Conduction on Periodic-Flow Heat Exchanger Performance," ASME Journal fl = Bi - (3)
of Engineering for Power, Vol. 86, Apr. 1964, pp. 105-119.
where Bi = hiRmByL/\i is the Biot number corresponding to the de-
parting drop size fimax and T is the time interval at which the con-
Effect of Knudsen Number on denser surface is swept by departing drops. Equation (2) is applicable
Dropwise Condensation to the following size range with the exception of the neighborhood of
the extremely small end where droplet sizes are comparable either
to the spacing between the nucleation sites or to the primary drop
Hiroaki Tanaka 1

The purpose of this technical note is twofold. One is to point out R*/ft £ 0.1 (4)
the effect of a Knudsen number on dropwise condensation. The other
When equation (2) is valid at the characteristic size R* = 2, we can
is to give a clear physical sketch of the mechanism of heat transfer in
estimate the instantaneous fraction / of the condenser surface which
dropwise condensation, which was discussed in previous papers [1,
can be utilized by the active droplets in the size range fl* g 2 without
2] in an exact but fairly sophisticated manner. We will begin with the
being occupied by drops in the size range fl* > 2 as
second part.
It seems that, under dropwise condensation of steam, heat is mainly f = 1 - 5(fl* = 2) = (2/fl>
transferred by microscopic active droplets whose sizes are small
The time average / of this fraction is calculated with the aid of equa-
enough for the heat flow rate through themselves to be virtually
tion (3) as
controlled by mass-transfer resistance at the liquid-vapor interface.
As to this resistance, the kinetic theory of gases gives the following
equivalent interfacial heat-transfer coefficient [3]. '-^.'"-(TO) -p/U+p) t
(i+p)
/Bi\-p

2a h,Is Here we consider the limiting case where the nucleation site density
h. (1)
2 - a (2TTRTS)^2 ueTs is infinitely high as well as the primary drop size is effectively zero.
(As a matter of fact, the thermodynamic critical size of a water droplet
where a is the condensation coefficient, R the gas constant, Ts the is about 2 X 10~ 5 mm at a subcooling of AT = 1.0°C. It is inversely
saturation temperature, h/g the heat of vaporization, and vg the spe- proportional to AT and is almost invariable of the saturation tem-
cific volume of the vapor. In Table 1, values of hi are calculated for perature.) In this case, equation (2) can be applied down to zero radius
water and mercury at several saturation temperatures. It should be and it predicts very naturally the complete coverage of the surface
noted that in this calculation, the condensation coefficient a was as- by drops. Then the fraction / is fully covered by the active droplets
sumed to be unity. Umur and Griffith [4] have achieved an analysis which convey heat at an almost uniform rate of 2/i; AT" per unit base
of heat conduction through a single drop of hemispherical shape of area according to the foregoing solution by Umur and Griffith. Fur-
radius r, with a uniform subcooling AT at the base and the heat- ther, if we neglect the contribution to heat transfer of drops in the size
transfer coefficient hi at the liquid-vapor interface. Here, in order to range fl* > 2 as a first approximation, we have for the time-averaged
make drop size r dimensionless, we may introduce a Biot number R* heat-transfer coeffficient of dropwise condensation
= hiv JXi, where X; is the thermal conductivity of the liquid. Then
Umur and Griffith's analytical result of heat-transfer rate through
a single drop is as a first approximation divided into the following two
regions by a characteristic size R* = 2. Namely, as to a droplet in a
Table 1
region R * < 2, the heat-transfer rate through the droplet can well be
expressed as Q = 2irr'lhiAT, because in this case thermal conduction Substance;
resistance through the droplet is almost negligible yielding a uniform firaax Ts hi L
observed K W/m 2 K mm mm
heat flux distribution of hi AT over the whole liquid-vapor interface.
As a result of this case the heat-transfer rate per unit base area of the 295 8.55 X 10 6 1.42 X 10- 3 1.6 X lO" 3
droplet becomes 2hiAT. On the other hand, for a drop within the Water; 306 1.45 X 10 6 8.56 X 10-" 8.8 X 10-"
flmax =
l--^ 321 2.78 X 10B 4.60 X 10-" 4.3 X 10-"
region R* > 2, the conduction inside the drop almost controls the heat
transfer, resulting in a much lower value than 2fe,AT for the heat- mm
342 6.10 X 10 8 2.17 X 10-" 1.8 X 10""
373 1.57 X 107 8.69 X 10~B 6.1 X 10- 5
1 378 4.35 X 10 3 4.88 1.4 X lO" 1
Associate Professor, Department of Mechanical Engineering, The Uni- Mercury; 412 1.74 X 10" 1.30 3.0 X 10" 2
versity of Tokyo, Bunkyo-ku, Tokyo, Japan. 452 6.62 X 10" 0.362 6.9 X lO" 3
Contributed by the Heat Transfer Division for publication in the JOURNAL flmax = 0.15
OF HEAT TRANSFER. Manuscript received by the Heat Transfer Division mm
492 1.95 X 10 s 0.129 2.0 X lO" 3
December 29,1980.

606 / VOL. 103, AUGUST 1981 Transactions of the ASME

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 4 Iliffe, M. A., "Thermal Analysis of the Contra-Flow Regenerative Heat transfer rate per unit base area of the drop. The characteristic drop
Exchanger," Proceedings of the Institute of Mechanical Engineers, Vol. 159, size in the actual dimension r i = 2X;//i;, which corresponds to fl* =
1948, pp. 363-372.
5 Lambertson, T. J., "Performance Factors of a Periodic-Flow Heat Ex- 2, is listed in the fourth column of Table 1.
changer," Trans. ASME, Vol. 80, Apr. 1958, pp. 586-592. The so-called steady dropwise condensation on an inclined surface
6 Nahavandi, A. N., and Weinstein, A. S., "A Solution to the Periodic-Flow is in reality composed of transient dropwise condensation which is
Regenerative Heat Exchanger Problem," Applied Scientific Research, Sec. taking place repeatedly on tracks left by departing drops. The in-
A, Vol. 10,1961, pp. 335-348.
7 Hausen, H., Warmeiibertragung im Gegenstrom, Gleichstrom und stantaneous drop-size distribution during this transient dropwise
Kreuzstrom, Springer-Verlag, Berlin, 1950, pp. 262-444. condensation occurring locally on a newly exposed surface follows a
8 Jakob, M., Heat Transfer, Vol. 2, John Wiley & Sons, New York, 1957, form [2]:
pp. 261-341.
9 Maclaine-cross, I. L., "Effect of Interstitial Fluid Heat Capacity on (fi*/fl> (2)
Regenerator Performance," ASME JOURNAL OF HEAT TRANSFER, Vol. 102,
Aug. 1980, pp. 572-574. where 5 denotes the fraction of the area covered by drops with sizes
10 Hausen, H., "Survey of the Heat Transfer Theories in Regenerators," greater than R* and exponent p has a value around 0.3. JS is the in-
Heat Exchangers: Design and Theory Sourcebook, McGraw-Hill, New York,
1974, pp. 207-222. stantaneous, effective maximum drop size in the dimensionless form
11 Romie, F. E., "ApproximatejTreatment of Transverse Conduction in and it grows with time t elapsed from the beginning of the transient
Regenerators," ASME JOURNAL OF HEAT TRANSFER, Vol. 102, Aug. 1980, dropwise condensation as [1]
pp. 574-576.
12 Bahnke, G. D., and Howard, C. P., "The Effect of Longitudinal Heat
Conduction on Periodic-Flow Heat Exchanger Performance," ASME Journal fl = Bi - (3)
of Engineering for Power, Vol. 86, Apr. 1964, pp. 105-119.
where Bi = hiRmByL/\i is the Biot number corresponding to the de-
parting drop size fimax and T is the time interval at which the con-
Effect of Knudsen Number on denser surface is swept by departing drops. Equation (2) is applicable
Dropwise Condensation to the following size range with the exception of the neighborhood of
the extremely small end where droplet sizes are comparable either
to the spacing between the nucleation sites or to the primary drop
Hiroaki Tanaka 1

The purpose of this technical note is twofold. One is to point out R*/ft £ 0.1 (4)
the effect of a Knudsen number on dropwise condensation. The other
When equation (2) is valid at the characteristic size R* = 2, we can
is to give a clear physical sketch of the mechanism of heat transfer in
estimate the instantaneous fraction / of the condenser surface which
dropwise condensation, which was discussed in previous papers [1,
can be utilized by the active droplets in the size range fl* g 2 without
2] in an exact but fairly sophisticated manner. We will begin with the
being occupied by drops in the size range fl* > 2 as
second part.
It seems that, under dropwise condensation of steam, heat is mainly f = 1 - 5(fl* = 2) = (2/fl>
transferred by microscopic active droplets whose sizes are small
The time average / of this fraction is calculated with the aid of equa-
enough for the heat flow rate through themselves to be virtually
tion (3) as
controlled by mass-transfer resistance at the liquid-vapor interface.
As to this resistance, the kinetic theory of gases gives the following
equivalent interfacial heat-transfer coefficient [3]. '-^.'"-(TO) -p/U+p) t
(i+p)
/Bi\-p

2a h,Is Here we consider the limiting case where the nucleation site density
h. (1)
2 - a (2TTRTS)^2 ueTs is infinitely high as well as the primary drop size is effectively zero.
(As a matter of fact, the thermodynamic critical size of a water droplet
where a is the condensation coefficient, R the gas constant, Ts the is about 2 X 10~ 5 mm at a subcooling of AT = 1.0°C. It is inversely
saturation temperature, h/g the heat of vaporization, and vg the spe- proportional to AT and is almost invariable of the saturation tem-
cific volume of the vapor. In Table 1, values of hi are calculated for perature.) In this case, equation (2) can be applied down to zero radius
water and mercury at several saturation temperatures. It should be and it predicts very naturally the complete coverage of the surface
noted that in this calculation, the condensation coefficient a was as- by drops. Then the fraction / is fully covered by the active droplets
sumed to be unity. Umur and Griffith [4] have achieved an analysis which convey heat at an almost uniform rate of 2/i; AT" per unit base
of heat conduction through a single drop of hemispherical shape of area according to the foregoing solution by Umur and Griffith. Fur-
radius r, with a uniform subcooling AT at the base and the heat- ther, if we neglect the contribution to heat transfer of drops in the size
transfer coefficient hi at the liquid-vapor interface. Here, in order to range fl* > 2 as a first approximation, we have for the time-averaged
make drop size r dimensionless, we may introduce a Biot number R* heat-transfer coeffficient of dropwise condensation
= hiv JXi, where X; is the thermal conductivity of the liquid. Then
Umur and Griffith's analytical result of heat-transfer rate through
a single drop is as a first approximation divided into the following two
regions by a characteristic size R* = 2. Namely, as to a droplet in a
Table 1
region R * < 2, the heat-transfer rate through the droplet can well be
expressed as Q = 2irr'lhiAT, because in this case thermal conduction Substance;
resistance through the droplet is almost negligible yielding a uniform firaax Ts hi L
observed K W/m 2 K mm mm
heat flux distribution of hi AT over the whole liquid-vapor interface.
As a result of this case the heat-transfer rate per unit base area of the 295 8.55 X 10 6 1.42 X 10- 3 1.6 X lO" 3
droplet becomes 2hiAT. On the other hand, for a drop within the Water; 306 1.45 X 10 6 8.56 X 10-" 8.8 X 10-"
flmax =
l--^ 321 2.78 X 10B 4.60 X 10-" 4.3 X 10-"
region R* > 2, the conduction inside the drop almost controls the heat
transfer, resulting in a much lower value than 2fe,AT for the heat- mm
342 6.10 X 10 8 2.17 X 10-" 1.8 X 10""
373 1.57 X 107 8.69 X 10~B 6.1 X 10- 5
1 378 4.35 X 10 3 4.88 1.4 X lO" 1
Associate Professor, Department of Mechanical Engineering, The Uni- Mercury; 412 1.74 X 10" 1.30 3.0 X 10" 2
versity of Tokyo, Bunkyo-ku, Tokyo, Japan. 452 6.62 X 10" 0.362 6.9 X lO" 3
Contributed by the Heat Transfer Division for publication in the JOURNAL flmax = 0.15
OF HEAT TRANSFER. Manuscript received by the Heat Transfer Division mm
492 1.95 X 10 s 0.129 2.0 X lO" 3
December 29,1980.

606 / VOL. 103, AUGUST 1981 Transactions of the ASME

Copyright © 1981 by ASME
Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 for all active droplets. Further, we assume as before the limit of the
h= 2hi(l+p)\—\ complete coverage of the surface by drops. Then, the fraction / of the
condenser surface which is completely covered by active droplets
If we put p = 0.3, we finally have conveys heat per unit area not at the foregoing rate of 2/i; AT but at
h/ht = 3.20Bi-°- 3 (5) the reduced rate of hi At, because in this case the whole microscopic
area of the liquid-vapor interface is not uniformly effective but the
In the previous paper [2], the following equation (its original form macroscopic superficial area seems to be effectively concerned with
is Nu = 5.64^ 1 -0 - 7 where Nu = W? max /X; and £i = 2/Bi) was derived mass transfer [6] (a hemisphere has a projected area equal to \ the
for the limiting case of the complete coverage of the surface by drops outside area). As a result of this case, reducing the numerical constant
after a fairly troublesome theoretical process. of equation (6) by a factor of two, we have as a first approximation
h/ht = 3.47Bi-°- 3 (6) h/hi = 1.74Bi-°-3 (8)
We find that the resultant equation (5) of the foregoing bold ap- In the previous paper [2], the author estimated the condensation
proximation agrees with the exact result (6) fairly well. Here it must coefficient of water at around 0.2 by comparing available experimental
be noted that equation (6) is valid with a collateral condition (7), data obtained at various pressures with the prediction from equation
because R*/R with R* = 2 and R in equation (3) has to satisfy con- (6) on the assumption that the nucleation site density is infinitely
dition (4) during a main portion of time t over the sweeping period high. However, if we employ equation (8) on this occasion, the con-
T. densation coefficient of water is a = 0.45.
Finally the case of mercury will be briefly mentioned, as it offers
Bi» 1 (7)
a very special case. From Table 1, the characteristic size r\ far exceeds
This condition is well satisfied by water, while it completely fails of the departing drop size flmax especially at low temperatures. This
mercury (see Table 1 by taking note of Bi = 2fl m a x /ri). means that heat conduction resistance through any drop can be ne-
In the previous paper [t, 2] the Nusselt number hRmax/\i was in- glected as compared with the interfacial mass-transfer resistance.
troduced to nondimensionalize the heat-transfer coefficient for Further, at the lowest temperature in Table 1, the mean free path L
dropwise condensation h. It seems more reasonable, however, to is nearly the size of the departing drop. Hence, for the majority of
employ the ratio of the heat-transfer coefficients, h/h,, as in equation droplets that exist on the condenser surface, the Knudsen numbers
(6), because this ratio is directly related to the fraction of the con- are greater than unity. As a result, the heat-transfer coefficient for
denser surface covered by the active droplets, as has been made clear dropwise condensation will take the same value as the interfacial
by the foregoing discussion. heat-transfer coefficient hi. This is the reason why there is no ap-
Now we have come up to the stage to discuss the effect of Knudsen preciable difference between the experimentally obtained heat-
number. The interfacial matter transfer characterized by equation transfer coefficient for dropwise condensation and that for filmwise
(1) is in its origin based on the kinetic theory of gases. Then we become condensation of mercury a t low temperatures [7].
aware that the foregoing theory [4] of heat conduction through a single
drop by Umur and Griffith, in which a uniform heat-transfer coeffi-
References
cient hi was assumed at the liquid-vapor interface, holds true in the 1 Tanaka, H., "A Theoretical Study of Dropwise Condensation," ASMK
JOURNAL OF HEAT TRANSFER, Vol. 97,1975, pp. 72-78.
case where the size of the drop r is much greater than the mean free 2 Tanaka, H., "Further Developments of Dropwise Condensation Theory,"
path L of molecules in the vapor phase. In other words a Knudsen ASME JOURNAL OF HEAT TRANSFER, Vol. 101,1979, pp. 603-611.
number defined by Kn = L/r should be much smaller than unity. At 3 Schrage, R. W., A Theoretical Study of Interphase Mass Transfer, Co-
this juncture, values of the mean free path are listed in the last column lumbia University Press, New York, 1953.
of Table 1 [5]. We find from Table 1 that, for the case of water, the 4 Umur, A., and Griffith, P., "Mechanism of Dropwise Condensation,"
ASME JOURNAL OF HEAT TRANSFER, Vol. 87,1965, pp. 275-282.
mean free path L is almost equivalent to the characteristic drop size 5 Kennard, E. H., Kinetic Theory of Gases, McGraw-Hill, New York,
r\. Therefore, as for the most of the active droplets in the size range 1938.
R* =£ 2 i.e. r £ n, the Knudsen number becomes greater than unity 6 Melville, H. W., "A Note on Evaporation from Irregular Surfaces,"
and Umur and Griffith's theoretical result proves no longer appli- Transactions of the Faraday Society, Vol. 32,1936, pp. 1017-1020.
7 Necmi, S., and Rose, J. W., "Heat-Transfer Measurements During
cable. Dropwise Condensation of Mercury," International Journal of Heat and Mass
For the sake of brevity we may consider the case where Kn > 1 holds Transfer, Vol. 20,1977, pp. 877-881.

Journal of Heat Transfer AUGUST 1981, VOL 103 / 607

Downloaded 20 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm