Proceedings of the World Congress on Engineering 2011 Vol III
WCE 2011, July 6 - 8, 2011, London, U.K.
Flame Radiation Characteristics of Open
Hydrocarbon Pool Fires
E. Ufuah and C. G. Bailey
Abstract— The fundamental subject of fire research with Generally, for the range 0.2m ≤ D ≤ 1.0m, the radiative
problems involving hydrocarbon pool fires focuses on thermal mode of burning is said to be optically thin while for pool
radiation from the flame surface. Smoke obscuration and pool diameters greater than 1.0m the fire is known to have become
fire size are parameters of influence. The object is to establish
optically thick and turbulent [1,5]. The object is to establish
the temperature and heat flux profiles, and assess the hazard
consequences that may arise from these fire actions. This the temperature and heat flux distribution together with the
investigation, therefore, provides a correlation for the surface hazard consequences deriving from such events. In this
emissive power of flame based on observed data taken from five investigation, a correlation for thermal radiation from
different experimental studies. It is shown in this study that due hydrocarbon pool fires is developed based on the
to spatial and temporal variations associated with trials methodology utilised by Shokri and Beyler [8]. The validity
conducted at different times and locations on free-burning pool
of the model is drawn from a set of data for pool fire radiation
fires, model predictions for surface emissive power of flame can
be very different. The present correlation provides a reasonable to nearby structures situated at ground level. A
prediction for liquefied natural gas and aviation fuels compared comprehensive algorithm for the computation including the
to those of Shokri and Beyler, and Mudan and Croce. However, relevant assumptions is presented, and the model is compared
the investigation reveals that a coefficient of variation between 6 with those of Shokri and Beyler [8], Mudan and Croce [9]
and 17 per cent can be found by separately adopting the flame and McGrattan et al. [10].
height models of Heskestad and Thomas for fire diameters up to
An explicit account of wind effect on flame geometry has
120 m. This highlights the need to exclusively utilize a given
methodology, which gives a clear description of all sub- models not been considered in the present model since the
used in the derivation of surface emissive powers of correlations on which a comparison is drawn are based on
free-burning hydrocarbon pool fires. It is noted that the quiescent air conditions. However, if wind effect will be
significance of thermal radiation model also broadens the considered the flame height correlation of Heskestad [11]
means by which the acceptable separation distances between does no longer apply. An alternative geometrical
radiation source and targets _ people and structures can be
characterization of the flame model can be made using the
determined.
models proposed by Thomas [12] and Moorhouse [13].
Index Terms— Flame height, pool fire diameter, radiation Apparently, when developing the effective emissive power of
model, soot. a flame it is vital to exclusively adopt a given methodology.
Table 1 summarises the discrepancies in the flame height
model based on Heskestad [11] and Thomas [12] for the zero
I. INTRODUCTION wind conditions using gasoline pool fuel in the diameter, D
The primary subject of research with problems involving range of 1.0m ≤ D ≤ 120m. The differences account for a
hydrocarbon pool fires centres on thermal radiation from the coefficient of variation between 6 and 17 percent. This is
pool surface. Over the last few decades there have been sufficient to introduce errors in the prediction of flame
numerous experimental studies conducted to establish behaviour if not correctly implemented, particularly for
thermal radiation models from large pool fires [1]-[7]. To smaller pool diameters as depicted in Fig. 1. Fig. 2 illustrates
address the above subject, it is necessary to consider how big various flame height models that can be utilized to compute
and fast the fire is burning as well as the consequent effect of radiant heat flux to nearby structures. The effect of wind has
smoke on radiation and wind on the resultant flame shape. been included in the correlations of ‘Thomas 2’ and
Systematically, one may look at the operating fire regime that ‘Moorhouse’ as shown in Fig. 2.
characterizes the scope within which radiation is seen to
dominate in heat transfer. Babrauskas [1] explains that for
diameters of pool fire, (D > 0.2m) heat transfer is dominated
by radiation. Fuel type is known to be a single factor that
characterizes the transition between optically thin and
optically thick flame.
E. Ufuah is a PhD research student at the University of Manchester, UK, on
study leave from Ambrose Alli University, Ekpoma, Nigeria (email:
[email protected])
C. G. Bailey is a Professor of Structural and Fire Engineering at the
University of Manchester, UK (email:
[email protected])
ISBN: 978-988-19251-5-2 WCE 2011
ISSN: 2078-0958 (Print); ISSN: 2078-0966 (Online)
Proceedings of the World Congress on Engineering 2011 Vol III
WCE 2011, July 6 - 8, 2011, London, U.K.
Fig. 1 Degree of variation between Heskestad and Thomas 1
flame height models for gasoline pool fire u w3 a 3
u
(4)
gDm
Beyler [15] remarks that the Heskestad [11] correlation
under quiescent air conditions best represents large diameter
pool fires. The correlation of Moorhouse [13] based on
several large scale test of liquefied natural gas (LNG) pool
fires under wind conditions can be represented in the form:
H
6.2Fr
0.254
u10 0.044 (5)
D
Fig. 2 Comparison of flame height correlations for gasoline
Where Fr is the non-dimensional combustion Froude’s
pool fire
number as given by equation (3) and u10 is the
non-dimensional wind velocity of equation (4) with the wind
speed measured at a height of 10m above the ground.
II. POOL FIRES AND FLAME GEOMETRY
Heskestad flame height model for zero wind condition is
The properties of open pool fires are markedly different commonly expressed as follows:
from those within enclosures. For pool fires, the spill could
result from the rupture of a pipe or container as well as from 2
possible overfilling of a storage tank [7]. When this happens, H 0.23Q 5
1.02 D (6)
there is the likelihood of fire occurring subject to possible
ignition sources. Different types of fire likely to occur in the
offshore following accidental releases of flammable
III. THERMAL RADIATION CHARACTERISTICS
hydrocarbon fuels can be found in the handbook of society of
fire protection engineers (SFPE) [14]. It is expected that a The flame spread rate for pool fires is considered to be
flammable liquid spill will continue to spread until it is liquid-controlled owing to its low initial momentum. The
bunded or at equilibrium when the burning rate of the fuel maximum liquid-phase spread velocity of pool fires is
equals the spread rate. Moreover, the parameters that conservatively assumed as 0.1 m/s [14], [16]. Radiations
characterize the geometry of free pool fire flames include the from hydrocarbon fires come mostly from the glowing soot
flame height and the equivalent pool fire diameter. Fig. 3 from within the flames. An exemption can be found for some
illustrates the schematics of geometrical shapes of pool fire clean burning pool fires such as methanol, in which thermal
models. The tilt and drag of the pool flame are dictated by radiations come from the hot gases namely; carbon dioxide
wind effect. and water vapour. For large diameter soot-producing
hydrocarbon pool fires, copious amount of soot can be
produced thereby creating the tendency for the unburnt soot
to escape from the flame. The soot then congregates and
forms a film around the flame surface, thus limiting the
radiation to external structures. Fig. 4 represents the classical
and the modified cylindrical flame models for pool fires by
the point source and surface emitter concepts. Most
modellers have considered the following assumptions to
Fig. 3 Schematic of cylindrical flame models. (a) Still air characterize radiation models for soot producing flames [3],
flame model (b) Flame model under wind conditions [7], [16]-[17]:
Surface emissive power, SEP
The most widely used flame height correlations are those Radiative fraction of the total energy of combustion
of Heskestad [11], Thomas [12] and Moorhouse [13]. Atmospheric transmissivity and absorptivity
Equation (1) is used for still air investigation while that of (2)
can be used under wind conditions as proposed by Thomas
[12].
H
42Fr
0.61
(1)
D
H
55Fr
0.67
u 0.21 (2)
D
where H is flame height, D is fire diameter
m
Fr (3) Fig. 4 Pool fire thermal radiation models (a) modified
a gD cylindrical flame model (b) classical cylindrical flame model
(c) point source model
ISBN: 978-988-19251-5-2 WCE 2011
ISSN: 2078-0958 (Print); ISSN: 2078-0966 (Online)
Proceedings of the World Congress on Engineering 2011 Vol III
WCE 2011, July 6 - 8, 2011, London, U.K.
The emissive power of a flame can be modelled based on IV. SHAPE FACTOR COMPUTATION
the point source theory and the surface emitter theory. While The amount of radiation reaching the surfaces of external
point source concept appears to be overly conservative, the targets from pool fire flames vary with shape and position of
surface emitter theory averages the emissive power of the the fire relative to the target. These factors are commonly
flame over the entire surface of interest. Cowley and Johnson known as shadow and position effects. Generally, the shape
[7] highlighted on three different ways by which the emissive
factor F12 , between the flame source and the target can be
power, SEP of a flame can be determined, namely;
Spot measured emissive power determined from the following correlation [19].
Average measured SEP
Model average measured SEP cos 1 cos 2
F12 dA1 (11)
The model average surface emissive power appears to be 2
S R
most commonly used in conjunction with the simplified
geometrical shapes assumed for hydrocarbon fires. Using this where 1 and 2 are the angles between the normal to the
concept, however, the surface areas of the model flame surfaces and the line of sight for the emitter and the receiver
shapes seem to be larger than those of the visible flame respectively.
surfaces [7]. The point source model of Drysdale [18] based R = distance between the flame source and the target along
on the narrow angle radiometer data for the radiant heat flux the line of sight.
to external target can be expressed as follows: dA1 = the elemental surface area of the emitter viewed by the
receiver.
r Q cos Using the mathematical form of (11), the following
q (7) correlations have been derived to calculate the shape factors
2
4R in still air for targets at the base or top level of the flame [20].
Where Q = the total combustive energy For horizontally placed targets, (12) is used whereas in the
case of vertically positioned targets, (13) can be
χr = radiative fraction
implemented.
θ = the angle between the normal to the external object and
the line of sight between the object and the point source
1
location B
F12 , H S B 1 S 1
R = the sight distance between the fire point source location arctan
and the object.
B2 _1 B 1 S 1 (12)
is a
For liquid hydrocarbon fuels, the heat release rate Q 1
A
, its lower heat of
function of the fuel mass loss rate m
S
arctan
A 1 S 1
A 1 A 1 S 1
combustion hc and the pool size D as follows:
Q m hc A (8a)
F12 ,V
1
arctan
h h
arctan
S 1
S 1 S
m m 1 exp k D (8b)
S S 1 (13)
Ah A 1S 1
Where m = fuel maximum mass loss rate, kβ is a property arctan
of the fuel defined as the mean beam length corrector-flame
S A 2
1 A 1S 1
attenuation coefficient product. Where
The surface area of a pool fuel can be determined using the 2L , 2H , h2 S 2 1 1 S 2
S h A , B
correlation of Gottuck and White [14] for liquid pool and D D 2S 2S
spill fires. L = distance between centre of pool fire and edge of target
H = flame height
A p 1.4V ; for pool 95 litres D = equivalent flame diameter
(9)
0.36V ; for pool 95 litres V. THERMAL RADIATION MODEL
In the case of an unconfined spill, Gottuck and White [14] The thermal radiation model proposed in the present study
proposed that the resultant pool size can be taken as 55 per adopts the Heskestad flame height correlation and uses the
cent larger than the corresponding confined pool fire size. methodology for surface emitter models. Shokri and Beyler
[8], Drysdale [18] and Mudan and Croce [9] present various
As 1.55 A p (10) algorithms to estimate the heat flux from flames to external
An alternative correlation that can be used to determine targets. Shokri and Beyler [8] proposed the following
fuel maximum mass loss rate and maximum regression rate correlation to calculate thermal radiation from hydrocarbon
can be found in [14]. pool fires.
SEP 58 10 0.00823* D
(14)
ISBN: 978-988-19251-5-2 WCE 2011
ISSN: 2078-0958 (Print); ISSN: 2078-0966 (Online)
Proceedings of the World Congress on Engineering 2011 Vol III
WCE 2011, July 6 - 8, 2011, London, U.K.
Indeed, this correlation appears to under-predict the
surface emissive power for liquefied natural gas fuel as
demonstrated in Fig. 5. According to Mudan and Croce [9] a
uniform surface emissive power of flames for smoky
hydrocarbon fuels can be determined as follows.
SEP E max . exp( kD ) E s 1 exp( kD (15)
where Emax = equivalent ideal radiator emissive power of fuel,
140KW/m2
Es = maximum smoke emissive power, 20KW/m2 Fig. 6 Comparison of model predictions with measured data
k = flame extinction coefficient, 0.2m-1 at ground level
D = flame diameter The statistical correlations commonly considered
appropriate for comparing observed data against predicted
models were used according to [4],[21],[22]. These models
are shown in the appendix and the computation is illustrated
in Table 2. The assumptions made in formulating the present
model include:
The prevailing wind velocity is less than the critical
velocity.
The cylindrical flame geometry is based on
Heskestad model.
Fig. 5 Surface emissive power models as a function of fire The spill fuel under consideration is confined.
diameter For the purpose of considering the effect of soot
production on thermal radiation from the flame
The data set used to develop the present model comprises surface, the clear flame height can be taken as one
most of the results of trials utilized by Shokri and Beyler [8]. half of the Heskestad geometric flame model.
The motive for selecting this data set is to have a basis for The nomenclature - gasoline 1 and diesel 1 as given in
acceptable comparison between the models. Table 1 Figs. 7 and 8 are used to characterise the flame envelope
summarises the data set used in the present formulation. A parameters based on measured data that were utilized to
total of 15 trials and 59 data points are included with the pool determine the angle factors. Similarly, gasoline 2 and diesel 2
diameter varying from 1.22 to 24.1m. The effective emissive characterise parameters based on the calculated flame height.
power for each trial was based on the number of data points The calculated flame height model over-predicts flame
shown in Table 1. Using the method of least squares, the geometry compared with the observed data. However, it is
following model is proposed for the radiant heat flux to shown in Fig. 7 that a similar trend is achieved in the
external targets. prediction of radiant heat flux to external structures. Suppose
the various models are extended to cover pool diameters up to
1 .61 300m, it becomes obvious that the Shokri and Beyler
L
q 20 . 7 (16) correlation zeros out as can be seen in Fig. 10. For diameter
D approximately 50m, the Mudan and Croce model reduces to
The model of (16) accurately predicts the heat flux for 20KW/m2, which is the maximum surface emissive power of
LNG and aviation fuels fairly compared to that of Shokri and smoke. In a similar manner, McGrattan et al. proposes that
Beyler [8] as illustrated in Fig. 6. It may be used for initial the surface emissive power of the flame should be taken as
fire design and evaluation. The proposed flame surface 100KW/m2 irrespective of the pool size. To account for the
emissive power based on the studied fuels is given by the smoke obscuration effect, the Heskestad modified flame
following expression. height model, H* was used according to (19). The luminous
factor is taken as 0.5. This concept is based on the fact that
SEP 70 exp kD (17) experiments have demonstrated that radiant heat fluxes at
flame heights within pool fire diameters are relatively
Where k = average flame extinction coefficient, 0.00165m-1.
constant [3,5].
D = equivalent fire diameter, m
The incident heat flux can, therefore, be calculated from 2
the following expression: H 0.12Q 0.51D
5
(19)
In order to further assess the radiative energy flux
q F12 * SEP (18) predictable from the various models, radiant heat flux from
flame surface is plotted against pool fire diameter as
illustrated in Fig. 11. It is the radiant heat flux per unit length
of equivalent flame perimeter (calculated as SEP multiplied
by the corresponding flame height). There are situations
when it becomes necessary to limit the amount of radiation
ISBN: 978-988-19251-5-2 WCE 2011
ISSN: 2078-0958 (Print); ISSN: 2078-0966 (Online)
Proceedings of the World Congress on Engineering 2011 Vol III
WCE 2011, July 6 - 8, 2011, London, U.K.
reaching external targets so that the acceptable separation In the present method, an evaluation of thermal barrier is
distance can be reduced. When these barriers are constructed, addressed by considering the modified Heskestad flame
the view from the targets is altered. McGrattan et al. [10] height model using a luminous factor of 0.5 to calculate the
recommend that when analysing for thermal barriers, the SEP angle factors. The results from this correlation are found to be
may be halved while the clear flame height is doubled so that reasonable when compared with those of McGrattan et al.
a constant heat flux per unit length of flame is yet maintained. [10] model.
Table 1 Summary of geometric and radiative properties of the studied fuel fires
Fuel Type Fire Diameter Data Points Average Emissive Power Source
[m] [KW/m2]
Gasoline 1.5 4 59.23 Muñoz et al.[4]
Gasoline 3.0 4 76.24 Muñoz et al.[4]
Gasoline 4 4 81.76 Muñoz et al.[4]
Diesel 3 4 67.57 Muñoz et al.[4]
Diesel 4 5 71.71 Muñoz et al.[4]
JP-4 1.22 3 69.06 Fu[5]
JP-4 2.44 5 44.63 Fu[5]
JP-5 2.44 3 46.47 Fu[5] /Dayan & Tien[3]
JP-5 3.05 6 79.99 Fu[5] /Dayan & Tien[3]
JP-5 5.5 4 53.31 Fu[5] /Dayan & Tien[3]
LNG 14.64 4 49.06 May & McQUEEN[6]
LNG 16.17 4 50.56 May & McQUEEN[6]
LNG 18.3 4 53.76 May & McQUEEN[6]
LNG 20.13 1 49.44 May & McQUEEN[6]
LNG 24.1 3 44.55 May & McQUEEN[6]
Table 2 Statistical matrix to compare the various models with observed data
Model Fractional Bias Normalised Mean Square Error
Present model -0.059 0.0743
Shokri & Beyler [8] 0.223 0.144
Mudan & Croce [9] -1.181 7.241
Fig. 7 Comparison of predicted radiant heat flux with
Fig. 9 Comparison of predicted radiant heat flux with
experimentally measured data for the present correlation
experimentally measured data for the correlation of Mudan
and Croce
Fig. 8 Comparison of predicted radiant heat flux with
experimentally measured data for the correlation of Shokri Fig. 10 Prediction range of surface emissive power models
and Beyler
ISBN: 978-988-19251-5-2 WCE 2011
ISSN: 2078-0958 (Print); ISSN: 2078-0966 (Online)
Proceedings of the World Congress on Engineering 2011 Vol III
WCE 2011, July 6 - 8, 2011, London, U.K.
no matter the size of the fire, the clear flame height will never
exceed the numerical value of 6.4x10-3q”. This may not truly
represent the flame behaviour because flames of fire
diameters up to 50m can still produce thermal radiations at
heights quite above that limit.
Remarkably, the Heskestad flame height model provides a
sensible path to the prediction of thermal radiation for some
cryogenic hydrocarbon pool fuels such as LNG and aviation
fuels such as JP-5. The large deviation from observed data by
Fig. 11 Prediction range of radiant heat flux as a function of
the Mudan and Croce model suggests that most of the data
fire diameter for a gasoline pool fire
used to calibrate their model and those analysed in the present
model are quite different. The much over-prediction of the
Mudan and Croce model, as revealed by the statistical
VI. ILLUSTRATION
measures, is contributed mostly by the LNG fuel since their
The example problem analysed by McGrattan et al. [10] is correlation is only for a smoky flame. It can be concluded that
chosen to describe the use of the proposed correlations in thermal radiation prediction is spatially and temporally
evaluating the effectiveness of thermal barriers. It involves a dependent, and also subject to uncertainties inherent in the
gasoline tank farm separated from a shopping complex by a modelling techniques and assumptions.
road. The following data are provided:
Dike width = 60m
Distance of target from the flame edge = 55m for personnel
Distance of target from the flame edge = 85m for building APPENDIX
It is proposed that a 7.3m vertical barrier will be built at the 1 N Xo X p
boundary of the tank farm to help screen the target from FB 2*
N 1 Xo X p
thermal radiation. The McGrattan et al. [10] model assumes
that the maximum flame height is 6.4x10-3q”. Thermal 1 N Xo X p 2
radiation to external target is calculated to be 7.9KW/m2 with NMSE
N 1 XoX p
no thermal barrier, and 5.9KW/m2 when a thermal barrier of
height 7.3m is constructed. The present method uses the Where FB = Fractional bias
modified flame height model of (19) to determine the NMSE = Normalised mean square error
luminous flame height, thus eliminating the conservatism X0 = Measured value
inherent in the conventional model. Based on this concept, Xp = Predicted value
the heat flux incident on external targets has been calculated N = Number of data points
as 8.10KW/m2 with no barrier and 5.78KW/m2 when a
barrier of height 7.3m is considered. ACKNOWLEDGMENT
E.Ufuah wishes to thank the federal government of Nigeria
under the platform of petroleum technology development
VII. DISCUSSION AND CONCLUSION funds (PTDF) and the management of Ambrose Alli
Thermal radiation from pool fire flames and the radiant University, Ekpoma for their financial support.
heat flux to adjacent structures have been developed. For the
fuel types considered in this study, the present model predicts REFERENCES
fairly well compared to the correlations of Shokri and Beyler [1] V. Babrauskas, “Estimating large pool fire burning rate”, Fire
[8], and Mudan and Croce [9] particularly for LNG fuel. Technology, vol. 19, (4), 1983, pp. 251-261.
[2] V. Babrauskas, Heat Release Rate, SFPE Handbook of Fire Protection
However, it should be noted that these models are exclusively Engineering, 3rd ed., National Fire Protection Association, Quincy,
dependent on the data used to calibrate them. Although, Figs. Ma., Section 3, 2002, pp. 3-1 to 3-37.
7 and 8 appear to be similar it can be easily surmised from [3] A. Dayan, and C. L. Tien, “Radiant heating from a cylindrical fire
column, Combustion Science and Tech., vol. 9 (1), 1974, pp. 41-47.
Fig. 5 that the present formulation presents a more precise [4] M. Muñoz, J. Arnaldos, J. Casal, and E. Plana, “Analysis of geometric
prediction for thermal radiation from a flame surface. and radiative characteristics of hydrocarbon pool fires”, Combustion
Moreover, the fractional bias of Table 2 shows that the and Flame, vol. 139, 2004, pp. 263-277.
[5] T. T. Fu, “Heat radiation from fires of aviation fuels”, Fire Technology,
present model over-predicts the incident heat flux. The model vol. 10 (1), 1974, pp. 54-67.
of Shokri and Beyler [8], however, is shown to under-predict [6] W. G. May, and W. McQUEEN, “Radiation from large liquefied
the heat flux. The over-prediction shown by the present natural gas fires”, Combustion Science and Technology, vol. 7 (2),
1973, pp 51-56.
model is negligible when compared with that of Mudan and
[7] L. T. Cowley, and A. D. Johnson, Oil and Gas Fires: Characteristics
Croce [9]. It is obvious in Figs. 5 and 9 that although the and impact, Offshore Technology Information, Report No. 596, 1992,
surface emissive power model of Mudan and Croce The Steel Construction Institute.
under-predicts the surface radiation for the LNG fuel, the [8] M. Shokri, and C. L. Beyler, “Radiation from large pool fires”, J. Fire
Protection Engineering, vol. 1, 1989, pp. 141-150.
incident heat flux to external targets is still over-predicted. [9] K. Mudan, and P. Croce, Fire Hazard Calculations for Large Open
This is due to the fact that LNG fuel data were not used to Hydrocarbon Fires, SFPE Handbook of Fire Protection Engineering,
calibrate their correlation. Besides, one of the prominent 1st ed., National Fire Protection Association, Quincy, Ma., Section 2,
1988, pp 2-45 to 2-87.
assumptions behind McGrattan et al. [10] correlation is that
ISBN: 978-988-19251-5-2 WCE 2011
ISSN: 2078-0958 (Print); ISSN: 2078-0966 (Online)
Proceedings of the World Congress on Engineering 2011 Vol III
WCE 2011, July 6 - 8, 2011, London, U.K.
[10] K. B. McGrattan, H. R. Baum, and A. Hamins, Thermal Radiation from
Large Pool Fires, National Institute of Standard and Technology,
Report No. NISTIR 6546, 2000.
[11] G. Heskestad, “Luminous height of turbulent diffusion flames”, Fire
Safety Journal, vol. 2, 1983, pp. 103-108.
[12] P. H. Thomas, “The size of flame from natural fires”, Symposium
(International) on Combustion, vol. 9, 1963, pp. 844-859.
[13] J. Moorhouse, “Scaling criteria for pool fires derived from large-scale
experiments”. Inst. Chemical Eng. symposium, 1982, pp. 165-179.
[14] D. T. Gottuk, and D. A. White, Liquid Fuel Fires, SFPE Handbook of
Fire Protection Engineering, 3rd ed., National Fire Protection
Association, Quincy, Ma., Section 2, 2002, pp 2-297 to 2-316.
[15] C. L. Beyler, “Fire plume and ceiling jets”, Fire Safety Journal, vol. 11,
1986, pp. 53-75.
[16] J. A. Fay, “Models of large pool fires”, J. Hazardous Materials, vol.
B136, 2006, pp. 219-232.
[17] P. K. Raj, “Large LNG fire thermal radiation: Modelling issues and
hazard criteria revisited”, Process Safety Progress, vol. 24, 2005, pp.
192-202.
[18] D. Drysdale, An Introduction to Fire Dynamics, 2nd ed., John Wiley and
Sons, New York, 1999.
[19] European Committee for Standardization CEN, Eurocode 1, Actions on
structures- Part 1-2: BS EN 1991-1-2, General actions-Actions on
structures exposed to fire, 2002.
[20] K. S. Mudan, “Geometric view factors for thermal hazard assessment”,
Fire Safety Journal, vol. 12, 1987, pp. 89-96.
[21] S. R. Hanna, D. G. Strimaitis, and J. C. Chang, “Evaluation of fourteen
hazardous gas models with ammonia and hydrogen fluoride data”, J.
Hazardous Materials, vol. 26, 1991, pp. 127-158.
[22] P. J. Rew, W. G. Hulbert, and D. M. Deaves, “Modelling of thermal
radiation from external hydrocarbon pool fires”, Trans IChemE, vol.
75, 1997, pp. 81-89.
[23] G. Heskestad, Fire Plumes, Flame Height, and Air Entrainment, SFPE
Handbook of Fire Protection Engineering, 3rd ed., National Fire
Protection Association, Quincy, Ma., Section 2, 2002, pp 2-1 to 2-17.
ISBN: 978-988-19251-5-2 WCE 2011
ISSN: 2078-0958 (Print); ISSN: 2078-0966 (Online)
