/h’ .

j
i . .

1’ 8

NASA Conference Publication IO181

Seals Code
Development
Workshop

Proceedings of a workshop held at

NASA Lewis Research Center
Cleveland, Ohio
June 14-15,1995
 NASA Conference Publication IO181

Seals Code
Development
Workshop

Proceedings of a workshop
held at and sponsored by
the NASA Lewis Research Center
Cleveland, Ohio
June 14-15,1995

National Aeronautics and

Space Administration
Office of Management
Scientific and Technical
Information Program
1996
 PREFACE

1. Large changes in direct operating costs plus interest (DOC +I) for a small investment in seal
technology particularly for small engines are supported by OEM’s and well documented by
Munson and Steinetz. For example, development of a film riding rim seal could reduce SFC by
1.8% in the regional engine and 5% in the turboshaft engines.

1.1. Engine secondary flows, cooling and parasitic, are 15 to 18 percent of the power stream
flows (up to 3 % per blade row); parasitic losses are 5 to 6% for regional and 8 to 9 % for
turboshaft. Seals and multiply connected cavities represent the interactive fluid dynamic
interface coupling.

1.2. Effective thermal management of engine flows calls for a clean sheet design approach.

2. Many IHPTET I1 goals have been met in tests, however, IHPTET I11 goals are difficult to
achieve. And, without the seals/secondary/primary flow interactions program, phase-I11 goals
can not be met - neither can AST or HSR goals. Critical sealing parameters include:

High surface speeds 1650 Wsec

High operating temperatures 1600 F
Larger structural and aerodynamic loading (up to 2x for HSR)
Transonic flow fields

3. All rotating systems have runout and dynamic unbalance. Seals should be and in many cases
are used to enhance rotor stability

3.1 All seals must afford compliance either by clearance control or compliant materials.
Reduced engine dynamics or tolerance to shaft excursion implies closer control of leakages
Reduction of leakages implies increases in engine pressure ratios, larger operating envelopes, and
higher efficiencies.

3.2 Smaller clearances imply increased sensitivity to dynamics, e.g., unshrouded rotor stages.

3.3 Applications of seal enhanced rotor stability to magnetic bearings will reduce dynamic
bearing load, control power, and backup bearing requirements as well as dampers.

3.4 Seals were the nemeses of the SSME turbopumps until Black, von Pragneau, Fleming,
Childs, Hendricks, Rocketdyne, NASA et al. provided data and developed methods to stabilize
these turbomachines seals and bearings. The success of the program represents a major
milestone in Space Exploration.

4. Significant benefits of tribopairing of interface materials can be derived in general and for
brush seals in particular.

...
111
4.1 Bristle and bare shaft material transfers in both heated air and cryogenic testing leaves
interface wear and life to be the Achilles heal of the brush seal.

4.2 Coatings for cryogenic operations as well as heated air should be supported. For example,
solid lubricant additive hard coatings as PMS 212 and its modifications (e.g., Triboglide),
Teflon impregnated Cr, ceramics as aluminum oxide, zirconia, silicon nitride or carbide sleeves
in addition to hardened and bare metallic shafts.

4.3 New brush bristle materials and applications program, both metallic and ceramics, should be
initiated.

4.4 For direct bristle shaft contact, watch the shaft interface for metallurgical changes and
cracking propagation. With highly loaded surfaces bare rubbing may induce microcrack defects
that lead to shaft failures (aero engines), while such surface defects may be of little consequence
for very large low speed rotors (aeroderivative engines).

4.5 Large diameter brush seals (diameter > 30 inch) are being fabricated for both aero and
aeroderivative gas turbine engines. The use of brush configurations looks promising.

4.6 Finger seals afford low cost compliant sealing; hystersis problems require further analysis
and materials tribopairing problems are similar to the brush seal system.

5 . New data and dynamic testing results have led Childs et al. to reexamine the linear theory and
conclude it to be insufficient to predict the instabilities associated with honeycomb stators
interfacing with labyrinth or smooth rotors. These systems may require frequency dependent
solutions and a two model solution is being prepared by Childs et al. The current approach is to
identify transfer functions as in controls modeling.

Fx= D E x Where D = Kd ( S +a)/(S + b )

[Fy] [-E D l [ y l E = Ke/(S+b)

The INDSEAL code work ,per Shapiro’s analysis anticipates frequency dependent results but is
not adequate to handle flow details of the honeycomb type. Honeycomb types can only be
simulated in INDSEAL through average wall roughness; however, hole depth and effective cell
parameters are significant and honeycomb seals reduce the effective fluid bulk modulus (and
sonic speed) which can not be handled in INDSEAL. A coupled acoustic solution will be
required as conventional time dependent CFD (SCISEAL) for high frequencies becomes
inaccurate and expensive; coupled aeroacoustic-aeroelasticand shock effects will also enter.

6. New and used codes in various states of development are available. Some are available to the
US industrial community and others are available for an initial investment. These codes are
beginning to find utility in the prediction of cavity flows, seal leakages, and rotordynamics.
A partial listing of such codes is limited to those discussed by Athavale, Braun, Pelf rey,
SanAndres, and Shapiro at this workshop with more detail found herein. Some NASA sponsored
Conference Publications are also cited.

iv
INDSEAL Industrial design version of seals code (Face, narrow groove theory spiral
groove , shaft with Ng-Elrod Turbulence model, labyrinth (USAF version),
dynamics). Data and numerical comparisons are given in this and previous
workshops (Roberts, Keba, Childs, San Andres, Scharrer to cite a few )
- in general, flows and stiffnesses are usually comparable, while damping and
“cross”or “quadrature dyanmic stiffness” terms often disagree.
Available to US industry through NASA Lewis Research Center

SCISEAL Scientific 3-D time accurate version of seals code ( interactive multiply
connected power/secondary/seal flow shaft seal flow fields with conjugate
heat transfer ) 33-validation tests are available including multiple connected
cavities in UTRC SSME turbopump simulation and Allison T-56/50 1D, 4-
stage turbine. Available to US industry through NASA Lewis Research Center

HYDROFLEX Bulk-flow code for evaluation of static and rotordynamic force characteristics
HYDROTRAN of laminar or turbulent flow hydrostatickydrodynamic bearings, e..g, damper
seals, and journal ,externally pressurized pocket, tilting pad, simple foil
bearings. Transient response of point mass rotor supported on rigid surface
bearings. Extensively validated with available experimental data.
Available through Texas A&M University Technology Licensing Office.

HYDROB3D Pratt-Whitney code for 3-D fluid film bearing hydrodynamics ( based on work
of Braun/Dzodzo (B&C Engr. Akron Ohio))

FLOWCONl Brush seals 2-D laminar CFD modeling including full flow field; also for
pin-fin flow devices and equivalent porous media flows. Fully validated
robust code ( University of Akron)

FLOWCON2 Power stream/secondary/seal cavity flows axisymmetric laminar CFD

modeling (University of Akron)

RSR - RSR Software library for seals and bearings including examples

PUBLICATION Seals Flow Code Development annual NASA Conference Publications date
fiom 1991.

Rotordynamic Instability Problems in High-Performance Turbomachinery are

available as biannual NASA Conference Publications dating fiom 1980.

Rotordynamic coefficient data sets available for labyrinth, honeycomb,

smooth (tapered, stepped, constant clearances), and helically grooved seals.
(Prof. Childs, Texas A&M Turbomachine Labs)

V
7. New emphasis is required on oiVair sealing flows, as customer driven sealing requirements
(Hendricks, Seals Code Development 1993). Oil vapors and coking smells are obnoxious at best
and health hazards at worst to the customer.

7.1 Ullah, Allied Signal, is instituting a consortium to enhance life of oil seals by 10 and reduce
the leakage of oil smoke into the cabin. Seals have high temperatures, incorrect tracking, and
coking. The effects of wear and sudden rise in seal temperature for a throttle chop are under
investigation. Several programs have end of 95 deadlines.

7.2 Inputs from the two-phase programs of Hughes et al. (Carnegie-Mellon retired), Hsu et al.
(NASA retired)two-phase and mist flows, Allison cooling work (Dr. Paul), Yasuna (Camegie-
Mellon) transient flows, Zimmerman (MTU-Siemans) two-phase flows in oil heat exchangers,
Marek fuels coking, Meyer and Lime (NASA) heat transfer, Glahn et.al. (Universitat Karlsruhe)
and others should be used as a nucleus for the program. Seals programs at Allison and Purdue
could also contribute to lip sealing.

8. Fluid film devices are receiving much attention in turbomachines.

8.1 Some bearings, are integrated into the IHPRPT program at Phillips AFB. Fluid film
bearings offer compact, low part count turbomachines.

8.2 Alternately, Si3N4 ceramic ball bearing tests are looking for 6.3M DN and the Dimofte
wave bearing is shown to have high resistance to whirl and operates well within the clearance
when whirl does occur.

8.3. Hydrostatic pocket film characterization requires a fully 3-D simulation. The effects are
pronounced in shallow and deep pocket flows as within the restrictor. In some cases, multiple
fluid passes from the restrictor to the pocket edge and back are required before the fluid exits the
pockethestrictor. The vonPragneau type sealhearing afford good stability and is often coupled
with the hydrostatic pockets.

8.4 Film riding seals as the aspirating seal, leaf seal, film riding brush and potentially the finger
seal, and proposed rim seals take advantage of the “lubricant film”, however very small
clearances (e.g., 0.4 mil) with critical flatness in harsh environments affords design challenges.

9. Turbulence measurements for code validation at 25 and 50 percent eccentricities will continue
under Morrison at Texas A&M with consortium sponsorship.

10. The What’s coming section includes structural-thennohydrodynamic coupling, predictive

maintenance/monitoring ,flow balancing throughout the engine, active/passive seals for rotor
stabilization (fan stabilization), strict environmental restraints, increased aeroderivative
enhancements, emphasis on film riding sealing, large diameter seal research, counter rotating
systems,and acoustic coupling.

Vi
10.1. SUGGESTED GOAL: Clean Sheet Approach to Engine Design ;specifically reduce the
number of compressor and turbine stages, increase reliability/life, and improve dynamic
response. Studies to date assume cooling and power stream requirements are known and
satisfied. Interactively this is an incompatable assumption; the clean sheet approach is required.

10.2 WHAT’S NEEDED (Under section 10.2, the percentages are given for aero and
aeroderivative gas turbines)
( Greater percentages are anticipated for space propulsion
turbomachines)

10.2.1. RIM SEALS ( order of 2 percent payoff)

Rim seals appear in both the compressor and the turbine. They afford the highest payoff,
the biggest challenge and provide the greatest opportunity for a new engine. The subtle warning
we are providing is that proper engineering of the seals will lead the manufacturers to a clean
sheet engine that will indeed be revolutionary as blades,stators, drums, cavities, radial and axial
dimensions are not constrained by conventional means.

To properly deal with rim seals will involve time unsteady (or suitable effective stress
parameters)interactive multicavity power/secondary and seal flow characteristics:

A. Start with the inlet flow parameters and determine the characteristics of the compressor,
the combustor, and the turbine with the exhaust parameters all satisfying the environmental
constraints.
B. This sounds a lot like NPSS (numerical propulsion simulation system), and it is to some
extent.
C. At this point we propose using all that’s available, including the interactive multicavity
power/secondary/seal flow code SCISEAL- CFDACE and -ADPAC (Allison).
D. New unstructured grid coding will enhance the solutions in the Euler form, and these
serve as first order solutions for the Navier Stokes forms plus conservation equations (energy,
mass,continuity, strain).
E. The new features of these codes will be the interaction with the strain codes to produce
the displacements associated with rotational and thermal management effects (nonunifoxmly
heated components with aerodynamic and centrifbgal loads).
F. The dynamics of such machines and their response to the seals have - to this point- not
been integrated into the engine analyses; however in our approach, these seals become a major
source of engine stability.
G. Controlled vortex generation would permit different platform flows and thermal
management scenarios. The basis remains in the lid-driven flow studies(Athavale et.al.
AIAA93-0390) and the success of the swirl brake in controlling unwanted flow fluctuations
(Childs and Ramsey: NASACP3 122). With some insight, such “tire-tracking” of the seal tips
and bladelvane curvature could be used to control flow. The seal tips through blade tip
treatments and shroud cellular materials (Bill, Wolak, Wisander, NASA TP 1835 (1981);
Tolokan, Jarrabet, Howe, ISROMAC (1992), p.571) and statodrotor curvature (Huber, Rowey,

vii
Ni; AIM-85-1216; Dring, Sprout, Weingold ASME95-GT-190) could be used to control stall
and enhance efficiency.

10.2.2. TIP SEALS (order of payoff 1-2 percent, depends on analyst)

The same arguments are given for tip seals; depends how the air losses are accounted for by
the engine analysis, but these seals provide the opportunity to reconfigure the shroud and case in
the same manner as the RIM SEAL. A potential key factor is the unsteady interaction between
the vortex shedding of the tip and hub regions with the subplatform cavities. When margins are
close to instability or singular points small perturbations can affect large changes in the entire
system. (e.g., fluid amplifiers). (e.g., see also Yamamoto et.al. ASME93-GT-404; Weingold et
al. ASME95-GT-380: Dring et al. ASME95-GT-190).

10.2.3. CDP SEALS ( order of 1 percent payoff)

The most expensive air in the engine is immediately downstream of the compressor. Several
engine companies are looking to “stopper the leakage” However, we know through analysis and
TEST RESULTS (NASA TM 106360/ ARL-MR-232) that the major benefits are not the
parasitic air recovery but the enhancement of the pressure ratio across the engine.

10.2.4. OIL/AIR SEALS (CUSTOMER DRIVEN REQUIREMENT - HIGHEST PAYOFF)

Cabin borne smells of oil and fuel are unacceptable to the paying customer and usually occur
after the engine is in service costing 1000 to 10000 times the initial investment to correct the
problem in the design phase. Air oil seals involve tracking the particle laden fuel, and oil as it
flows through the engine and egresses through the exhaust. Clean compressor air is affected by
leakages and subsequent ingestion of the air into the cabin either through these ducts or as
“sucked” in from the exhaust sources.

11. Emphasis is shifting toward supporting of both the engine and component industries. NASA
and these workshops will attempt to accommodate industrial requests and expand the base to
include manufacturing. The latter will require non-disclosure agreements among the
participants.

12. Future meetings:

Sept.-Oct.
NASA Interactive Seals Secondary Power Stream Workshops will be
limited representation to OEM’s and their designees.

June
Joint Propulsion Conference
Seal/SecondaryPower Stream Flows Sessions will be public forum.

...
v111
Liang
Seals code development overview and dissemination of INDSEAL and SCISEAL, codes
Hendricks
Rotordynamic coefficient data sets available for labyrinth, honeycomb, smooth (tapered, stepped,
constant clearances), and helically grooved seals.

SCISEAL successfully extended to compute multiple connected cavity flows with seals
interacting with the power stream with validation using UTRC SSME HPFP simulated
configuration results and turbine rig data.

Rim seal flow ingresdegress for small static displacements simulating and eccentric rotor
showed increased seal-neck heating for the forward facing step and diffuse heating for the
backfacing step.

Numerical simulation of the pressure in a synchronous whirling seal (Morrison water rig data)
indicated similar axialcircumferential mapping at fixed radial position; however, the magnitudes
are about 213 that of the experiment. Local radial-circumferential contours show qualitative
agreement with higher experimental shear.

Sic fiber testing on bare and coated rotors show decreases in friction with speed. Haynes 25
alloy fiber testing shows decreasing wear at elevated temperatures.

New sealing technology including piezoelectric face seals, compliant metallic leaf, and finger
seals are discussed.

Steinetz
AST Goals: reduce DOC+I up to 5%
SCF up to 10%
NOx over 50%
Noise by 7dB
2.5% can be gained by seals
Seals / Secondary Air Delivery goals:
Evaluate low leakage seals and cavity design requirements
Demonstrate sedsecondary air flow in engines
Validated time resolvedaverage CFD

Aspirating seal, rim seal development, with current research proposals under evaluation

ix
Cooperative programs with Allied Signal, Cross, Technetics, Williams, USAF, US Army

T-700 engine test data demonstrated over 1% reduction in SFC replacing BOM CDP labyrinth
seal with a dual brush seal. These tests clearly demonstrated that the compressor pressure ratio
increased thereby changing flows throughout the entire engine.

New SBIR starts for brush CFD analysis and a floating brush seal was initiated. Shape ;memory
alloy clearance control testing moves into phase 11. Compliant high temperature seals to over
1500F continue with evaluation of seal durability..

Proctor
A 2" diameter Haynes-25 standard construction brush seals were tested up to 4.3 hours in
cryogens nitrogen and hydrogen at surface speeds of approx. 300 and 540 f p s . Flow data for a
single brush were 0.5 to 0.3 that of a 12-tooth labyrinth for pressure drops to 175 psi.
Teflon impregnated chromium and chromium carbide coatings are preferable to zirconium oxide
and uncoated inconel 7 18.

Addy
Data for Sic brush and a finger seal are compared to 4-tooth labyrinth seal illustrating some
hystersis of the finger seals with enhanced performance over that of the brush and labyrinth.

Mayhew
IHPTET initiative to double turbine engine capability by 2000; thrustlweight over 100% and
compressor exit temperature by +400F. Secondary flow impact on the engine HPC: 4.4%
efficiency and HPT 4.2 % eficiency. Sealing requirements 1400 to 1650 f p s and over 1600 F
Brush seal development Haynes 25 and 214 with 130 psi pressure drop for single seal. Coatings
as A1203 and CrC (triboglide) have been evaluated. High speed testing at P&W to 1500F and
1650fjps. Toward a ceramic brush exceeding the 1600F goal. Hybrid and all ceramic seals are
being investigated. A user friendly secondary gas path analysis code is being developed.
Brush seals are in the Fl19 and F100-PW-229 engines. IHPTET goals can not be met without
the seals/secondary air programs.

Lowenthal
Sealol brush seals with backplate relief (low hystersis design) show lower leakage over time and
respond more favorably to excursions than conventional Sealol designs. Testing includes speeds
to 1080 fps, temperatures to 1200F and pressures to 130 psid and seal diameters to 9 inches (not
all applicable to the same test). Projected for NTC-76 demonstrator engine.

Bagepealli
Aspirating seal geometries and loadings from seals codes are reported. The primary components
of the aspirating seal are a labyrinth seal axial seal combined with a cavity fed inner face seal and
a hydrostatically fed outer face seal. Windmilling restarts (6) and ten other operational envelope
characteristics are cited. Film and rotor growth parameters are given.

Shapiro

X
INDSEAL: GCYLT, I C n , (cylindrical seals gas turbulent and incompressible laminar)
GFACE,IFACE (face seals, gas and incompressible)
SPIRALG,SPIRALI (spiral groove seals, gas and incompressible)
DYSEAL (dynamic seal analyses)
KTK (labyrinth seal, from USAF)
KBS (knowledge based system)
GCYLT : uses G-factors for Poiseuille and Couette turbulence coefficients. Ng-Elrod models
SPRTALI : is updated to include turbulence and inertia, but maintains the narrow groove theory
KTK : labyrinth seal code handles straight or stepped seals
DYSEAL: provides dynamics for the seal geometry
KBS: knowledge based system, which remains to be expanded into reality

Roberts
Comparison between Muijderman spiral groove bearing (Philips Tech. Labs, 1966) for a 324 mm
x 388 mm seal operating at 320 d s ( l 0 0 0 fps), 700C(1300F) and pressure drop of 1350
kPa(200psi) and SPIRALG show significant departures. Below 10 microns (0.04 mils) one
would suspect the results of either code and above 100 microns (0.4 mils) some transition to
turbulence is expected. No comparison to data are cited. In any case the thermoelastic effects
would have significant influence on the results.

Keba
Comparisons of LD=1 hydrostatic bearing run at Texas A&M and a generic L/D=0.37
hydrostatic bearing run at Rocketdyne with code predictions are provided. In general the flow
rates and direct stiffness as predicted by HYDROPAD and ICYL code are similar; however
significant differences appear in cross-coupled stiffness and damping predictions. These results
illustrate that the codes are not accurately modeling the physics.

Athavale, Ho, Przekwas

SCISEAL capabilities for cylindrical seals and multiply connected cavity flows interacting with
the power stream are delineated. Colocated grids, high order differencing, turbulence models,
rough walls. Multi domain solution methodology. Efficient solutions for entrance regions and
seal clearances, stepped and straight labyrinth seals, rim seals, face seals, conjugate heat transfer,
passive scalar transport (mass transport) 2D/3D treatments. Rotor loads, torques, and
rotordynamic coefficient calculations; full CFD solutions for centered orbits and small
perturbation methods for eccentric seals. The validation effort includes some 33 validation
cases.
Results are presented for:
Synchronous whirling water seal data of Morrison (Texas A&M, 1992 and 1995)
Illustrates good qualitative comparison between measured and calculated results for average
velocities. Note that these methods do not calculate the Reynolds stress tensor; however the
Reynolds stress tensor is a measured result of Morrison. A more fundamental approach is
required in the computations.

Rims seal gas ingestion (Graber et al UTRC 1987)

The comparison with data are quite good. These are averaged quantities and smear the local

xi
 details - like any good integration scheme.

Allison engine turbine cavity data (Munson and Forry of Allison)

Comparison with two different seal clearances illustrates how gas may be ingested into the disc
cavities and how non uniform heating of the discs can occur. While the simulation in static the
inference of dynamics can be made, Le., rotor perturbations

Large scale rig UTRCMSFC SSME HPFTP (Daniels and Johnson, 1993)
These results illustrate that the cavities are connected and the analysis must support multiple
connected cavities that interact with the power streams if the correct flow and mass distributions
are to be computed.

San Andres
Description of HYDROFLEX code to cdculate flows in bearings and seals. Considers laminar,
transition and turbulent flows with surface roughness and variable thermophysical properties. 32
force impedance coeficients for stiffness and damping The code handles cylindrical seals and
bearings with hydrostatic or hydrodynamic forces. HYDROTRAN provides force response to
transient loads. Fixes to improve dynamic stability in hydrodynamic journal bearings:
Pneumatic Hammer ( limit recess axial length and reduce recess volume; use end seal restrictors
or wear rings; change type of fluid inlet restriction, i.e., inherent compensation)
Hydrodynamic Instability: (use rough bearing surface; fluid injection against shaft rotation;
introduce bearing asymmetry; use flexible pad bearing geometry, i.e., flexure pad bearings)
Reduced load and capacity and direct stifiess are to be traded off for dynamic stability. Flexure
pivot tilting pad hydrostatic bearing has no restriction for hydrodynamic stability.
HYDROFLEX has been extensively validated with available experimental data for
' water/oil/liquid hydrogedliquid nitrogen hydrostatic bearings, watedliquid nitrogen damper
seals, oil and air journal bearings and tilting pad bearings, oil flexure-pivot tilting pad bearings,
and other applications.

Pelfiey
Analytical models of Reddecliff and Vohr were Reynolds equation finite difference solutions
which accounted for turbulence in hydrostatic bearings but has some convergence problems at
high eccentricity and clearances. HYDROB could account for grooves and is a finite element
code with better turbulence and inertia modeling and has no convergence problems. At present
HYDROFLEX and HYDROTRAN provide a 2-D bulk flow model finite difference solution
with transient and compliant bearing pad capability. RSR Software library includes a 2-D
incompressible bulk flow model for thrust bearings; however HYDROB is used for compressible
thrust bearings. MTI INDSEAL codes are cited. HYDROB3D (Braun/Dzodzo) full 3-D Navier
Stokes finite difference finite volume accounts for turbulence and inertia and does not assume
uniform pocket pressure. Pocket flows are highly 3-dimensional and code validation comparisons
with experimental data are excellent. Further testing of components will include LH2 and LOX
at Phillips and MSFC as well as ARPA TRP under a 4 yr $15M program with N A S M S F C .
The vonPragneau damper bearing geometries are to be tested as a part of the IHPRPT program.

xii
It is important to note the accomplishments of the PW ATD turbopumps; building on
technological advancements the pump has hours (vs minutes) of operational time - a significant
achievement.

Parallel processing using up to 500 engineering worksations, equivalent to 50 Cray XMP’s ,can
provide solutions overnight.

Palazzolo, Venkataraman, Padavala

Ryan, Vallely, Funston
A code is presented for liquid annular seals as a joint effort between Texas A&M and NASA-
MSFC. Details are not provided; however the bulk flow model is similar to that of San Andres.
References to INDSEAL and SCISEAL are absent as are comparisons. Comparisons with the
data of Iwatsubo indicates good agreement for fixed and flexible surfaces. The latter appears
difficult to assess.

Ullah
Air/oil seals that result in oil-in-cabin or smoke-in-cabin odors are a major cause of customer
complaints and warranty costs at Allied Signal. A consortium of various Allied Signal business
units and research center along with Arizona State Univ.and Georgia Tech.has been established.
Their current goals are to increase seal life by at least 10 times and understand the fundamentals
such as coking that is believed to be a primary contributor to the seal leakage problem. Current
problems being addressed include: Improving the thermal management of seals by improved
modeling/analysis methods; obtaining heat transfer data for active cooling schemes (jets,
proprietary film cooling etc); prediction of seal face tracking dynamics; understanding oil coking
mechanisms; and quantifying coking rates. Steady state thermal management models are being
studied first with 1995 year end deadlines; coking research is expected to continue through 1996;
and ,transient problems, such as sudden rise in seal temperatures during a throttle chop, will be
addressed thereafter.

Munson, Stienetz
Results of a comparative study of engine internal airflow systems for regional (AE-3007)and
turboshaft (LHTEC T801) illustrate 50% of leakage through 2 turbine blade/vane gaps; Sealing
effort at few locations yield large retums (regional or turboshaft). Development of a film riding
rim seal could reduce SFC by 1.8% in the regional engine and 5% in the turboshaft engines. The
potential for a clean sheet approach to engine design can provide even larger gains. Better
sealing is an inexpensive way to boost engine performance.

Dimofte, Hendricks
Unloaded journal bearing with a wave can operate over a range of speed free of subsynchronous
whirl. When whirl occurs the equilibrium radius is less than the bearing clearance and the
bearing runs stably. At large clearances the two wave exhibits whirl frequencies near 1/4
synchronous while the three wave whirl frequencies are close to % synchronous. At small
clearances, both exhibit half synchronous whirl at increased whirl thresholds.

...
XI11
 TABLE OF CONTENTS
SEALS CODE DEVELOPMENT WORKSHOP INTRODUCTION 7

Anita D. Liang, NASA Lewis Research Center ....................................... 1-C ) ' ,

TURBOMACHINE SEALING
Robert C. Hendricks, NASA Lewis Research Center ................................... 5 -/
AEROSPACE SEAL REQUIREMENTS
Bruce M. Steinetz and Robert C. Hendricks, NASA Lewis Research Center ................ 41'
CRYOGENIC BRUSH SEAL TEST RESULTS
Margaret P. Proctor and James F. Walker, NASA Lewis Research Center ................. 55 -9
RESULTS OF CONTACTING SEAL TESTING AT NASA LEWIS RESEARCH CENTER
Gene Addy, NASA Lewis Research Center ......................................... 67 Rq

AIR FORCE SEAL PROGRAMS /

Ellen R. Mayhew, Wright Laboratory, Wright Patterson Air Force Base ...................73
PRDA-I1 AND I11 BRUSH SEAL DEVELOPMENT PROGRAMS AT EG&G
Robert G. Loewenthal, EG&G Mechanical Components Technology Group Research
and Development ............................................................. 81 -6
ASPIRATING SEAL DEVELOPMENT: ANALYTICAL MODELING AND SEAL
TEST RIG
Bharat Bagepalli, General Electric Company ........................................ 89 7
INDUSTRIAL COMPUTER CODES
Wilbur Shapiro, Wilbur Shapiro & Associates, Inc. .................................. 115 3
COMPARISON OF SPIRALG PERFORMANCE PREDICTIONS WITH A SPREAD
SHEET SOLUTION
Martyn G. Roberts, Atlanta Engineering, Rolls-Royce, Inc. ............................ 139 -7
COMPARISON OF CODE PREDICTIONS TO TEST MEASUREMENTS FOR
TWO ORIFICE COMPENSATED HYDROSTATIC BEARINGS AT HIGH
REYNOLDS NUMBERS
John E.Keba, Rocketdyne Division, Rockwell Corporation ........................... 145 -/ 0

SCISEAL: A CFD CODE FOR ANALYSIS OF FLUID DYNAMIC FORCES

IN SEALS
M.M. Athavale, Y.-H. Ho, and A.J. Przekwas, CFD Research Corporation ............... 159 -11
THERMOHYDRODYNAMIC ANALYSIS OF CRYOGENIC LIQUID TURBULENT
FLOW FLUID FILM BEARINGS
Luis San Andres, Mechanical Engineering Department, Texas A&M University ........... 191 -I&

THE ACOUSTIC INFLUENCE OF CELL DEPTH ON THE ROTORDYNAMIC

CHARACTERISTICS OF SMOOTH-ROTOR/HONEYCOMB-STATOR ANNULAR
GAS SEALS
Dara W. Childs, Mechanical Engineering Department, Texas A&M University ............ 209 -/3

xv
WALL SHEAR STRESS,WALL PRESSURE AND NEAR WALL VELOCITY FIELD
RELATIONSHIPS IN A WHIRLING ANNULAR SEAL
Gerald L. Momson, Robert B. Winslow, and H. Davis Thames 111, Turbomachinery
Laboratory, Texas A&M University .............................................. 211

P&W CRYOGENIC FLUID-FILM BEARING AND SEAL TECHNOLOGY DEVELOPMENT

AND IMPLEMENTATION I
Philip C. Pelfrey, Pratt & Whimey ............................................... 223 15

INTERSTAGE DISK-CAVITYBRUSH SEAL NUMERICAL FLOW VISUALIZATION

STUDY
V.V. Kudriavtsev and M.J. Braun, Department of Mechanical Engineering, The University
of Akron ................................................................... 237 1
6
TWO-DIMENSIONAL DYNAMIC SIMULATION OF A CONTINUOUS FOIL BEARING
M.J. Braun,F.K. Choy, M. Dzodzo, and J. Hsu, Department of Mechanical
Engineering, The University of Akron ............................................ 267 1 '7
THREE DIMENSIONAL FLOW AND PRESSURE PATTERNS IN A HYDROSTATIC
JOURNAL BEARING
M.J. Braun and M.B. Dzodzo, Department of Mechanical Engineering, The University
ofAkron ................................................................... 281 / 9
!

THREE DIMENSIONAL FLOW AND PRESSURE PATTERNS IN A SINGLE POCKET

OF A HYDROSTATIC JOURNAL BEARING
M.J. Braun and M.B. Dzodzo, Department of Mechanical Engineering, The University
ofAkron ................................................................... 285 i -1
LIQUID ANNULAR SEAL RESEARCH
Alan B. Palazzolo, Balaji Venkataraman, and Sathya S. Padavala, Department of
Mechanical Engineering,Texas A&M University; and Steve Ryan, Pat Vallely, and
Kerry Funston, NASA George C. Marshall Space Flight Center ........................ 299 9 0

SEALS RESEARCH AT ALLIEDSIGNAL

M. Rifat Ullah,AlliedSignal Engines, AlliedSignal, Inc. .............................. 315 -21

INNOVATIVE SEALING TECHNOLOGY FOR NEXT GENERATION SUBSONIC

ENGINES
John Munson, Allison Engine Company, and Bruce Steinetz, NASA Lewis
ResearchCenter .............................................................. 327 2

FRACTIONAL WHIRL MOTION IN WAVE JOURNAL BEARINGS

Florin Dimofte and Robert C. Hendricks, NASA Lewis Research Center ................. 337 -23
COLORPLATES .................................................................. 341

ATTENDANCELIST .............................................................. 353

 SEALS CODE DEVELOPMENT WORKSHOP INTRODUCTION

Anita D. Liang
NASA Lewis Research Center
Cleveland, Ohio

SEALS CODE DEVELOPMENT WORKSHOP

PRESENT WORK PERFORMED UNDER NASA CONTRACT NAS3-25644,

"NUMERICAIJANALYTICALEXPERIMENTALSTUDY OF FLUID
DYNAMIC FORCES IN SEALS"

0 PROVIDE A FORUM FOR INFORMATION EXCHANGE BETWEEN THE

GOVERNMENT AND INDUSTRY ON SEALBEARING TECHNOLOGY
DEVELOPMENT

1
 0 CONFERENCE PROCEEDINGS TO BE DISTRIBUTED I N
SEPTEMBER, 1995

-
CODES TO BE DISSEMINATED IN JUNE JULY, 1995

OBJECTIVE DEVELOP CODES FOR ANALYZING AND DESIGNING

OPTIMIZED ADVANCED SEALS FOR FUTURE AEROSPACE AND
ADVANCED ROCKET ENGINE SYSTEMS

ORIGINAL 0 SEVEN YEAR EFFORT

SCOPE 0 DEVELOP 3-D CFD SCIENTIFIC CODES
0 COMPILE AND GENERATE SETS OF VERIFIED 2-D
INDUSTRIAL CODES
0 GENERATE A KNOWLEDGE BASED SYSTEM TO BE
COUPLED TO THE CODES
0 TECHNOLOGY TRANSFER V I A WORKSHOPS
CODE VALIDATION VIA PUBLISHED DATA, IN-HOUSE TEST
WORK AND COOPERATIVE PROGRAMS
0 PROGRESS MONITORED BY A PEER REVIEW PANEL

2
 SE-OUT OF ItXihi CO-CT NA83-25644

0 CONTRACT RENEGOTIATED IN N 9 4 AS A FIXED PRICE CONTRACT

0 CONTRACT CLOSE-OUT I N FY95

0 FINAL DELIVERABLES FROM MECHANICAL TECHNOLOGY, INC.

INDUSTRIAL CODES (INDSEAL) - ICYL, GCYLT, IFACE, GFACE, SPIRALG,

SPIRALI, DYSEAL, AND KTK

SCIENTIFIC CODE (SCISEAL) - MULTIDOMAIN CODE WITH

ROTORDYNAMIC CAPABILITY

-
EXECUTABLES TO BE DISSEMINATED I N JUNE JULY, 1995

0 SOURCE CODES TO BE DELIVERED BY THE END OF CY1996 TO COSMIC

OR EQUIVALENT

0 CODE DEVELOPMENT ACTIVITIES WILL BE COORDINATED WITH

OTHER PROGRAMS AND FUND SOURCES

S TO BE ADDRWWD BY THE Pl?&UEVTEW

INDUSTRIAL CODE OPERATING ENVIRONMENT

FUTURE WORKSHOPS

3
 2/47
TURBOMACHINE SEALING
Robert C. Hendncks
NASA Lewis Research Center
Cleveland, Ohio

Overview

Designers and customers are demanding higher performance turbomachine systems that have
long life between overhauls and satisfy the more restrictive environmental constraints. This
overview provides sources of design data, numerical, and experimental results along with
selected new seal configurations and static sealing challenges such as in the combustors.

The following categories are presented:

1. Seal Rotordynamic Data Base (experimental analytical program: Dr. Childs at Texas A&M)

2. Secondary Flow Interactions (validation studies with Dr. Athavale at CFDRC, Huntsville AL)

3. Contact Sealing (selected types with finger seal model)

4. Environmental Constraints ( emphasis on combustors)

5. What’s Coming

1. Seal Rotordynamic Data Base

At least 95-percent of the published data for dynamic coefficients of annular gas seals available
in the world today were generated under grant NAG3-181 with Dr. Dara Childs at Texas A&M
University. These data and the Workshop Proceedings will continue to serve as resource
materials and validation test results for analysis codes, future experimentation, and design.

Rotordynamic coefficient characterization of several different seal configurations including

labyrinth, honeycomb, brush, cylindrical, spiral grooved, and damper. Stiffness and damping
coefficients were assessed for centered whirling seals where the force input was horizontal and
the rotor suspended as a pendulum. Flow rates, pressure and accelerometer measurements were
taken to define the forces and leakages and displacements monitored by Bently probes. A
limited amount of eccentric rotor data (static eccentricity) results were taken.

5
For small motion about a centered position, the linear reaction-force model for gas path annular
seals becomes

and test data are available on the following seal configurations (figure 1, ref. 3):

(a) Tooth-on-stator labyrinths

(b) Tooth-on-rotor labyrinths, smooth stators
(c) Tooth-on-rotor labyrinths, honeycomb stator
(d) Interlocking seals
(e) Brush seals (4,5, and 6 stages)
(f) Smooth seals, constant clearances
(g) Smooth seals, tapered geometries
(h) Smooth seals, step configuration
(i) Honeycomb seals
Q) Helically-grooved stator/smooth-rotor seals
This program was the first to demonstrate the remarkable advantages of the honeycomb seals as
compared to see-through or interlocking labyrinth seals contrary to some early German test
results which suggested honeycomb should be avoided. However, because of the large damping
the long honeycomb seals are much better than the labyrinth seals, and leak less. These seals are
now gaining acceptance as replacements for labyrinth seals in centrifugal compressors and steam
turbines.

The brush seal results still remain the only available set of dynamic measurements and reduce the
cross-coupling to nearly zero. As a result brush seals generally improve the rotordynamics of gas
turbine engines.

See-through annular teeth-on-stator configurations are slightly more stable than teeth-on-rotor
configurations; for teeth-on-rotor and smooth and honeycomb stators are comparable.
Interlocking seals showed substantial reductions in direct damping.

Helically-grooved-stator/smooth-rotorseals showed that a negative cross coupling could be

achieved, which opposes forward whirl. Analytical treatments of the turbulent flows and
commercialization remains to be explored.

Swirl brakes showed that significant reductions in cross-coupling can be achieved; for the SSME
HPOTP the reduction could be a factor of two. These brakes have found commercial
applications in the electric power industry particularly where labyrinth seals are involved.

The testing of eccentric seals with smooth bores showed the seal to be more sensitive to

6
eccentricity than predicted and a continued effort is needed.

Swirl brake characteristics were explored and have been introduced to stabilize several
turbomachines in the field as well as new designs.

2. Secondary Flow Interactions

Fluid flows within an engine or turbomachine are coupled. The power stream and the secondary
flow streams that cool components, function as working or purge fluids are coupled through the
seal and cavity flow fields. The behavior of these flows becomes critical to the thermal
management of the component and ultimately to the performance of the turbomachine. Flow
studies in a turbine and within the UTRC simulated SSME HPFTP (figure 2, ref. 1) provided
validation studies of the SCISEAL code and the CFD-ACE codes,

As turbomachine power systems mature the ability to refine component efficiencies declines with
efforts focused on small percentage gains. In conventional turbomachine analyses interaction
between the power stream and the secondary flow paths such as beneath the blade platforms,
beyond the blade/vane tips, around the diffuser, and combustor sections are not coupled. For
large changes in efficiencies this approach is valid; however, for small changes the coupling
becomes of major significance. Current numerical and experimental work is focused on
determining the interaction between the power stream and multiply-connected multicavity sealed
secondary flow fields and conjugate thermomechanical response.

UTRC KPFTP Simulation

The results of a recent numerical study of the experimental work of Daniels and Johnson at
UTRC using C02 as the tracer gas (ref. 4) suggests the following:

1. Multiple cavity analyses capture interactive power/scondary flow stream effects that can not
be realized for uncoupled single cavity treatments; in short, uncoupled results are inadequate for
determining small changes in performance. For the UTRC simulation of the SSME HPFTP a
flow thread is defined that works its way through the upstream first stage rotor seal throughout
the dual rotor cavities and exits downstream of the second stage rotor seal.

2. Generally there is good agreement with the experimental results for gas ingestion and flow
egress, although the egress was lower than calculated. Tables I, 11,111, ref. 1.

3. Comparisons of the concentrations (C02) in the central cavity regions was good but only fair
agreement at the blade shanks. The shanks were simulated as rotating slots and such simulations
are known to be inadequate as the flow field is complex 3-D with spiral and angled jetting from

7
the holes. A second reason for disagreement could be in the scanning of the prints for griding.

4. The purge cooling thread (figure 3, ref. 1) tends to provide cooling for the front side of rotor 1
and aft side of rotor 2. The ingested gas effects the aft part of rotor 1 and the front side of rotor
2. Such nonuniform cooling and heating leads to nonuniform stress in the rotors.

5. These results illustrate that small changes in seals and cavity flows effect flows
throughout the entire engine or turbomachine. Further, these results are corroborated by
test results from the T-700(figure 4,ref. 11) where the compressor discharge seal was changed
out to a dual-brush seal. The engine performance showed a SFC decrease of at least one
percent (figure 5, ref. 11). The seal flows were reduced but more importantly, the compressor
discharge pressure was increased implying changes throughout the entire engine.

6 . With the strength of these numerical tools, the experimental data, and design expertise, it is
suggested that the entire secondary flow path and seal design be reevaluated. Those cooling and
purge gases that are of marginal importance can be eliminated along with the structural distress
associated with such holes. Further it is suggested that this ‘clean sheet’ approach be applied
to all the associated components with thought given to active controls for a more efficient long
life low maintenance engine.

Turbine Interstage Seal

A study of pressure and thermal effects in a labyrinth seal between stage 1-2 turbine disk cavities
provide the validation results. For orientation, the turbine vane is shown in white and the color
results represent the flow field computations. The geometry is that of a six-tooth-on-rotor -
labyrinth seal with a smooth stator. The clearances were fixed at 0.012 inch and 0.024 inch. The
analysis shows the following results:

1. When the labyrinth seal clearances are small ( 0.012 in.), turbine gas ingestion into the cavity
is limited and cavity temperatures appear uniform, especially in the seal area.(figure 6 ) (Blue to
blue-green color). However when the seal clearances are opened by operations or design to
0.024 inch, the powerstream flow is ingested into the cavity and appears to heat the labyrinth seal
as well. If the rotor is whirling in some manner the thermal distortions on the disk and the
labyrinth seal teeth will be nonuniform as there is a large thermal gradient across the first tooth.
Such nonunifoxmities can lead to durability problems.

2. The pressure drops reflect the increased leakage flows with the major gradient across the first
tooth (figure 7). The increased loading coupled with the thermal effects can increase the
durability problem.

Vortex Control

8
In a lid-driven flow blade simulation, the blade flow carryover initiated a vortex on the opposite
wall (pressure side) (figure 8, ref. 10). While these vortices are usually confined to the upper 10
to 20 percent of the blade they represent a considerable loss of performance. The concept of
introducing counter and co-rotating vortices could be used to control vortex losses, noise sources,
and stall margins. The vortices could be introduced by controlled jets or swirl vanes in the
platforms of the stators.

Unsteady Ingestion

Turbomachine rotation is eccentric to some degree, with whirl representing an extreme orbital
motion that can destroy a machine(figure 9, ref. lo). The fluid ingestion due to the rotor being
offset half of the seal clearance into the power stream is significant; also the fluid egress due to
the rotor being half a seal clearance below the power stream is large. While the coupled effects
were not studied, the inference is that unsteady thermomechanical distress can become acute
under dynamic loadings.

Acoustics

Not addressed due to the lack of data, but of major significance is the acoustic fields generated
and transmitted by the multiple cavities and their associated seals. A part of the long range seals
code development program, but this program funding has been stopped.

Synchronous Whriling Annular Seal - Experimental

The measurements of turbulence in a 0.5 eccentric synchronous whirling annular seal operating
at a Reynolds number of 24,000 and a Taylor number of 6,600 was measured using a 3-color
laser Doppler anemometer system. Flush mounted pressure and shear stress probes measured the
normal and shear stresses along the stator. Phased averaged pressure and shear along with mean
velocities and turbulence kinetic energy (at 0.16 clearance from the stator wall) were measured.
The results are complex and do not follow conventional seal models:

1. The mean pressure decreases rapidly at the seal inlet as the flow accelerates into the clearance
gap resulting in large shear stresses and flow fluctuations (figure 10, ref. 2).

2. The coherence levels are large (-8 -.9)for pressure and shear as rotor whirl dominates the
development of the flow field. Pressure is high on the pressure side of the rotor and low on the
suction side with shear stress being nearly 90-degrees out of phase.

3. Progressing through the seal, pressure and shear stress fluctuations decrease as turbulence
levels decrease with the axial and tangential velocities becoming more uniform (figure 11, ref. 2).

9
4. The mean shear stress angle steadily increases from 15-degrees at the inlet to 35-degrees at
the outlet with a sudden increase to 45-degrees outside of the seal within the ‘plenum’. The
circumferential velocity steadily increases with a sudden decrease in axial velocity at the exit.

5. At the exit the axial velocity is largest on the pressure side as compared to the suction side at
the inlet resulting in the phase averaged velocity becoming negative on the pressure side and
positive on the suction side at the sed exit.

6. The simple assumptions of no circumferential variation in pressure or shear stress and the
assumption that the axial velocity will be maximum within the maximum clearance are invalid.
Modeling of the flow field becomes a difficult task.

Synchronous Whriling Annular Seal - Numerical

The code SCISEAL was applied to the synchronous whirling annular seal data (see figures 10,
1I). Transforming the system to a rotating frame made the flow quasi steady. In this framework
the flow fields showed fair to good agreement with the experiment and generally consistent
picture of the flow physics. experimental turbulence is anisotropic, but the isotropic k-e model
was used in the numerical calculations. Before validation can be accepted, the low-Reynolds
model or multi-layer models should be applied to the data. Although M e r measurements are
desirable now that the codes are available, funding for the effort has been curtailed.

3. Contact Sealing

Brush seals

Brush seals suffer from high incipient wear, which reduces their effectiveness. Attempts to
identify rub-runner coatings and tribological pairing to reduce wear have not been to successful
and perhaps the better solution is to permit the bristles to rub an uncoated or sacrificial layer of
the shaft.

Ceramic brush developments, currently centered on Sic bristles, indicate that Sic is a good
cutting tool. Tribological pairing of Sic, A1203, and glass with various coatings provided some
initial screening of the fibers and Sic was the fiber of choice with tuft testing carried out at
surface speeds ( 500 <Vs< 1100 ft/s ) (figure 12, ref. 12). standard rub-runner coatings such as
CrC was stripped by the Sic fibers and other coatings were sought. One coating developed for
wear reduction was co-sprayed PSZ with boron-nitride added for dry high temperature
lubrication and more conventional coatings, plasma sprayed PSZ and vapor deposited PSZ,
alumina, and Triboglide (not available for testing) were also selected. Tuft testing of the SIC
running against coated rotors indicated that none of the coatings provided good wear
characteristics. friction coefficient was also high and in some cases decreased a little with
running speed (figure 13, ref. 12). Although the ceramics afford a hard surface, the operation on

10
bare shaft 17-4PH proved to be an equally good rub tolerant surface to the Sic.

In separate tuft testing at low surface speeds (< 80 Ws)showed a significant reduction in wear at
650C even though the fiiction coefficient was nominally the same (figure 14, ref. 5).
Four compliant seal concepts (finger seals, piezoelectric-controlled face seals, and shaped
memory-alloy shroud seals) while not in “use” have significant potential. In most cases, the
contacting seals experience rapid incipient wear.

Alternate contact seals

While a significant amount of effort has been given to brush seals and face seals ( hydrostatic,
hydrodynamic, aspirating, ring lift-off) development efforts continue on new types of seals
(figure 15, ref. 16 and figure 16, ref. 15, and refs. 6,9, 13, and 17).

Shaped-memory allovs are used for attachment pins and are being investigated for use in a
shroud seal. For example, a compressor ring could be designed so that when steady state
temperature is reached, the ring would contract to less than 1 percent of the diameter. Such a
design would pennit a relative expansion between a magnesium case and nickel alloy blades and
provide good blade tip sealing at cruise conditions. Response to in flight dynamics remains to be
evaluated.

Piezoelectric face seal elements can be designed to provide face coning. Coning controls both
seal stability and leakage. One such experimental test configuration had a carbon-faced,
piezoelectric, active-controlled elements placed on both sides of a rotor. purpose to the seal was
to reduce the consumption of onboard helium and still separate cryogenic oxygen and hot
hydrogen-rich steam (figure 17, ref. 10).

Leaf seals combine the noncontacting features of fluid film seals and the compliant nature of the
brush seal. leaf seal consists of several thin sheets or leaves that resemble sections of lip seals
wiping the shaft. A second set of overlapping leaves completes sealing and the leaves can be
straight or angle cut. leaves have mismatched curvature with respect to the shaft for lift off with
300-500 micro-inch at the leading edge to perhaps twice that at the trailing edge. leaf flexes in
the radial direction to accomodate thermomechanical behavior and axial and angular
misalignments. Under no rotation, there is a net closing force to keep the leaves in contact with
the shaft (figure 18, ref. 10).

Finger seals are of simple construction. seal consists of two or more axially stacked, radially
compliant, overlapping disks or diaphragms that resemble and are often referred to as “hockey
sticks”(figure 19, ref. 7). fingers which are separated by gaps can have straight or logarithmic
curvature and usually are of constant thickness. overlapping disks arrays form a compliant seal.
Finger seals can be fabricated by several techniques, e.g., photoetched or machined from a
variety of materials such as metal, plastic, or ceramics. assembly is held together by simple
riveting and the final configuration can be ground and lapped depending on the use. These seals
leak less than l/lOth of a conventional labyrinth, but current versions suffer from hystersis
problems.

11
An elementary leakage model was developed based on the brush seal bulk flow model and
provides guidance for parametric variations and design (figure 20, ref. 7)

4. Environmental Constraints

Most communities have adapted some form of environmental controls and requirement for
leakage control, emissions, and noise (e.g., see ref. 8). For example, the high speed research
program is dedicated to achieving about 5g of NOx per kg-fuel burned.

Gas turbine exhausts CO, C02, H20,unburned hydrocarbons (UHC), particulate matter mostly
in the form of carbon, NOx and excess 02, and N2. While C02 and H20 are commonplace they
contribute to the global warming and are reduced only by burning less hydrocarbons. In
industrial gas turbines sulfur and oxides are also a problem and requires removal prior to
combustion. Controlling the combustion temperature between 1680K and 1900K is considered
by some as the temperature range for low emissions (COG5 and NOx< 15 ppm) and represents
the underlying principles in staged, variable geometry, and premixed combustors.

International Civil Aviation Organization (ICAO) has set standards that involve the emission
index(g/kg-hel), the SFC@g fuel/hr-kN), and the time of operation (hr) for the landing takeoff
cycle (LTO) (see Table IV).

Emissions (g/kN) = E1 x SFC x TimeinLTOmode

For the commercial powerplants, emissions of 10-20 ppm NOx are being sought and can
possibly be met by consuming natural gas.

These complex combustors require static seals, and several types are proposed including the use
of brush and rope seals (figure 21, ref. 14). Loss of high emission laden gases or injection of fuel
or air at the improper locations will lead to excessive engine emissions.

5. What’s Coming

A. Conjugate problems with coupled structural interactions with different time scales as a
potential problem.

B. Monitoring components for longevity (life) and predictive maintenance

C. Flow balancing through the compressor/turbine multiple cavities and subplatform flows

D. Active/passive seals rotordynamics for stabilizing both the aero and aero-derivative engine

12
E. Strict environmental restraints on fluid leakages and engine emissions

F. Aeroderivative systems increases in efficiencies by controlled subplatform flows and perhaps

the use of separate cooling fluids

G. Increased shift toward face seals for special applications (high payoff) but labyrinths will not
go away for a long time.

H. Large diameter seals, segmented, brush, face, lift, and floating types.

I. Counter-rotating systems.

Selected References

1. Athavale, M.M., Przekwas, A.J., Hendricks, R.C., Steinetz, B.M. (1995)Numerical Analysis
of Intra-Cavity and Power-Stream Flow Interactions EMultiple Gas-Turbine Disk-Cavities. Int.
Gas Turbine and Aeroengine Congress and Expo., Houston TX, June 5-8,1995, Paper 95-GT-
325. NASA TM 106886.

2. Athavale, M.M., Hendricks, R.C., Steinetz, B.M. (1995) Numerical Simulation of Flow in a
Whirling Annular Seal and Comparison with Experiments. To be presented at the Joint
Propulsion Conference, SanDiego CA.

3. Childs, D. W. (1995) Summary of Seal Rotordynamic Data NASA NAG3- 131. Final Report
1995. Texas A&M Turbomachine Laboratories.

4. Daniels, W.A., and Johnson, B.V. (1993), Experimental Investigation of Turbine Disk Cavity
Aerodynamics and Heat Transfer, Contract NAS8-37462, UTRC Report 93-957878-27.

5. Fellenstein, James A.and DellaCorte, Christopher: (1995) A New Tribological Test for
Candidate Brush Seal Materials Evaluation. Socitey of Tribologists and Lubrication Engineers,
Chicago, IL, May 15-20,1995. DOE/NASA/50306-6; NASA TM-106753.

6. Glienicke, J., Launert, A., Schlums, H.,and Kohring, B. (1993) Non-Contacting Gas
Lubricated Face Seals for High p.v -Values. Seals Code Development-93, NASA CP10136,
p.367.

7. Hendricks, R.C., O’Halloran, B., Arora, G., Addy, H.E., Flowers, J.,Carlile, J., and Steinetz,
B.M. (1994): Advances in Contact Sealing. Advanced Earth to Orbit Propulsion Technology
Conference, Huntsville AL, May 16-19, 1994. US Army Research Laboratories.

8. Hendricks, R.C. (1 993) Enviromental and Customer-Driven Seal-Requirements. Seals Flow

Code Development-93. Hendricks, R.C. and Liang, A.D. Eds., NASA CP 10134, p.67.

9. Hendricks, R.C., Liang, A.D., Childs, D.W., and Proctor, M.P.( 1992) Development of
Advanced Seals for Space Propulsion Turbomachinery. 1992- Aerospace Atlantic Int. Conf.
13
And Expo., SAE, Dayton OH April 6-10,1992. NASA TM 105659.

10. Hendricks, R.C., Steinetz, B .M.,Athavale, B.M., Przekwas, A.J., Braun, M.J., Dozozo, M.I.,
Choy, F.K., Kudriavtsev, V.V., Mullen, R.L., and Von Pragenau, G.L. (1995) Interactive
Development of Seals, Bearings, and Secondary Flow Systems with the Power Stream.
to be published Interanational Journal of Rotating Machinery. Presented at the Fifth
International Symposium on Transport Phenomena and Dynamics of Rotating Machinery
(ISROMAC-5) Kaanapali, HI, May 8-1 1,1994.

11. Hendricks, R.C., Griffin, T.A., Kline, R.R., Csavina, K.R., Pancholi, A., and Sood, D.
(1994) Relative Performance Comparison Between Baseline Labyrinth and Dual-Brush
Compressor Discharge Seals in a T-700 Engine Test. Int. Gas Turbine and Aeroengine Congress
and Expo., The Hague Netherlands, June 13-16,1994. NASA TM 106360. ARL-MR-118.

12. Howe, Harold: Brush Seal Shaft Wear Resistant Coatings. Contract NAS3-27383, Final
Report March 1995.

13. Leefe, S., (1992) Dynamics of Face Seals for High Speed Turbomachinery. Seals Flow Code
Development-92, NASACP 10124, p 197.

14. Lefebvre, A.H. : (1995) The Role of Fuel Preparation in Low Emissions Combustion. Int.
Gas Turbine and Aeroengine Congress and Expo., Houston TX, June 5-8, 1995, Paper 95-GT-
465.

15. Munson, John (1993) Compressor Discharge Film Riding Face Seals. Seals Flow Code
Development-93 NASA CP 10136 ,p.2 19.

16. Pope, A.N. (199 1) Areas of Seal R&D at GE. Seals Flow Code Development-9 1.
Hendricks, R.C. and Liang, A. D. Eds., p 109.

17. Withers, P.A. (1992) High Pressure Compressor Delivery Brush Seal of the International
Aero Engines (IAE) V2500-A1 Gas Turbine Engine. Seals Flow Code Development-92, NASA
CP 10124. P 275.

14
 -
TAB1 E L-l" PARAMETERSAND NOTATION
rpm prrsmrcRymlds~ccnrcr Afr
psir aumba a* cavity any
plrge Fpgt
n-4u ==4,, F3-414
102 605s 1 . 6 4 ~ ~ 9o m 0.012 0.012
5724 219x14 0.017 0.008 0.013
57.79 2%& 0.030 0.015 0.014

4.126 -0.094 0265 02.08

(w-m 0.19 -023 0.12 -025

0.1168 I I
0-0552 0.0904

numaicll
(30K Pzid)
0.1168 I I
0-0552 0.0904 0.6374 I 0.639 I
expimental 4224 -0.154 0257 0271 0.115 0-057 0.087 0.637 I 0528
(elprmm)/exp 0.05 4.10 4.11 030 0.02 0-03 0.04 0 -021

numerical I -0223 I -0.1025 I 0357 I 0381 0.208 I 0.103 0.0985 0.7351 0.738
e m d 14.186 1-0.113 1 0302 I 0.302 0.208 I 0.105 0.095 0.707 0.604
(e%prmm)/cxp 020 0.09 0.18 026 I 0 0.01 0.04 0.04 0.22

Runnumber 102 202 205

experimental 4 . 0 1 8 4.109 4.103
numerical -0.005 -0.OOO6 4.0029

15
I -
Region XI centcravity mdrouxI I

C42-numcrical
1 1
0.232
24
0.150
0.03
-12
0.378
1903
0
I I 1
0.465 0273
32
0299
28
0.173
exPin=d I 2 2 -18 -03 31 22
0.194 0 0388 0223
-67 0 -13 -11
0.987 0 0.006 0.004
9 0 .05 .04
0.992 0 om4 om2
-71 -03 .1 -04
htacavity d rotor 11

* CXp

c11 -ll-
,

CXFCCimcntal
-
Region N Aft cavity mdmorII
0.006 0.012
-03

16
TABLE IV. ICAO Gaseous Emissions Standards

Emission Subsonic turbojet/turbofan Supersonic turbojetlturbofan

New, takeoffthrust >26.7kN

HC 19.6 140(0.92)**Pr,takeoff

co 118 4550/(Pr,takeoff)** 1.03

NOx 40 +2Pr,takeoff 36 +2.42Pr,takeoff

Pr = engine pressure ratio

17
-21
E>

I I

18
I -

@) Model seal
computational

19
Contour levels
3 -6.540E-02
6 -~.~soE-o~
9 -5.160E-02
12 4.470M2
15 4.780E-02
18 -3.090E-02
21 -2.400M2
24 -1.710E-02
27 -1.02OE-02
30 -3.300E-03
33 3.600E-03
36 1.050-2
39 1.740E-02

@) See color plate on page 342.

Figure 3 - ( a ) Streamline pattern in regions I, II and the connecting blade shank region. @) Streamline
pattern in regions 111, IV, connecting blade shank region and the slot in the stator support. (CfFlow
detail at seals 1, 2, and blade shank region t o illustrate the complex vortical structure and Mainpath
flow ingestion.

20
 Contour levels
2 -6.615E-02
4 -5.845E-02
6 -5.075E-02
a -4.350~~~2
10 -3.535E-02
12 -2.765502
14 -1.995-2
16 -1.225E-02
18 4.550E-03
20 3.150E-03 See color plate on page 343.

Figure 3 -Concluded. (c) Flow detail at seal 1,2, and blade shank region to illustrate the complex vortical
structure and main-path flow ingestion.

Mixture Fraction (Concentrations)

Computed Values of F2 Concentrations, Run No.202

See color plate on page 345.

Figure 3 Ingestion of main-path flow in multiple

gas-turbine disc cavities
UTRCMSFC Large Scale Rig
21
 See color plate on page 346.
Mass Fraction of F4

See color plate on page 347.

Figure3 Ingestion of main-path flow in multiple

gas-turbine disc cavities
UTRClMSFC Large Scale Rig

22
 r StationaryCOP seal

\
\
d
\
Rotating COP seal
'L

(a) Labyrhth seal package and airnow.

2247-iI
2237

0.015
0.005
"

Radius
4 places -,
.
360
3402
- ~
,-0.120
, ~- 44"
, 4 0.040
6

0.020 R
~YP.-,,
1--
~

.*?
16" all teeth
: except14'
Nofive
front

step
0.1 46 -U
'.
No step allowed 1
allowed

7- Angled teeth Straight teeth

2.4295 diam
@) Schematic of labyrinth compressor discharge seal. (Seal teeth and axis established by diameters A and B to be concentric within
0.003 full indicator reading. No steps allowed on tooth face or at fillet radius. AI1 dimensions are in inches.)
Figure 44abyrinth compressor discharge seal system.

23
 Brush 5 2
Brush seal pack airffow.

A
A
I
73-mm
(2.875-in.I
diam
1.4 m t
(0.055

(b) iliustratian of dual-brush camp

e4 essor discharge seal and schematic of aim

24
 1.5
0
0
0
2
0
E C
E.$
o n

a 0
I 1.0

--
80
cad,
-a
c-
i!
z Dual-brush sed-
0
n

100 200 300 400 500 600 700 800 900 lo00 1100
Experimental testbed engine horsepower
Figure 5 --Expenmental testbed engine specific fuel consumption as a function of horsepower.

25
Figure 6 Temperature contours

Stage 1-2 Disk Cavities

K
1.360+03
1.3e+03
1.24e+03
1.19c+03
1.13e+03
1,07e+03
1.02e+03
962
905
848
792

-
A. Labyrinth seal clearance 0.012 in. B. Labyrinth seal clearance = 0.024 in.
See color plate on page 348.

Figure 7 Stream Function

Stage 1-2 Disk Cavities

Pa Pa

-
A. Labyrinth Seal clearance 0.012 in. -
B. Labyrinth Seal clearance 0.024 in.
See color plate on page349.
26
 C X

...........

*=
4 x ~ = 8

rl;, ~y=13
............
..............
................ .-\ ............
.........
--e\

.............
................. ....... ---
.....
\

.................
.................
.................
* . . I
...---e

........----e

.................
..................---/
............
...........
...........
- c _

............ -e-/

11-111
<X mr13

...............
...............
.............
' : .............
\: .............
\.--.
\\\--.
.\\--.-
....
...........
...........
\.---.-..-e-

- .............
-.
C X mtl3

, (4 (r)
Figure 8
- - -
Velocity vector plots for a liddriven cavity simulating a turbine blade passage at axial plane ry 13 (near midplane) and radial
-
-
planc xz 8 (near midplane). (a) Flow geometry. (b) Base test case velocity vector plots at three selected axial planes: xy 4 (mar inlet). xy
-
22 (near exit and radial plarrs), 4 1 (near cavity boaom), xy I5 (near cavity top). (c) Inlet corouting vortex injection ne= pressure side of
simulated blade. (d) Inlet corotating vortex injection near suction side of simulated blade. (e) Inlet counterrotating VONX injection near pressure
side of simulated blade. ( f ) Inlet counterrotating vonex injccrion near suction side of simulated blade.

27
 s
6
7
a
9
10
11
K
l3

I I

3
+
S
7
e
9
:e
11
22

I
Figure -
--
i ~ contours near rim seal neck for simulated mtor-sutor mismatch operation where R R , + e. Base level
- 7200 andlisotkrm

-- -
~ ~ a n
C
-
m c
q (see fig. 37). (3) Eccentriciry e 0. strcunlines. 0)e 0 (isotherms). (c) e 4.5 (wd neck thickness I ) , smunlines. (d)
4.5 (sed neck &cknas I), isotherms. (e) e 0.5 (sed neck thickness I). smamlims. (0 e 0.5 (sed neck thichess I ) , isoth~rms.

28
 4 - 1 Qjrcctian of rotair m o t i o n ia absofuteframe

1.0

Axial
distan Ge
ZAI

0.0

0.994

A xial
distance
28%

-0.004
0@
. a.5 1.0
Fraction of time
See color plate on page 35 1.
b. Experimental tp15wlt5

d a t e d a n d expenbnental n on -dimension
or wall, ..
P*= PI. / ( C A P )

29
 (a) x/L= O.oq125 x/L = 0.0

0)
x / t = 0.2125 x/L = 0.22

Numerical Experimental
Figure 1 1 4ountouw of uxNm at various cross-sections along the seal axial length x. Seal whirl and spin in counter-
clockwise direction. Seal clearance exagerrated for clarity.

30
(e) x/L=O.9875 x/L = 0.99

Numerical Experimental

31
 MOIENI *an

I W I I O L D E R 1UBE

tor GAS EXIT DUCT

l\C PORI (2 PLACES)’

)KIT G * S MIXING CbUMBER

INSVCATION

1
I I
J

SECTION V I E W S OF FIBER / C O A T I N G T E S T R I G
Figure 12, Reference 12

32
 Sic on BN/PSZ
Cold Air17.5 scfml /Large Wt.

0.42

0.38

(Thousands1
Speed If t/aecl

Sic on Bare Rotor

Ambient Air /Large Wt.

0.42

0.41

0.4

0.39

0.38

0.37

0.36

0.35

0.34

0.33

0.32

0.31

IThourandsl
Speed (ft/secl

Figure 13

33
 l,OOORPM&lQ
E3 l,OOORPM&5og
R l~RPM&loog
l2,,OOORPM&5Og

-8Nsh Wear factor, in rnrn3M-m for H25 cobalt based superalloy brides sliding
aSainSt l718 nickel based supemRoyshaft. Wear factor goal for adequate wear tie is
1Os rnrn3m.m
Figure 14, Reference 5

34
 I Primary Seal Elements
2 Outer Race. Rotating
3 Fwd Race. Stationary
4 Aft Race. Stationary
5 Axiai Spnngs
6 Piston Ring. Stationary
7 Radial Bushings
g Fiston Ring. Rotating
9 Raoial Splines. C:am?-Race
IO Carrier. Sktionary P:s:icn fiir.5
t i Clamp Ring
12 Clamp Spring
13 Adapter. Seal lo Frame
1 4 Flange. Eilgine Frame
15 Racial SpIincs. Fwd-Aft Races
16 Reiaincr Lugs. CarScn f e p e n :

LARGE DIAMETER HYDROSTATIC SEAL

Figure 15, Reference 16

35
 ROTATI G

ASPIRATING GAS BEARING FACE SEAL

CONTINUOUS ROYG SEAL DESIGN

Figure 15 Concluded

36
Low Leakage of FRFS will provide substantial reduction in
cycle specific fuel.consumption:
y700-
..-
o
v)

-600-
a)
&
0
K
50'0 -
* Film Riding Face Seal
0
Brush Seal
G A
Labyrinth Seol
_-
o 400-
I
1
al

0 100 200 300 400 500 600

Differential Pressure (psid)

0 1.5% relative to Labyrinth seal system

0 0.5% relative to Projected multi-stage brush seal

Figure 16,Reference 15

37
 To controlsystem
m
From control system Fromcontrol system

Uquidoxygen side

Figure 17 -Piezoelectricface seal configuration (a) Helium purge seal (clearancegreatly exaggerated).@) Coned
defomableface assembly. Coning, 6 = ho- hl. (c) hce holder and deformable face assembly.

Intyp.

PS

Figure 18 -Compliant. metallic leaf seal concept

-Argfe-cut leaf seal configuration

38
 l-6 -ye'-
R

j,/'
1'

(a) Configuration (b) Finger geometry at the interface

(d) Finger model

Figure 20 d d e d t c d furga seal foopint model.

39
 iGNITORS(2)
I
CONVENTIONAL h W N ZONE FUEL INJECTORS
MOUNTING SYSTEM E NEW OUTER
DIFFUSER CASE WALL
20 PILOT FUEL
INJECYORS

CONVENT I 0‘NAL
FIRST UW E
I NTER FACE

EXISTING PRE-DIFFUSER AND

INNER OlFFUSER CASE WALL

Figure 21, Reference 14 Pratt and Whitney axially-staged combustor

* Floatwall is a tradename of a product manufactured by the Pratt & Whitney. Trade names or
manufacturers’ names are used in this report for identification only. This usage does not constitute an
official endorsement, either expressed or implied, by the National Aeronautics and Space Administration.

40
 AEROSPACE SEAL REQUIREMENTS

Bruce M. Steinetz and Robert C. Hendricks

NASA Lewis Research Center
Cleveland, Ohio

Cycle studies have shown the benefits of increasing engine pressure ratios and
cycle temperatures t o decrease engine weight and improve performance of
commercial turbine engines. NASA is working w i t h industry t o define
technology requi rements of advanced engines and engine technol ogy t o meet the
goals o f NASA’s Advanced Subsonic Technology Initiative, As engine operating
conditions become more severe and customers demand 1 ower operating costs.
NASA and engi ne manufacturers are i nvesti g a t i ng methods of improving engi ne
efficiency and reducing operating costs. A number of new technologies are
being examined t h a t will allow next generations engines t o operate a t higher
pressures and temperatures. Improvi ng seal performance - reduci ng 1eakage and
increasing service l i f e while operating under more demanding con itions - will
play an important role i n meeting overall program goals of reduc ng specific
fuel consumption and ultimately reducing direct operating costs. This paper
provi des an overvi ew of the Advanced Subsonic Techno1ogy Program goals ,
discusses the motivation for advanced seal development, and high ights seal
technol ogy requi rements t o meet future engine performance goals .

Outline

o ASTProgram
o Joint NASNAnny Program
o SBlR Seals Development
o Compliant Seal Development (In House)

41
 Pratt & Whitney-PW40~
GD-94-84752

* Direct operatin erest) by up to 5%

Engine fuel burn up to 10%

* Noise by7dB

ess ne and

42
 Transport Aircraft Seal Technology Benefits
2005 EIS

Seal Technology Study Engine co. System Level

Benefits
Balance Piston Aspirating Seals GESWAST Transport GE -1 29% SFC
(3ea)
m
CDPPreswirl Locations
--
Regional Allison
Adv. Tech. Brush Seals AST Regional Allison -1.67% SFC
Turbine Rim Locations (2 ea)
Film Riding Circumferential Seals AST Regional Allison -1.8% SFC
Turbine Rim Locations (2 ea)
Turbine Blade Tip AST Regional Allison -1.36% SFC
Active Clearance Control
Compressor Blade Tip T-55 Allied/ -.3% SFC
Passive Clearance Control Lycoming

Aspirating Seal Development

0bjective:
Investigate feasibility of large
(36") diameter aspirating seals for
turbine balance piston locations
and demonstrate in engine test.

Approach:
+ Perform sub-scale and full-scale seals tests to optimize seal characteristics
+ Demonstrate ability to safely follow once-per-rev and maneuver tilt rotor run-outs
+ Evaluate seal's resistance to potential dust-ingestion
+ Develop and validate analytical methods to perform seal sensitivity studies under
off-design conditions. . .
+ Test seal on engine test bed and evaluate acceptability for engine service.

Near Term Schedule:

+Evaluate performance of robust, low leakage, large diameter seal FY96

45
 UTRC Cavitymim Seal Experiment
0biectives:
Characterize turbine cavityhim seal environment
and develop an optimized design to minimize purge
requirements and mitigate hot gas ingestion.

Amroach: I
+ - Obtain time-avg./time-resolvedpressure and velocity distributions to examine rim
seal ingestion mechanisms for an AST high work turbine.
+ Develop physical models of ingestion processes
+ Develop improved/optimized design concept based on operating conditions
+ Determinefilm cooling benefits of purge on gas-path side of blade platform
+ Evaluate nature of purge flow entering main stream and identify potential effects
on turbine efficiency.
+ Document findings in usable format for anchoring advanced time-avgltime-
resolved 3-D CFD analyses techniques.

Schedule:
Complete fabricationhnstrumentationof airfoils and rim seal cavity model 2Q FY95
Complete baseline geometry experiments 1Q FY96
Complete parametric geometry experiments 4Q FY96
Complete improved geometry experiments 2Q FY97
Deliver complete data set 3Q FY97

NASA Research Announcement (NRA)

Elements:
+ Two steD (PRDA-like) Drocess:
Company submission of abstracts responding to NRA Topics of Interest
Company submission of full-proposals

+ Term: FY96 - R O O with possible large-scaletechnology demonstration WOO-01

Schedule:
June 2 Receipt of abstracts
June NASA Technical review of abstracts
July 7 Letter of encouragement for full proposalsand detailed costs.
August 23 Receipt of full proposals
end Oct. Final proposal evaluation complete
January Contracts awarded

46
 Joint NASNArmy Seal Program

Why Seals?

e reduction and
percentage increased thrust or dec

ies offer significant leakage reductions

I PROG SCHEDULE

47
 Brush Seals i
Conven

T-700 Test Bed

mpressor Discharge Brush Seals
1.0
ine)

100 300 500 700 900 1100

Horsepower

48
 SBlR Seal Development
Hybrid Floating Brush Seal Feasibility Study
NASA SBlR Phase I
B&C Engineering/Univers*Qof Akron

Objective:
Assess feasibility of unique hybrid floating brush seal to reduce brush seal wear by
floating brush seal in an air-bearing carrier.

Approach:
+ Perform desi n analysesto size hydrostatic air bearing to float brush seal carrier
+ Designbuild$ b r i d seal
+ Evaluate seal easibilii using University of Akron facilities

Schedule:
Complete design March, 95
Complete fabrication June, 95
Complete feasibility demonstration June, 95

Potentialcustomers:
+ Aircraft engine industry
+ Durametallic Seal Company
+ Power generation industry
+ Other

CFD Brush Seal Analysis

NASA SBlR Phase I
CFD Research Cop.

Objective:
Develop integratedthermaVnuidlstructuresbrush seal analysis methodology

Approach:
+ Adapt unstructured 3-D Navier Stokes code to model the fluid flow through
w s brush
+ Eplernent the conjugate heat transfer and frictional heating capabilities to
assess brush seal thermal response to brush-rotor friction and local flow
conditions.
+ Validate analysis tools using available experimental data.
+ Initiate the development of dedicated brush seal flow model that can be coupled
with global secondary-air system analysis procedures.

Schedule:
CFD code and automatic grid generation adapted to brush seal April, 95
Complete brush seal flowheat transfer initial simulations June, 95

customers:
Brush seal vendors
Aircraft engine and power generation industries.

49
 oy Compensator Ring For
Clearance Control
Tech ~ o l ~ g j e ~ ~ y ~ o r n i n g
ed Development
Prototype Compensator Ring Enaine Compressor Demonstration
Phase 1 Phase 2

50
 Compliant Seal Development
High Temperature Compliant Seal Development
In-House
0bjective:
Develop high temperature compliant seals for high temperature structural interfaces
to minimize leakage and permit relative component growth, minimizing thermal
stresses.

Approach:
+ Develop advanced compliant seal concepts for 21500 F service.
+ Evaluate performance characteristics
+ high temperature leakage and durability.
+ flexibility and compliance retention with cycling
+ manufacturability, scalability.
+ Develop design guides to predict leakage and seal compliance over operating
ternperature regime.
+ Provide technology to engine and industrial partners.

Goals:
+ Develop necessary facilities to evaluate advanced compliant seals at
temperatures to 1500 F.
+ Extend NASP-derived high temperature compliant seal technology to meet HSR,
IHPTET, and industrial-partner goals.

51
 e Seals

CD-94-70758
1500°F Seal Testing

52
 HIGH-TEMPERATURE FLOW AND DURABILITY RIG

CUT-AWAY VlRN I

Summary
NASA and industry partners are pursuing engine performance
advancements to reduce fuel bum, reduce operating costs, and ensure
industry competitiveness into the next century and beyond.

Advancements in seal technology will play an important role in achieving

AST performance goals.

Significant reductions in SFC are possible through implementing

advanced seal technology. Engine studies show that over 2.5%
reduction in SFC is possible applying advanced seals to a few locations.

Costs of developing advanced engine seals are a small fraction of

re-designing & re-qualifying complete compressor or turbine components
with comparable performance improvements.

53
 53-37
CRYOGENIC BRUSH SEAL TEST RESULTS

Margaret P. Proctor and James F. Walker

NASA Lewis Research Center
Cleveland, Ohio

Brush seals are compliant, contact seals t h a t have long-life, low-leakage

characteristics desireable f o r use in rocket engine turbopumps. 50.8-mm (2.0
inch) diameter brush seals w i t h a nominal i n i t i a l radial interference of
0.127-mm (0.005 inch) were tested in liquid nitrogen a t s h a f t speeds u p t o
35,000 rpm and differential pressure loads up t o 1.21 MPa (175 psi) per b r u s h .
The measured leakage r a t e of a single brush was 2-3 times less than t h a t
measured for a 12-toothl 0.127-mm (0.005 inch) radial clearance labyrinth seal
used as a baseline. Stage e f f e c t s were studied and i t was found t h a t two
brush seals with a large separation distance leaked less than two brushes
tightly packed together. The maximum measured groove depth on the Inconel 718
r o t o r was 25.4 pm (0.001 inch) a f t e r 4.31 hours of shaft rotation. The
Haynes-25 b r i s t l e s wore approximately 25.4-76.2 pm (0.001-0.003 inch) under
the same conditions.
Three seal runner coatings, chromium carbide, Teflon impregnated chromium, and
zirconium oxide, were tested i n liquid hydrogen a t 35,000 and 65,000 rpm with
separate 50.8 mm diameter brush seals made o f Haynes-25 b r i s t l e s and having a
nominal i n i t i a l radial interference of 129 pm. Two bare Inconel-718 rotors
were also tested as a baseline. The test r e s u l t s revealed s i g n i f i c a n t d i f f e r -
ences between the wear characteristics o f the uncoated and coated seal
runners. A t both speeds the brush seal with the bare Inconel-718 seal runner
exhibited significant b r i s t l e wear with excessive material transferring t o the
runner surface. I n contrast, the coated seal runners inhibited the t r a n s f e r
and deposit of b r i s t l e material. The chromium carbide coating showed only
small quantities o f b r i s t l e material transferring t o i t s surface. T h e Teflon
impregnated chromium coating also inhibited material t r a n s f e r and provided
some lubrication. This coating, however, i s s e l f - s a c r i f i c i n g . The Teflon
remained present on the low speed runner, b u t i t was completely removed from
the high speed brush s e a l , which tested considerably longer. The t e s t s of
W ~ Z :

the Teflon coating revealed the importance of using a lubricating and low
friction coating for brush seals t o reduce b r i s t l e and seal runner wear. The
zirconium oxide coating exhibited the greatest amount of coating wear, while
the brushes incurred only s l i g h t wear. Further t e s t i n g o f ceramics i s
recommended before making a final judgement on the v i a b i l i t y o f ceramic
coatings for brush seals because of t h e contrast between the r e s u l t s reported
by Carlile and the results presented herein. S t r i c t l y based on the r e s u l t s
presented hereinabove , the chromi um carbide and Tef 1on impregnated chromi urn
coatings were considered preferable t o the uncoated Inconel-718 and zirconium
oxide coatings because o f t h e i r good wear resistance and c h a r a c t e r i s t i c s t o
inhibit b r i s t l e material wear and transfer t o the seal runner.

55
 SUMMARY OF LN2 BRUSH SEAL TEST RESULTS

LEAKAGE FOR A SINGLE BRUSH SEAL WAS 2-3 TIMES LESS THAN
FOR A 12-TOOTH LABYRINTH SEAL.

THE MAXIMUM TEMPERATURE RISE FOR A SINGLE BRUSH SEAL

WAS LESS THAN 30K (50R) AND OCCURRED AT 0.172 MPa (25 PSID)
ACROSS THE SEAL AND 35,000 RPM. (THIS TEMPERATURE RISE
WOULD BE GREATER WITH NO PRESSURE DROP.)

A STATIC BLOWOUT TEST DEMONSTRATED SEALING CAPABILITY

UP TO 3.79 MPa (550 PSID). THE'SEAL'LIMIT WAS NOT OBTAINED.

THE POWER LOSS FOR A SINGLE BRUSH AT 35,000 RPM AND 1.21
MPa (175 PSID) WAS 1824 W (1.73 BTU/S).

TWO BRUSHES FAR APART LEAK LESS THAN TWO BRUSHES

TIGHTLY PACKED.

ROTOR WEAR WAS APPROXIMATELY 19 pm (0.00075 MILS) AND

BRISTLE WEAR WAS 25.4-76.2pm (1-3 MILS) AFTER 4-112 HOURS.

OBJECTIVE: TO EVALUATE SEAL COATINGS FOR CRYOGENIC BRUSH

SEAL APPLICATION BY TESTING VARIOUS MATERIALS ON
SEAL RUNNERS IN A CRYOGENIC BRUSH SEAL TEST RIG.

CHROMIUM CARBIDE (CC)

TEFLON IMPREGNATEDCHROMIUM
ZIRCONIUM OXIDE (210,)
UNCOATED INCONEL-718 (BASELINE)

BACKGROUND:

THE IMPORTANCEOF GOOD MATERIAL SELECTION WAS REVEALED IN THE 1ST YEAR
OF TESTING CRYOGENIC BRUSH SEALS AT NASA LERC. W I L E THE EFFORT FOCUSED
ON LEAKAGE RATE PERFORMANCE IN LIQUID NITROGEN(LN2), PRELIMINARY DATA WAS
TAKEN IN LIQUID HYDROGEN (LH2). IN LH2, THE UNCOATED INCONEL-718 SEAL RUNNER
AND THE BRUSH MADE OF HAYNES-25 BRISTLES EXHIBITED POOR TRIBOLOGICAL
PERFORMANCE. BRISTLE WEAR WAS EXCESSIVE AND BRISTLE MATERIAL
TRANSFERRED TO THE RUNNER CAUSING GALLING DUE TO LIKE-ON-LIKE METAL
CONTACT.

56
CROSS SECTION OF CRYOGENIC BRUSH SEAL TESTER

Housing P A
height

Section A-A

FIGURE 7.. BRUSH SEAL

57
 SEAL RUNNER PROPERTIES

SEAL COATING SURFACE COATING INTER- TEST

RUNNER COATING DEPOSITION PROCESS MICK- FINISH HARD- FERENCE SPEED
COATING NESS (RMS NESS FIT
(mm) crm) (W (m) (rpm)
INCONEL-718
[UNCOATED)
- l 3 46
I
-+
35,000
INCONEL-718 38 65,000
(UNCOATED)
Cr C Plasma sprayed 67 35,000
CrC Plasma sprayed 100 s 66
Cr+Teflon Cr electro-deposit and ground. Then 50.8 .6 50
Teflon impregnated into fissures
created by chemically etching the
Cy.
~ ~

Cr+Teflon Cr electro-cleposit and ground. Then

Teflon impregnated into fissures
created by chemically etching the
Cr.
60

130 I 65,000

Zr02

Zr02
Plasma sprayed onto 100 pm thick
AMI 973 bond coat
Plasma sprayed onto 100 pm thick
AMI 973 bond coat
67

62 -I= 65,000

58
 TEST PROFILES

SPEED PRESSUREDROP

I .Z I MPA
I .-PA
0.86MPA
0.52MPA

TIME

TEST SPEED : 35,000 RPM.

TIME

TEST SPEED : 65,000 RPM.

59
 EVALUATION TECHNIQUES

SEAL RUNNER WEAR QUANTIFICATION:

A SURFACE PROFILOMETER WAS USED TO MEASURE SURFACE TRACES

PERPENDICULAR TO THE BRUSH TRACK AT FOUR LOCATIONS EQUALLY SPACED
AROUND THE CIRCUMFERENCE AFTER EACH TEST.

BRISTLE WEAR QUANTIFICATION:

AN OPTICAL COMPARATOR WAS USED TO MEASURE I.R. OF THE BRUSH AT MULTIPLE

LOCATIONS EQUALLY SPACED AROUND THE CIRCUMFERENCE OF THE BRUSH AFTER
EACH TEST. RADIAL WEAR WAS BASED ON THE AVERAGE OF THE MEASURED
VALUES, AND DETERMINABLE ONLY IF THE MEASURED VALUES BEFORE AND AFTER
A TEST WERE STATISTICALLY DIFFERENT.

SEAL RUNNER AND BRISTLE WEAR CHARACTERISTICS:

A SCANNING ELECTRON MICROSCOPE WITH A BACK SCATTER DETECTOR AND X-RAY

ENERGY DISPERSIVE ELEMENTAL ANALYSIS WAS USED TO INVESTIGATE MATERIAL
TRANSFER BETWEEN THE BRISTLES AND THE SEAL RUNNER.

60
 -
UNCOATFD INCONEL 718 SE& RUNNEB

R: 41 pm
- 17pm
SEAL RU~NER.

PROFILOMETER TRACE SHOWS BUILD-UP OF

TAKEN IN BACK-SCATTER
MATERIAL ACROSS THE TRACK.

61
 BRISTLE DAMAGE LlKELY CAUSED BY MATERIAL GAIIING.

CHROMIUM CARBIDE COATED SEAL RUNNJB

65,O

0 TESTED AT 65,000
 PROPERTIES OF THE COATING.

A BACK SCATTER DETECTOR. ME

63
 -
TEFLON IMPREGNATFD CHROMIUM COATED SEAL RUNNER TESTFD

RADIAL WEAR : 56 p m
-10’um

BRUSH TRACK

T EDGE OF BRUSH TRACK

64
 35,WORPM

SEAL RUNNER ACCUMULATIVE ACCUMULATIVE ACCUMULATIVE MAXIMUM

COATING TEST DURATION LINEAR SLIDING BRISTLE WEAR RADIAL
DISTANCE SEAL RUNNER
(min) -1 (w) (w)
INCONEL-718 43 213 64 -17
(UNCOATED) (DEPOSITED)
51 223 25 6
Cr+TEFLON 28 86 (NEGLIGIBLE) (NEGLIGIBLE)
ZrO, 47 222 5 90

SEAL RUNNER ACCUMULATIVE ACCUMULATIVE ACCUMULATIVE MAXIMUM

COATING TEST DURATION LINEAR SLIDING BRISTLE WEAR RADIAL
DISTANCE SEAL RUNNER
(mid (I(m) (m) wEAR.(LM)
INCONEL-718 38 300 41 -17
(UNCOATED) (DEPOSITED)
CrC I 58 I 450 38 20
Cr+TEFLC)N 1 66 I 571 56 12
Zfl,
I 48 337 18 85

THE COATINGS REDUCED BRISTLE WEAR AND TRANSFER OF BRISTLE MATERIAL TO

THE SEAL RUNNERS AS COMPARED TO THE UNCOATEDSEAL RUNNER.

THE GOOD PERFORMANCE OF THE TEFLON IMPREGNATED CHROMIUM COATING

REVEALED THE IMPORTANCE OF USING A LUBRICOUS (LOW FRICTION) MATERIAL TO
REDUCE BRISTLE AND SEAL RUNNER WEAR.

THE SEVERE WEAR OF THE ZIRCONIUM OXIDE COATING IS AlTRlBUTED TO POOR

CERAMIC COATING QUALITY AS SUBSTANTIATED BY THE LOW COATING HARDNESS.
FURTHER INVESTIGATION OF CERAMICS FOR BRUSH SEAL APPLICATION IS
WARRANTED.

BASED ON THE RESULTS OBTAINED IN THIS EFFORT, THE TEFLON IMPREGNATED

CHROMIUM AND THE CHROMIUM CARBIDE COATINGS ARE CONSIDERED
PREFERABLE TO THE ZIRCONIUM OXIDE COATING AND UNCOATED INCONEL-718.

65
 RESULTS OF CONTACTING SEAL TESTING AT NASA LEWIS RESEARCH CENTER

Gene Addy
e-
& L/- +) 7
NASA Lewis Research Center
Cleveland, Ohio

Advanced technology initiatives such as the Integrated High Performance Turbine

Engine Technology (IHPTET) program and the Advanced Subsonic Technology (AST)
program have recognized that advancements in seal technologies are a key element of the
overall effort to improve gas turbine performance and efficiency. These improvements in
performance and efficiency call for reduced seal leakagewhile requiring the seals to operate
at higher temperatures and surface speeds.

The High Temperature/High Speed Seals Test Facility at NASAs' Lewis Research Center
has evolved as a result of the combined efforts of three separate federal government
entities. The US. Air Force had the rig built under contract by Teledyne CAE in Toledo, Ohio
and a series of brush seal tests were conducted at Teledyne as part of the contract. At the
conclusion of the contract, the USAF chose to locate the rig at NASA Lewis based on Lewis'
proposal for continued seal testing under its seal programs. Support for seal testing in this
facility at Lewis is a combined effort of the Vehicle Propulsion Directorate of the US. Army
Research Laboratory located at Lewis and NASA Lewis. The US. Army provides primary
technical support for rig operation while NASA provides primary research and development
c a p bility.

Testing has been underway at Lewis in this test rig since April 1993. Various
contacting seals such as brush and finger seals, as well as labyrinth seals have been tested
to surface speeds of 1 100 feet per second, temperatures to 1 100 degrees Fahrenheit and
pressure differentials to 100 psid. Various rotor coatings have also been tested including
aluminum oxide, chromium carbide and zirconia as well as an uncoated lnconel7 18 rotor.

Brush seals with bristles made of cobalt based alloys (i.e. Haynes 25) and with bristles
made of silicon carbide have been tested for leakage performance at high speeds and high
temperatures. Finger seals made o f lnconel X750 (a nickel based alloy] have also been
performance tested. Results from these tests will be presented and discussed.

One sealing technology which has been gaining wider use is brush seals. These seals
have been shown to reduce leakage over the widely used labyrinth seals and have the
added advantage of being able to withstand significant shaft excursions without incurring
permanent damage. However, the higher temperatures of up to 1600 degrees F and
surface speeds up to 1600 Wsec targeted by the technology initiatives will require use of
materials other than the cobalt and nickel based materials, such as the Haynes and lnconel
alloys, currently used for the bristles in the brushes.

Silicon carbide is one such substitute material for the bristles. It has shown good
resistance to wear while maintaining its strength, stiffness, and resistance to oxidation at
temperatures to 1600 degrees F. It can also be formed into the small diameter filaments of
approximately 0.003 inch needed for the brushes.

67
 HIGH TEMPERATURE/HIGH SPEED
SEALS TEST PROGRAM

Objective:
To apply the low leakage and displace-
ment-tolerant characteristics of con-
tacting seal designs to the high speed,
high temperature conditions in gas
turbomachinery

Participants:
- U.S. Army Research Laboratory
- U.S. Air Force Wright Laboratory
- NASA Lewis Research Center
- Gas Turbine Industry
- Seal Manufacturing Companies

68
 Industry Partners:
- Allied Signal Engines
- Cross Manufacturing
- Technetics
- Williams International

Rig Capabilities:
- Surface speeds to 1100 fps
- Pressure differences across seal
to 100 psi
- Temperatures to approx. 1200 F

69
 Air Flow Path During Brush Seal Test

d
8d
3
3

s
LL
w

70
71
12
 AIR FORCE SEAL PROGRAMS

Ellen R. Mayhew
Wright Laboratory
Wright Patterson Air Force Base
Wright Patterson Air Force Base, Ohio

Seal technology development is an important part of the Air Force's participation

in the Integrated High Performance Turbine Engine Technology (IHl?TET) initiative, the
joint DOD, NASA, ARPA, and industry endeavor to doubie turbine engine capabilities by
the turn of the century. Significant performance and efficiency improvements can be
obtained through reducins ifitemal f l ~ w system !ezkzge, h i sed emironmenr
requirements continue to become more extreme as the engine thermodynamic cycles
advance towards these IHPTET goals. Seal technology continues to be pursued by the
Air Force to control leakage at the required conditions. This presentation briefly describes
current seal research and development programs and gives a summary of sed applications
in demonstrator and developmental engines.

OUTLINE
INTEGRATED HIGH PERFORMANCE TURBINE ENGINE
TECHNOLOGY (IHPTET) INITIATIVE

AIR FORCE SEALS R&D PROGRAMS

ENGINE APPLICATIONS
- IHPTET DEMONSTRATORS
- PRODUCTION MILITARY ENGINES
FUTURE DIRECTIONS

CONCLUSIONS

73
 IHPTET INITIATIVE
JOINT DOD/NASA/ARPA EFFORT
GOAL: DOUBLE TURBINE ENGINE PROPULSION
CAPABILITY BY THE TURN OF THE CENTURY
- FIGHTERIATTACKENGINE GOALS
)) F N M +00%
I
)) COMPRESSOR EXIT TEMP +400F

- SIMILAR TURBOSHAFTIPROP AND EXPENDABLE ENGINE GOALS

PHASED APPROACH FOR NEAR TERM AND LONGER
TERM PAYOFFS
- PHASE I DEMO COMPLETED 1994 (+30%)
- PHASE I1 DEMO 1997 (+60%)
- PHASE 111 DEMO 2003 (+loo%)

SIGNIFICANT SECONDARY FLOW IMPACT ON ENGINE

PERFORMANCE
- HPG: EFF +4.4%==> TIT -85F OR FN +7.6%
- HPT: EFF +4.2% ==> TIT -93F OR FN +9.7%
- REDUCE/ELIMINATELPT COOLING AIR

SECONDARY FLOWS GROUPED UNDER IHPTET

COMPRESSOR TECHNOLOGY DEVELOPMENT
- 60% LEAKAGE REDUCTION GOAL

ADVANCED CYCLES POSE CHALLENGES IN SEAL

OPERATING ENVIRONMENT
- CURRENT SEALS RESEARCH (FOR 1997 ENGINE DEMOS)
)) 1400-1650 FT/SEC
)) 16OOF

74
 MILITARY SEALS R&D PROGRAMS

1988 I 1989 I1990 I 19911 19921 1993 I 19941 1995 I1996 1

EXPER 8 ANALYT INVESTIGATION OF BRUSH SEALS
BRUSH SEAL WESTIGATION
HIGH TEMPERATURE BRUSH SEALS
BRUSH SEAL ROTORDYNAMICS PROGRAM
T406 BRUSH SEALS PROGRAM
BRUSH SEAL LEAKAGE FLOW MODELING
BRUSH SEAL DEVELOPMENT PROGRAM
HIGH PERF CD FILM RlFlNG FACE SEAL

CERAMIC BRUSH S
CERAMIC BRUSH S

CURRENT AIR FORCE PROGRAMS

EG&G: ADVANCED BRUSH SEAL DEVELOPMENT

PROGRAM

PRATT 4% WHITNEY: HIGH SPEED BRUSH SEAL

DEVELOPMENT PROGRAM

TECHNETICS: CERAMIC BRUSH SEALS PROGRAM

SCIENTIFIC RESEARCH ASSOCIATES: ADVANCED

SECONDARY GAS PATH METHODOLOGY

75
 EG&G: ADVANCED BRUSH SEAL DEVELOPMENT
PROGRAM
OBJECTIVE: DEVELOP A BRUSH SEAL DESIGN METHODOLOGY FOR
MAN-RATED ENGINE APPLICATIONS (UP TO 14OOF/14OOFPS)
APPROACH:
- SEAL DESIGN OPTIMIZATION VIA RIG TEST AND CFD
- TRlBOPAlR TESTING AND EVALUATION AT ELEVATED CONDITIONS
ACCOMPLISHMENTS:
- "LOW HYSTERESIS" DESIGN OPTIMIZED FOR LOW WEAR
)) REDUCED BRISTLE FLUTTER
)) ELIMINATED BRISTLE PUSH-DOWN EFFECT

- 130 PSID SINGLE-STAGE CAPABILITY SUCCESSFULLY DEMONSTRATED

- CFD MODELING OF BRISTLE TIP LIFT-OFF FORCES IN PROGRESS
- TRlBOPAlR INVESTIGATIONIN PROGRESS

P&W: HIGH SPEED BRUSH SEAL DEVELOPMENT

PROGRAM
OBJECTIVE: DEMONSTRATE HIGH-SPEED/HIGH-TEMPERATURE
OPERATION FOR IHPTET PHASE I1 DEMONSTRATOR ENGINES
APPROACH:
- APPLICATION STUDY OF PHASE II ENGINHMISSIONFLIGHT CYCLE
FOR SURFACE SPEED AND TEMPERATURE REQUIREMENTS
- DESIGN/FABRICATE/RIGTEST BRUSH SEALS TO VERIFY THEIR
CAPABILITY AT PHASE II CONDITIONS
ACCOMPLISHMENTS:
- APPLICATION STUDY COMPLETED
REQUIREMENTS TO 1500F, 1650 FT/SEC
))

- PRELIMINARY DESIGN OF BRUSH SEAL CONCEPTS COMPLETED

- RIG ADAPTIVE HARDWARE DESIGN/FABRICATIONIN PROGRESS

76
CURRENT AIR FORCE PROGRAMS

TECHNETICS: CERAMIC BRUSH SEALS

OBJECTIVE: DEVELOP A FULLY OR PARTLY CERAMIC BRUSH SEAL
FOR APPLICATIONS FROM 1600F - 2000F
APPROACH:
- PERFECT MANUFACTURING METHODS
- RIG TEST FOR PERFORMANCEl ROTORDYNAMICS, AND WEAR
- FABRICATE HYBRID SEAL FOR TEST IN A DEMONSTRATOR ENGINE
ACCOMPLISHMENTS:
- MANUFACTURING TRIALS OF ALL-CERAMIC SEALS DEMONSTRATED
LONG TERM PROMISE
- “HYBRID” CERAMtC BRISTLWETALLIC HOLDER SEALS CURRENT
FOCUS (1600F GOAL)
B CERAMIC COATINGS SHOWED BEST DURABILITY IN WEAR TESTS
HIGH CERAMIC/CERAMIIC FRICTIONAL HEATING, WEAR DEBRIS CONCERNS
LINE-ON-LINE DESIGN MAY BE DESIRABLE FOR CERAMIC BRUSH SEALS
)> HYBRID SEAL TESTS IN PROGRESS

SCIENTIFIC RESEARCH ASSOCIATES:

ADVANCED SECONDARY GAS PATH METHODOLOGY

OBJECTIVE: DEVELOP A SECONDARY GAS PATH DESIGN SYSTEM

BASED ON A 3-D NAVIER-STOKES CODE
APPROACH:
- APPLY GRAPHICAL USER INTERFACE (GUI) TECHNOLOGY TO MAKE CFD
DESIGNER-FRIENDLY
- EXERCISE ON A DEMONSTRATOR ENGINE CAVITY AND COMPARE TO
TEST RESULTS
ACCOMPLISHMENTS:
- GUI NEARLY COMPLETE
- USER TESTING IN PROGRESS
- DEMONSTRATOR ENGINE COMPUTATIONS COMPLETE
- ENGINE CURRENTLY IN TEST

77
IHPTET DEMONSTRATOR ENGINE
SEAL APPLICATIONS

BRUSH SEAL TESTING TO DATE

- TURBINE AND COMPRESSOR APPLICATIONS
- HAYNES 25 AND HAYNES 214 BRISTLES
- ALUMINUM OXIDE, CHROME CARBIDE COATINGS
- 2040% REDUCTION IN LEAKAGE OVER LAB SEALS
- GENERALLY GOOD DURABILITY
TURBINE AND COMPRESSOR APPLICATIONS

FUTURE TESTING
- BRUSH SEALS PLANNED FOR ALL DEMO ENGINES
- HIGHER SURFACE SPEEDS, TEMPERATURES
- EXPANDED ARRAY OF BRISTLE AND COATING MATERIALS
- ADVANCED CONFIGURATIONS

AIR FORCE PRODUCTION ENGINES

ADVANCED SEALS

F l l 9 ENGINE
- 4 STATIC BRUSH SEALS
- 2 DYNAMIC BRUSH SEALS IN LPT

FIOO-PW-229 ENGINE
- DYNAMIC BRUSH SEAL IN LPT

78
 FUTURE DIRECTIONS
SEAL TECHNOLOGY
INCREASED TEMPERATURE REQUIREMENTS
- ALL-CERAMIC BRUSH SEALS
)> SPECIALLY DESIGNED FIBERS, COATINGS
NON-CONTACTING SEALS
- WEAR BENEFITS, HIGH PRESSURE CAPABILITY

APPLICATIONS

FULL INTEGRATION OF ADVANCED SEAL

TECHNOLOGY INTO ENGINE DESIGNS
- NOT ONLY "SHOWCASE" TECHNOLOGY OR QUICK FIXES

CONCLUSIONS

EXTENSIVESEALSRESEARCHHASBEEN
CONDUCTED IN PURSUIT OF THE IHPTET GOALS
- ADVANCED BRUSH SEAL MATERIALS AND DESIGNS
>>' NEW TEMPERATURUSPEED REGIMES
>> EXTENDED SEAL LIFE
- DESIGNER-FRIENDLY TOOLS FOR EFFICIENT SECONDARY GAS
PATH DESIGN
>> 3-D NAVIER-STOKES FOR THE DESIGNER

PUSHING FOR ADVANCED SEAL TECHNOLOGY IN

IHPTET DEMONSTRATOR ENGINES

WORKING WITH PROGRAM OFFICES TO TRANSITION

SEAL TECHNOLOGY

79
 PRDA-I1 AND I11 BRUSH SEAL DEVELOPMENT PROGRAMS AT EG&G

Robert G. Loewenthal
EG&G Mechanical Components Technology Group
Research and Development
Cranston, Rhode Island

1990-1995

These programs come unda the Integrated High Performance Turbine Engine Technology
0 program-
EG&G Mechanical Components Technology Group R&D completed a BNsh Seal
Development Program under PRDA-II in late 1992. We started the Advanced Brush Seal
Development program, under PRDA-Iu, in 1993 and will complete it in 1996. Both
programs have been funded by the United States Air Force. In the first program, we made
signifjcant gains in the area oftribopairs (bristle materials vs. shaft coatings) and the 'Low
Hysteresis" design for brush seals. These were reported in two AIAA Propulsion Conference
papers (copies available), and the T o w Hysteresis" design has been patented. Seals were
delivered for test in an Air Force demonstrator at Allison. In PRDA-Dl, goals are to increase
the pressure sealing capability, and the swface speeds and temperatures at which brush seals
can be used. We have conducted part of the design and testing and have tested brush seals
successfully at more severe conditions than in the previous program. W e are continuing with
thc program, and will complete it in time to furnish brush seals for an Air Force Derrumsmtm
test in 1997.

PRDA ITml:B m h Seal Programs

PRDA-11
Developed Low Hysteresis Design - Patent isssued.
- AIAA and Propulsion Journal Papers
Optimized Tribopairs with extensive test program.
- AIAAPaper
PRDA-III Goals
Single Stage Seal for Higher Pressure. 15Opsid
Design for High Surface Speed. 1200-1400f@
Increase Temperature Capability. 1200-14OO0F
Reduce Pressure Closure to decrease wear.

81
 Uso05401036A
*United States Patent 1191 [ill Patent Number: 5,401,036
Basu [~SJ Date of Patent- Mar; 28, 1995

[54] BRUSH SEAL DEVICE HAVING A 1523768 11/1989 U.S.S.R. ................................ 277/53
RECESSED BACK PLATE 201495.1 9/1992 WPO ................................... 277/53

[75] Inventor: Prithwisb Basu, Pawtucket, R.I. Primary Examiner-William A . Cuchlinski, Jr.
Axsuranf Examiner-John ki Beres
173) Assignee: EC & G M o l , Inc., Crmston. R.I. Afromey. Agenr. or Firm-Finnegan. Henderson.
1211 Appl. NO.: 35,072 Farabow. Garrett & Dunner
122) Filed: Mar.22, 1993 [571 Al3srRAcr

[SI] Int Cl.6 ............................................. F16.J 15/447 A brush seal device for sealing a high pressure area
[52) US. Cl. ...................................................... 277/53 from a low pressure a r a The brush seal device com-
[58] Field of Search ....................... 277/51. 53, 188 R, prises an annular retaining plate having a first side fac-
41W173.5, 174.5 ing the high pressure area and a second side opposite the
first side, and an annular back plate having a first side
1561 References Cited facing the low pressure area and a second side opposite
U.S. PATENT DOCUMENTS the first side, the second side of the back plate having an
4,202,554 5/1980 Snell ...................................... 277/53 outer peripheral portion and an inner peripheral por-
5.029.875 7/1991 Spain et al. ....................... 277/53 X don. A plnrality o f bristles are between the second side
5.066.024 11/1991 Rdsinger et al. ..................... 2’771’53 of the retaining plate and the outer peripheral portion of
5.071.138 1U1991 Mackay et al. __ the second side of the back plate such that they extend
5.106.104 4/1992 Atkinson et al. inwardly from the outer peripheral portion of the sec-
5,108.116 4/1992 J o h n et al. .. ond side of the back plate The inner peripheral portion
5,201,530 4/1993 Kclch et al. ................ 277/53
of the second side of the back plate has a recessed sur-
FOREIGN PATENT DOCUMENTS face formed therein to inhibit the plurality of bristles
0293140 11/1988 European Pat 0
from contacting the inner peripheral portion of the
2021210 11/1979 U d a l Kingdom . second side of the back plate.
1598926 9/1981 United Kingdom .
22582n U1993 United Kingdom 43 claims.7 Drawing Sheets

*Published, with permission, from EG&G Mechanical.

82
REFERENCES

Basu, P., et al.: Hysteresis and Bristle Stiffening Effects in Brush Seals. J. Propul. P., vol. 10, no. 4, pp. 569-
175.
Derby, J.; and England, R.: Tribopair Evaluation of Brush Seal Applications. A I M Paper 92-3715, 1992.

Basu, P., et al.: Brush Seal Design Guide. WL-TR-93-2029, 1993.

Short, J., et al.: Brush Seal Development Program.Final Report. WL-TR-93-2064, 1993.

PRDA-11 1990-1993
- Baseline Testing 12 seals 184 hours
- Tribological Evaluation Ring-on-ring samples
and 9 small (2”) brush seals
- Characterization Testing 17 seals 484 hours
- Design Selection Testing 8 seals 280 hours
D Design Selection: All were low hysteresis design
- Performance Testing 3 seals 101 hours

PRDA-111 1992-1996
Characterization testing to date:
- 6 seals -160 hours
- Approx. 300 more hours testing in 1995 and 1996, incuding
200 hour endurance test.

83
 Brush Seal Testing 1990-1995
Maximum conditions of successful testing brush seals to date:
-
- Highest temperature ever tested 120OOF
-
- Highest surface speed ever tested 1080 feet per second
-
- Highest pressure differential across a single stage seal
13Opsid

- Tested 9 inch diameter seal with .019 inch radial excursions for
500 cycles,
-
- Pressure differential for that test 57psid.
- Low Hysteresis design. 5OOOF 95Ofps.
-
- Total test time 80 hours.
- Brush seal tested satisfactorily.

84
 Flow Parameter History
Design Comparison

14 . press: 36-26 Psid ......... __

TUP:300-200 F
12 ...... w 900 Wsec .............. .. ...... ......

+
Design #I
+
L.H. Std. Dsgn.

I I I I
0 I I I 1 I I

0 10 20 30 40 50
Time hours

Flow Parameter

& = lbmass
T= temperture in 'Rankine
Pu = upstream pressure psia
Di = ID in feet

85
 Excursion Performance
DeSign#l

8 4
7.5
FLOW -- 3.75
7
+ YARAlwlm T

6.5 3.5
0
*
~6 .-
0
3.25

3 4
0
i
i 3.5 2.5
3
2.25
2.5
2 f I
I , I
1
I
I
1
I 4
I
I
I I
I
I
I
I 12
0 50 100 150 200 250 300 350 400 450 500
Excursion ##

Flow Parameter

=(&dTTI(PuDi)

= lbmass
T= temperture in 'Rankine
Pu = upstream pressure psia
Di = ID in feet

86
 Moving toward best design for IHPTET Goals.
XTC-76 - New demonstrator engine being built by a
GE and Allison team.
Positions in the engine being selected for demonstration
of the PRDA-III seal technology.
PRDA-III Advanced Brush Seal Development Program
will complete in late 1996.

87
ASPIRATING SEAL DEVELOPMENT: ANALYTICAL, MODELING AND SEAL TEST RIG

Bharat Bagepalli
General Electric Company 2-7 - 0’7
Schenectady, New York
J FYY9

ABSTRACT
This effort is to develop large diameter (22 - 36 inch) Aspirating Seals for application in
ahcraft engines. Stein Seal Co. will be fabricating the 36inch seal(s) for testing. GEs task
is to establish a thorough understanding of the operation of Aspirating Seals through
analytical modeling and full-scale testing. The two primary objectives of this project are to
develop the analytical models of the aspirating seal system, to upgrade using GE’s funds,
G E s 50-inch seal test rig for testing the Aspirating Seal (back-to-back with a
corresponding brush seal), test the aspirating seal(s) for seal closure, tracking and
maneuver transients (tilt) atoperaMg pressures and temperatures, and validate the
analytical model.

The objective of the analytical model development is toevaluate the transient and steady-
state dynamic pedormance characteristicsof the seal designed by Stein. The transient
dynamic model uses amulti-body system approach: the Stator, Seal face and the rotor
are treated as individual bodies with relative degrees of freedom. Initially,the thirty-six
springs are represented as a single one trying to keep open the aspirating face. Stops
(Contact elements) are provided between the Stator and the seal (to compensate the pre-
load in the Wy-open position) and between the rotor face and Seal face (todetect rub). The
secondary seal is consided as part of the stator. The film’s load, damping and stiffness
characteristics as functions of pressure and clearance areevaluated using a separate
(NASA) code GFACE. Initially, a laminar flow theory is used. S p e d two-dimensional
interpolation routines are written to establish exact film load and damping values at each
integration timestep. Additionally, other user-routines are written to read-in
actual pressure, rpm, stator-growth and rotor growth data and, later, to transfer these as
appropriate loads/motions in the systemdynamic model. The transient dynamic model
evaluates the various motions, clearances and forces as the seals are subjected to Werent
aircraft maneuvers: Windmilling restart; start-ground idle; ground idle-takeoe takeoff-
burst chop, etc. Results of this model show that the seal closes appropriately and does not
ram into the rotor for all of the conditions analyzed.
The rig upgrade design for testing Aspirating Seals has been completed. Long lead-time
items (forgings, etc.) have been ordered.

89
 Analytical Modeling of Aspirating Seals

TRANSIENT DYNAMICS
- Time Variation of Aspiration Motion in Moving From One
Operating Regime to Another

STEADY STATE DYNAMICS

- The Behavior and Dynamic Characteristics of the Aspirating
Seal in the Vicinity of an Operating Regime under Cyclic
Perturbations

Transient Dynamics Cases

Windmilling Restart (6 cases)

Initial closure to ground idle

Ground idle to takeoff

Takeoff to ground idle chop (deceleration)

M O O 0 maneuver

Ground idle to takeoff via worst case burst

Takeoff to worst case chop ground idle

Takeoff to cruise

90
 Steady State Dynamics Cases

Ground Idle

Take Off

Cruise

ASPIRATING SEAL 2 LOW PRESSURE SIDE

\
GARTER
BLOCK

91
92
 F1 - Load Capacity
700 1 I I I I 1 I I I
I I I I I I I 1 I
I I I I I 1 I I I
I 1 I I I 1 I I I
600 1----1----1----1----1----,----

500

400

300

200

100

0
( 10 0.50 1.00 1.50 2.00 2.50 3.00 3.50 4.00 4.50 5
Film Thickness, mils
Aspirating Seal

F1 - Load Capacity

0.00 0.50 1.00 1.50 2.00 2.50 3.00 3.50 4.00 4.50 5
Film Thickness, mils
Aspirating Seal

93
 Seal Closure to GI

FLlM THICKNESS VS. TIME

0.14 I I

0.12 - -
0.10 - -
0.08 - -
-
0.06 - -
VI
W -
I
g
- 0.04 --
- -
0.02
--
0.00 I I

Seal Closure to GI

FACE LOAD = F1 + F2

80

0 50 100 150
TIME. SECONDS

94
 Seal Closure to GI

PRESSURE LOAD
140-
--
- BACK
I I

120 - -
100-
- -
80 - -

60 - -
-
-

- -
-20 I I

Seal Closure to GI

SPRl FORC VS. TIME

-10

-
-15- -
-

-20 - -
?
v,
g -25- -
0
Q
e
-30 - -
-

-35 - -

-
-40 I I

95
 Seal Closure to GI

0 50 100 150
TIME, SECONDS

Seal Closure to GI

LOW PRESSURE
I I

I I

96
 Seal Closure to GI

HIGH PRESSURE
T I
I I

8; -

I I

GI to Takeoff

CLEA POSl VS. TIME

' " " " " I ' " " " " I " " ' " " I " " " ' " I ' " ' " " '

0 10 20 30 40 50
TIME(SEC.)

97
 GI to Takeoff

GI to Takeoff

2000 " " " " ' I ' . " " " ' I " ' . " " ' I ' " ' " " ' ~ " " " " '

-
-c
1500 - -

-
1000 - -

98
 GI to Takeoff

GI to Takeoff

SECONDARY SEAL FRICTION

1.O~""""'l""""' I " " " " ' I " ' " ' " ' I " " " ' "

0.5 -

0.0

-0.5 -
tn
!3
-1.0 - -
U
8
-1.5 - -
V
-2.0,. - - * I I . . . l m a . I . e . 3 I I . , . . I I I I C I I I I I I I S I B 1 0 . I . I I I . .

99
 GI to Takeoff

10 20 30 40 50
TIME, SECONDS

GI to Takeoff

LOW PRESSURE

10 20 30 40 50
TIME, SECONDS

100
 GI to Takeoff
ROTOR GROWTH
I " " " " ' I " " " " ' I " " " " ' I " " ' " '

-10

10
STATOR GROWTH
20
TIME, SECONDS

GI to Takeoff
30 40 50

1 ~ . . . ~ . . . . 1 . , . . . , . . . I . . , . . . . , . I . . , . , . , . .

20
TIME, SECONDS
30 40 50

101
Chop to GI

Chop to GI

102
Chop to GI

Chop to GI

103
 Chop to GI

LOW PRESSURE
1 ' ' ' ' " ' ' ' 1 " " ' ' " ' I " " " " ' I ' " ' " ' "

-
-

-:
-
-
-

I . . , . . . . . . I . . . . . . . . . I....,,.,.I,.....,..

Chop to GI

10 20 30 40 50
TIME, SECONDS

104
 Chop to GI

c FACE LOAD = F1 + F2

Chop to GI

ROTOR GROWTH
I " " ' " " I " ' " " " I " ' " " " I " " " "

l l l . I . I . l l l l l l l

10 20 30 40 50
TIME, SECONDS

105
 Chop to GI

c STATOR GROWTH

-
3

f , . . . . . . . , I , . . . , . , . . I . I ,. I . . I 2 1 . . , . , . . . .-

106
 STEADY STATE DYNAMICS

Face

3D or Face

Axial Perterbations: ?Ax, k A y , 2 A z

Angular Perterbations +A8,, + b e y

Result in runout at seal face (TIR)

Nominal Design Requirement:
+Ax, ?Ay, ~ A =z2 2 mils (?)

TIR = rt 10 mils at 36 in.

APPROPRIATE DEGREES OF FREEDOM, (5)
STIFFNESSES,AND INERTIAS OF THE SEAL ARE CONSIDERED

107
I
DYNAMIC ANALYSIS OF SEALS
GE ASPIRATING SEAL, CRUISE CONDITIONS, CASE 2
lI- HMIN.--

4.000 I I
I
i
I
!
I

3.000

2.000

t
1.ooo
!
I
t
I
I
f
!
I
!
I i
.ooo
.00 2.00 4.00 6.00 8.00 10.00 12.00
REVS.

108
 Minimum Film Thickness
equilibrium = 2.25 mils

GROOVED

- . .- - _.- - .- -
I- HMIN. *03p'~i3
DYNAMIC ANALYSIS OF SEALS I

GE ASPIRATING SEAL, TAKEOFF CONDITIONS, CASE 2

2.800
I
i
i i ! I

1
I
I !
!

2.400 I

2.000
B
w
1.600 ! !
I

i
1.200
! I
i
.oo 2.00 4.00 6.00 8.00 10.00 12.00
REVS.

109
 x, = Y o=z,= 2 mils
misalignment = 210 mils TIR

GROOVED

- ZS-DISP.
DYNAMIC ANALYSIS OF SEALS , ---- 2-DISPL.
--- .
GE ASPIRATING SEAL, IDLE CONDITIONS, CASE 2

4.000

I
I
!
I

2.000

.WO

-2.000
I
!
!
i
!

I !
i i I

I
4.000
.oo 2.00 4.00 6.00 8.00 10.00 12.00
REVS.

110
 X, = Y o=Z, = 2 mils
misalignment = 2 10 mils TIR

Axial Vibrations
Synchronized

GROOVED

DYNAMIC ANALYSIS OF SEALS

GE ASPIRATING SEAL, CRUISE CONDITIONS, CASE 2

4.000

2.000

.oOO

-2.000

i I I
-4.000 I
.00 2.00 4.00 6.00 8.00 10.00 12.00
REVS.

111
 X, =Yo=Z,= 2 mils
misalignment = 2 10 mils TIR
f = 0.2

Axial Vibrations
Synchronized

GROOVED

.. -

I---
ZS-DISP. 5fF4L
DYNAMIC ANALYSIS OF SEALS
Z-DISPL. &7D<
GE ASPIRATING SEAL, TAKEOFF CONDITIONS,CASE 2

4.000

2.000

.*$ .ooo

-2.000

! I
-4.000 .
.oo 2.00 4.00 6.00 8.00 10.00 12.00
REVS.

112
 Aspirating Seal Test
Test Program Objectives
Phase I Determine seal closing pressure.
Determine seal leakage at cruise and take off pressure conditions at room temperature, at 400” F and
at operating temperature.
Demonstrate seal tracking capability over full range of relative axial motion.
Phase ZZ Determine sed behaviour under maneuver deflections - tilt mechanism.
Determine seal behaviour with rotor runout -- 0.005” and 0.010” TIR machined into rotor

Test Rig Capabilities

Aspirating seal and brush seal inlet air flow measured independently; inlet pressure controlled
independently.
Rotor can be moved axially -.35/+.95”.
Aspirating seal cone designed to allow retrofit of tilt mechanism.
Aspirating seal instrumentation includes:
1 accelerometer -- axial for rub detection
3 prox probes - axial for film thickness measurement
3 thermocouples -- seal exit air temperature measurement

Aspirating Seal Test Rig

113
 c
INDUSTRIAL COMPUTER CODES -BcG/
Wilbur Shapiro
Wilbur Shapiro & Associates, Inc.
2yy3-0
Niskayuna, New York

Industrial Codes

b GCYLT* * Updated
b GFACE **.New
SPIRALG
b SPmALI*
b IFACE
b ICYL
DYSEAL**
b KTK**
ms**

GCYLT
(Gas - Cylindrical-Turbulent)

Multi=GeometryCode
Steps

Tapers

Hydrostatic

115
 GCYLT Turbulent =

Theory

Reynolds equation for turbulent compressible flow for journal bearings is as follows:

The equation is made dimensionless with the following definitions. (Upper case variables are
dimensionless).

A=- 6paR
P,C.
,to
12@
POCO
2'
,G, and G, = turbulence modifiers

Substituting the dimensionless variables into the turbulent Reynolds equation produ e
dimensionless equation.

116
Couette Turbulence Coefficients

b
c
0
P

0.01 1 I I I I Ill1 I I I I I Ill1 I I 1 I I I I l l

100 200 500 1,000 2,000 5.000 20,000 100,000

Reynolds Number, hrw p / p

Poiseuille Turbulence Coefficients

0.01
103 104 10s 108 107 10’
Pressure-Based Reynolds Number, Re*

117
 GCYLT - Turbulent
Theory (Continued).

The turbulent G factors are dependent upon the Couette and Poiseuille Reynolds numbers which
are at each grid pointp1.
The Couette Reynolds number is

where the subscript c refers to the cell comer point (e.g., for Q,;;P, = PI>.The PoiSede
Reynolds number is defined as:

r 'M

118
 GCYLT Comparisons

CoueUe Turbulence Comparison Poiseuille Turbulence Comparison

GCYLT GBEAR CCYLT GBEAR

0.252 0.252 h,mils 0.25 0.25
1270 1269 W Ib 9.513 9.449
4.92 4.63 Y 0.03 0
1.2
-
1.2
-
=
QI
hP
ws
9.48 x IO-’ 9.43 x 10-5
-0270 .0267
43,740 40.917 K, 1Wi. 10.05 -
6,458,000 6,443.64 Kr 1Wi. 46,170 47.27 1
-5,062,000 1 1Win. -38,860 -37,188
919,000 - IWi. 23.16 0.78
1920 S,OSS,S 1 D, Ib-din. 728.7 739
-270.7 6 D, lb-din. -0.60 -
334.3 813,689 DyK Ib-din. 0.82 0.88
2649 1936 D, lb-din. 862.7 875
16 -166 M, lb -.35 -2.8
23,987 -147 a, rpm 511 497
2650
12 N =IOOOrpm E?, = 14,700 psig
24,064 Pa= 500 psig Excitation Frequency = 0

N = 50,OOO rpm P, = 14,700 psig

& = 48.995 Excitation Frequency = 0

119
SPIRAL1 (Spiral-Groove - Incompressible)
b Spiral Groove Seals
b Parallel Grooves
b Helical Grooves
b Arbitrary Axi-Symmetric Geometry
b Cylindrical and Face
b Turbulence and Inertia
b Limitations
Concentric
Hir's Bulk Flow Model
Narrow Groove Theory

SPIRAL1 - Local Inertia Neglected

I
t hr

I 1

120
121
SPIRAL1 - Ungrooved Cylindrical Seal Comparison

-
Table I.Comparison of test cases 5 8 with data published by D. W. Childs (Ref. 4)

Childs, D. W. "Finite Length Solutions for Rotor Dynamic Coefficients

of Turbulent Annular Seals, ASME, J. Lub. Tech., V105,3, (1983),pp 437-445

122
 SPIRAL1 = Definition of Coefficients

Overall Seal Discharge Coefficient

C,.= AP
1 pV2
2

Radial Force Coefficient

-f, = K + co - Ma2 4 Kef - Mefo2

Tangential Force Coefficient

-f, = CO - k = Cefo

SPIRAL1 - Parallel Groove Pressure Breakdown Seal

Rotor
Q-
N.G I WL .I WG I HG I C
9 Grooves( 2.38 mm I 2.38 mm I 0.38 mm 10.356 mm

Radius = 50.8 mm

Childs, D. W., Nolan S . A., Kilgore J. J., "Test Results for Turbulent Annular Seals
Using Smooth Rotors and Helically Grooved Stators", ASAME, J. Tnb. V 112,2, (1990),
pp 254-258

123
SPIRAL1 = Parallel Groove, Flow Coefficient

0.8 -

7200
0.2 -

01 I I I I I
1 I I
0 5 10 15 20 :25 30 ' 35 40
AP (bar)

SPIRAL1 - Parallel Groove, Direct Stiffness

4.0 ,-

-1.0 -

I I I 1 I I I
-3.0
0.75 1.0 1.25 1.5 1.75 2.0 2.25 2.5

Re, x l o 5

124
 SPIRAL1 = Parallel Groove, Damping

4.00 -
2.00-

0.00 I I I . 1 I I

SPIRAL1 - Helical Groove Flow

8 60'
140 -

KO -
do t.00-
0
I m 60'
om- =-
30'
0.50 - f Y f *
Y 15.
0 0 U U U

-
0.40

OM -
aoof I I I 1 I 1 I

0.00 5.00 U.00 544 20.00 25.00 3OD3 35.00 40.00

loo0 RPM. hkt loss coefGcW = 1.0
AP (bar)

125
SPIRAL1 = -
Helical Groove Direct Stiffness
6.00

5.00

4.00

E
g
\

is.
3.00
Y

2.00

too

0.00
0.00 100 2.w 3.00 4.00 5.00
R , ~
1OOO RPM. hlel bss coellicient E 1.0

SPIRAL1 - Helical Groove - Effective Damping

Data for C,
0 15'
w.00 - 130'
60'
14.00 -
3
12.00 -
xc
u-
I
x).oo -
0

8.00 -

6.00 -

4.00 -
2.00 -
0.00 I I I I 1
0.00 100 2.00 3.00 4m 5.00
R, 1o - ~
loo0 RPM. inlel kss coellicienl= 1.0

126
 KTK - Labyrinth Seal Code

b Knife to Knife Approach (KTK)

b Empirically Based
b Leakage and Internal Pressure
Distribution
b Optimun Geometry Option
b Straight or Step Seals
Flow going up or down the step

KTK - Labyrinth Seal Code - Loss Zones

b Contraction 1 to 2,4
to 5
b Venturi and Friction -
2 to 3,5 to 6
b Partial or Full
Expansion 3 to 4,6 to
7

127
KTK Labyrinth Seal
=

Code = Straight Seal Parameters

128
KTK Labyrinth Seal
=

Code Step Seal Parameters

=

LTSD Flow Down the

Land Step (not shown)

Geometric parameters for straight and stepped seals

Knife height (KH)

Knife pitch (KP)
Number of knives (a)
Knife angle (KO)
Knife tip thickness (KT)
Knife taper angle (KB)
Knife tip leading edge radius (KR)
Clearance (CL)
Surface roughness (01

Additional parameters considered for stepped seals

Step height (SEI)

Distance to contact (DTC)
F l o w direction (LTSD or STLD)*

F l o w parameters

Overall pressure ratio (Pg)

Inlet stagnation pressure (Pu)
Fluid temperature distribution (T)
Flow rate (W)

*LTSD = Large to Small Diameter.

STLD = Small to Large Diameter.

129
 KTK - Labyrinth Seal
Code - Parameter Ranges
Parameter Ranges of Data in Labyrinth Seal Data Base

Seal TVDe SteDDed Seal

Single Straight
Parameter knife seal STLD dit. LTSD dir.

KN min 2 2 2
UlaX 12 6 6

KT/CL min 0.21 0.21 0.50

UUX 4.4 2.64 1-50

K8 min 60 50 50
UlaX 90 90 90

m/cL min 2.7 5.1 5.1

Inax 31.3 29 e 4 28.0

KP/CL min 4.0 6.4 9.2

tUX 56.3 53 40

&CL) min 0 0 0
max 0.030 0 0.030
--
SH/CL min
m X _- 2.0
29.4
4.0
12.5

DTC/CL min _- 0.85 4.1

IMX -- 40 19.4

(KP-KT) /CL min 3.5 6.2 8.9

max 55 .o 51.8 38.5

130
KTK - Labyrinth Seal Code - Optimization

131
KTK - Labyrinth Seal
Code - Optimization (Continued)

50

40

30

20

10

0
0 0.01 0.02 0.03 0.04

0 ORIGINAL + O P CLEARANCE('N)
T o o OPT(STEPPE0)

Lealuge versus Clcutunce -

TCP, SSO, FP-172 I n l e t Seal
(PU = 795 psia, PD = 715 psi., AP = 80 p s i )

132
DYSEAL - Dynamic Response of Face and Ring Seals

b 5 DOF for face seals, 2 for ring seals

b Piston ring, O-ring, Contact face
secondary seals
b Coulomb friction of secondary seals
b Interface represented by cross-coupled
stiffness and damping
b Forward integration in time algorithms
based on Newmark's method
b Continuation Option
b Plotting included

DYSEAL - Face Seal Configurations

133
 DYSEAL - Ring Seal Configurations

--

Low Pressure

DYSEAL - Interface Stiffness and Damping Coefficients

X Y 2 B a F\A X Y Z B a

-Ringseals

134
 DYSEAL - output
b Seal Ring - x, y, z, a, b
b Shaft - x, y, z, a, b
Film thickness,
b Minimum film thickness
b Friction Forces and Moments- x, y, z, a, b
b CG, Mass, It, Ip
b Closing Forces
b Continuation parameters

DYSEAL - Examples of Plotted Output

5
WNUU:ANUYSlS OF S W

I -1 -1

135
DYSEAL - Examples of Plotted Output (Continued)

!
L

DYSEAL - Friction Algorithm

i
I ‘\1
Fa= Applied Force
F, Friction Force
I
L‘
dt-I
IB

136
 DYSEAL - Ringsealcheck

(.bo0

I . 000

3.000
0
L

4.000

.(.009 -+
4.000
-
t--l
-I.oQQ .wo 1.000 *.ooo
XS-DXSP.

.
I
00 4.00 8. 12.00 16.00
REVS.

137
 DYSEAL - Face Seal Check

T i e (msec)

b) Slnuaoldal Seal Seat Mode; Amplltude P 50 pm (2 mll);

Frequency P 100 Hz

I DYNAIUC m Y S I S O? S W

14000 UP? ALIGNED, 2 MIJ,. 100 HZ A X I U EXCIYRTION

r

138
 COMPARISON OF SPIRALG PERFORMANCE PREDICTIONS WITH A
SPREAD SHEET SOLUTION

Martyn G. Roberts
Atlanta Engineering
Rolls-Royce, Inc.
Atlanta, Georgia

The aero design of an inward pumping spiral groove face seal using an in-house
spread sheet was compared with predictions from the NASA code SPIRALG. The
high pressure compressor exit of an aero gas turbine was chosen as the location for
the candidate seal. This is a challenging environment as rotational velocity, pressure
drop, and temperature are high.

This presentation compares the resulting lift forces, leakages, and friction loss for
various ride heights. Within practical ranges of ride height, the lift force predictions
agreed well. However, both leakage and friction loss predictions were significantly
different.

Design Constraints

Velocity = 320 m/s (1OOOft/s)

Temperature = 700°C (1300°F)

Pressure Drop = 1350 kPa (200psi)

RR Spreadsheet based on:

t+ E.A. Muijderman, Spiral Groove Bearings
Philips Technical Library 1966

139
 Lift Force

-
22000

20000 1 1 I I I 1

Lift Force

28000 -
RR SPREADSHEET

26000:-

A
-
3
E 24000 -
2

22Ooo -

20000. I 1

140
 Friction Loss

Ride Height (micron)

Friction Loss

u 1 I I I I
0 10 20 30 40 50
Ride Height (micron)

141
 Leakage

8o 1
70 -
60-

50 -
8 40-
m
4 30-

20 -
10 -

01- , I I I I

-10
-10 P 50 100 150

Ride Height (miwon)

200 250 300

Leakage

0.6 -
0.5 -
A 0.4-
0.4
PD

B& 0.3-
0.3
s
Q
9 0.2 -
0.1 -
0
0 10 20 30 40 50

Ride Height (micron)

142
 Prediction Differences (1)
I VU70

8OYo

60%

4040

20Yo

0%

-20%

-40%

-609/0

-80%

-100%
RMe HeigM (Microns)

Prediction Differences

Friction Loss
\\ \ \

Ride Height (Microns)

143
 COMPARISON OF CODE PREDICTIONS TO TEST MEASUREMENTS FOR TWO ORIFICE
COMPENSATED HYDROSTATIC BEARINGS AT HIGH REYNOLDS NUMBERS
c-
John E. Keba >/o -37
Rocketdyne Division
Rockwell Corporation
Canoga Park,California
3 953-2

ABSTRACT

Rotordynamic coefficients obtained from testing two different hydrostatic bearings are
compared to values predicted by two different computer programs. The first set of test
data is from a relatively long (UD=l) orifice compensated hydrostatic bearing tested in
water by Texas A&M University (TAMU Bearing No.9) under Dr. Dara Childs (1). The
second bearing is a shorter (UD=.37) bearing and was tested in a lower viscosity fluid
by Rocketdyne Division of Rockwell (Rocketdyne “Generic” Bearing) at similar rotating
speeds and pressures (2). Computed predictions of bearing rotordynamic coefficients
were obtained from the cylindrical seal code “ICYL“ (3),one of the industrial seal codes
developed for NASA-LeRC by Mechanical Technology Inc., and from the hydrodynamic
bearing code “HYDROPAD” (4) developed by Dr. Luis San Andres of Texas A&M
University.

The comparison highlights the difference the bearing has on the accuracy of the
predictions. The TAMU Bearing No. 9 test data is closely matched by the predictions
obtained for the HYDROPAD code (except for added mass terms) whereas significant
differences exist between the data from the Rocketdyne “Generic” bearing the code
predictions. The results suggest that some aspects of the fluid behavior in the shorter,
higher Reynolds Number “Generic” bearing may not be modeled accurately in the
codes.

The ICYL code predictions for flowrate and direct stiffness approximately equal those of
HYDROPAD. Significant differences in cross-coupled stiffness and the damping terms
were obtained relative to HYDROPAD and both sets of test data. Several observations
are included concerning application of the ICYL code.

145
 REFERENCES

1. Unpublished data. Testing was conducted in the Hybrid Bearing Test Stand at
Texas A&M under Rocketdyne Div., Rockwell sponsorship. For a description of the test
method, see: D. W. Childs, “Theory Versus Experiment for the Rotordynamic
Coefficients of Annular Gas Seals: Part 1 - Test Facility and Apparatus”, ASME
Trans., Vol. 108, July 1986.

2. Unpublished data. Testing conducted by Rocketdyne Div. Rockwell. For description

of test method, see TECHNOLOGY TEST BED ENGINE - EXPERIMENTAL
VERIFICATION OF ROTORDYNAMIC ANALYSIS, FINAL REPORT; John E. Keba,
Report No. RSS-8856-24, Rocketdyne Div., Rockwell. Prepared for NASA-Marshall
Space Flight Center under Contract No. NAS8-40000 Task 31.

3. USERS MANUAL FOR COMPUTER CODE ICYL - CYLINDRICAL SEALS

LUBRICATED BY INCOMPRESSIBLE FLUIDS, Antonio F. Artiles, Report No. 92TM13
Rev. July 1993, Mechanical Technology Incorporated. Prepared for NASA-Lewis
Research Center under Contract No. NAS3-25644.

4. “HYDROPAD -- User’s Manual”, L. S. San Andres, Texas A&M University, Oct.

1992.

146
 INTRODUCTION

BEARING GEOMETRY

(units) I TAMU No. 9 I Rocketdyne GENERIC

Nomini Diameter (D) inch 3.00 2.69
Length(L1 inch 3.00 1 .o
Clearance (radial) (ho) inch .00475 .0049
Recess Geometry
~~
No. of Recesses~
5 9
Depth inch .OlO .020
Axial length inch 1.064 .485
Circumferential length inch 1.064 .485
Area Ratio .200 .250
Orifice Geometw
Throat Diameter
~~ ~ ~
inch I .098 I .085
Exit Chamfer Counter Bore Sharp Edge
Surface Roughness
Land -- between uin
~~
I ~~ ~
13 ~
I 12
Land -- side pin 13 12
Rotor uin 13 6

ORIFICE COMPENSATED HYDROSTATIC BEARING

147
 BEARING TEST CONDITIONS

(units) TAMU No. 9 Rocketdyne GENERIC

I TEST FLUID WATER
~
HFC-134a
I Temperature "F 130 100
Viscosity Ibmlft-sec 340. E-6 122.5 E-6
I Density Ibm/cu ft 61.5 71.7
SHAFT SPEED rPm 10,200 to 24,600 0 to 24,000
I
Rotational Re oRhh to 23,000 to 67,000
1 AP psi 600 to 1000 800 to 1600
FLOWRATE GPM 29 max. 39 max.
Flow Re Q/(xDv) to 15,000 to 72,000

SIGNIFICANT DIFFERENCES BETWEEN DATA SETS

TAMU No. 9 GENERIC

Land Axial Length to

Film Thickness Ratio 200: I 50: 1
Orifice Exit Geometry:
Orifice Diameter .098 inch .085 inch
Recess Depth .010 inch .020 inch
Exit Geometry Counter Bore Sharp Edge

Pressure Ratio:
(Precess-Pout)/(Pin-Pout) .58 .32
Fluid CompressibiIity:
Bulk Modulus (adiabatic) 320,000 psi @ 68 "F 45,000 psi @ 950 psig, 100"F
Density Change -
Podpin Negligible 1.075 @ 800 psid, 30°F AT
Reynolds No.
Rotational 22,200 67,000
Flow (avg. axial on lands) 12,300 49,800

148
 PREDICTIONS VERSUS RESULTS -- TAMU BEARING NO. 9

MEASURED BEARING PERFORMANCE

0 Direct and Cross-Coupled Stiffnesses relatively large, approximately equal,

and both strongly dependent on speed and bearing AP.

TAMU Test Series 9

Stiffness vs. Speed and Pressure
1000000 *
900000
I ! I I I I I I I I I I

800000
700000
600000
500000
400000
300000
200000
100000
0
I
-100000
I ! I
I I I I I I I I 1

Shaft Speed, 10200 17400 24600 10200 17400 24600 10200 17400 24600
rPm
Bearing Delta PI 1000 psid 800 psid 600 psid
psid

149
MEASURED BEAR1NG PERFORMANCE (con't)

0 Direct damping large, relatively insensitive to speed and bearing AP

TAMU Test Series 9

Damping vs. Speed and Pressure
1000 I I I I I I I I I I 1 - 4 3 -

I I I I I I I i I

800
700
600
500
400
.-F 300
E" 200
d I00
I I I I I Y I I
I
,"" I
II I
I
I
I
I
I
II II I
I
I
I I
I 1
I

Shaft Speed, 10200 17400 24600 10200 17400 24600 10200 17400 24600
rpm
Bearing Delta P, 1000 psid 800 psid 600 psid
psid

150
HYDROPAD PREDICTIONS

0 Orifice Flow Coefficient selected to match flow data at one operating point.
Match of flow prediction to data at other operating conditions shows
that orifice coefficient is constant.
0 Direct Stiffness, Direct Damping and Cross-coupled stiffness predictions
were within +/- 20% of test measurements
Measured cross-coupled stiffness and direct damping increased less
with speed than predicted

TAMU Test Series No. 9

HYDROPAD vs Test Measurements
2.00
1.80
1.60
1.40
1.20
1.oo
0.80
0.60
0.40
0.20
0.00
Shaft Speed, 10200 17400 24600 10200 17400 24600 10200 17400 24600
rPm
earing Delta P, 1000 psid 800 psid 600 psid
psid
I ~~~

-a- Flowrate
~

-E+ Direct Stiffness +Cross-Coupled K+ Direct Damping I

151
ICYL PREDICTIONS

0 Orifice Flow Coefficient matched to measured flow data

Independent of operating conditions
Direct Stiffness within 20% of test data except at lowest speed
0 Cross-Coupled Stiffness and Direct Damping approximately 50% of
measured values

TAMU Test Series No. 9

ICYL Predictions vs Test Measurements
2.00
1.80
5 1.60
c

6 1.40
L

2cu 1.20
8 1.00
\

.- 0.80
Y
0
*e0.60
Q,
0.40
0.20
0.00
Shaft Speed, 10200 17400 24600 10200 17400 24600 10200 17400 24600
rpm
earing Delta P, 1000 psid 800 psid 600 psid
psid

rLFlowite ~ -e- Direct Stiffness +-Cross-Coupled K+ Direct Damping I

152
 PREDICTIONS VERSUS TEST RESULTS “GENERIC” BEARING --

MEASURED BEARING PERFORMANCE

0 Direct stiffness roughly one-half of TAMU No. 9 bearing at similar

operating conditions
0 Cross-Coupled Stiffness and Direct Damping about one-tenth of
TAMU No. 9
0 Direct Damping is speed dependent

Generic Bearing
Measured Stiffness vs. Speed, Pressure
500,000
450,000
400,000
z 350,000
.-2 300,000
5 250,000
v)- 200,000
v)
150,000
100,000
50,000
0
(50,000)
(100,000)1 I I I I I I I I I I I I I
Shaft Speed, 0 8k 16k 24k 0 8k 16k 24k 0 8k 16k 24k
rpm
Bearing Delta P, 1600 Psid 1200 psid 800 psid
psid

153
MEASURED BEARING PERFORMANCE (con’t)

Generic Bearing
Measured Damping vs. Speed, Pressure
140

120
5 100
.-c
2 80
60
J
6, 40
.-C
E 20
g o
-20
-40
Shaft Speed, 0 8k 16k 24k 0 8k 16k 24k 0 8k 16k 24k
rpm
Bearing Delta P, 1600 psid 1200 psid 800 psid
psid

154
HYDROPAD PREDICTIONS

Orifice Flow Coefficient selected to match flow data

e Coefficient significantly smaller than TAMU bearing because of sharp
edge geometry
Direct Stiffness underpredicted by constant ratio
Direct Damping overpredicted at low speed, under predicted at higher
speeds
Cross-Coupled stiffness overpredicted but magnitude still small compared to
Direct Stiffness
Discrepancies not pressure dependent

GENERIC BEARING
HYDROPAD vs Test Measurements
2.00
1.80
c
1.60
E 1.40
L

z 1.20
([I

91.00
\

.-5 0.80
U
0
-50.60
Q
2 0.40
0.20
0.00
Shaft Speed, 0 8k 16k 24k 0 8k 16k24k 0 8k 16k24k
rpm
Bearing Delta P, 1600 Psid 1200 psid 800 psid
psid
Flowrate -e- Direct Stiffness 4- Cross-Coupled K +Direct Damping I

155
ICYL PREDICTIONS

Orifice Flow Coefficient matched to measured flow data

Independent of operating conditions
Direct Stiffness underpredicted, similar to HYDROPAD prediction
Direct Damping roughly one-third of measured value
Cross-Coupled Stiffness magnitudes are small
Discrepancies are not pressure dipendent

GENERIC BEARING
ICYL Predictions vs Test Measurements
2.00
1.80
1.60
E 1.40
p!
3 1.20
([I
SJ 1.00
\

.-
*
0.80
0
0.60
a,
0.40
0.20
0.00
Shaft Speed, 0 8k 16k 24k 0 8k 16k 24k 0 8k 16k 24k
rpm
Bearing Delta P, 1600 psid 1200 psid 800 psid
psid

I +I- Flowrate -H- Direct Stiffness +Cross-Coupled K +Direct Damping I

156
 NOTES REGARDING ICYL USAGE

Code was run on Silicon Graphics IRIS workstation

Provided faster execution but lacked interactive input and graphics capability
of PC/OS2

Bearing modeled as follows:

0 Bearing symmetric about axial center plane (ISYM =I)

0 Pressures at circumferential start of model equal to pressures at end
(IPER=1)
0 Grid mesh size (TAMU Bearing Model) --
0 Axial - 7 total on half bearing; 5 on land including step, 2 in recess
0 Circumferential - 8 per recess sector: 5 between recesses including
steps, 3 in recess
0 Friction factor used IFRICT=O. (Smooth, Ng-Elrod turbulence model.
Perturbed equations would not solve with IFRICT=3. Solution time with
IFRICT=4 was approximately 45 minutes per case; similar results obtained)

Results from ICYL

0 Flowrate must be multiplied by 2 to obtain the quantities presented.

Flow balance checked correctly (against orifice flow calculation) when
factor of 2 applied

157
 CONCLUSIONS

0 GOOD PREDICTION OF TAMU BEARING PERFORMANCE

0 ICYL prediction of direct stiffness accurate

0 SIGNIFICANT DISCREPANCIES FOR “GENERIC”BEARING

0 Underprediction of direct stiffness unexpected
Direct damping error is speed dependent

0 SHORT, HIGH CLEARANCE BEARING AND LOW VISCOSITY FLUID MAY BE

CAUSE
0 High Reynolds Number and short lands result in inertia dominated flow
losses which have greater uncertainty
0 Model assumes fully developed flow in land -- entrance region error may
become significant in short bearings
Low pressure ratio of GENERIC bearing causes stiffness prediction to be
more sensitive to errors in predicted pressure losses in land region
0 Effect of fluid downstream of test bearing may be significant with shorter
bearing

158
SCISEAL: A CFD CODE FOR ANALYSIS OF FLUID DYNAMIC FORCES
IN SEALS

M.M. Althavale, Y.-H. Ho, and A.J. Przekwas

CFD Research Corporation 5/-3$(
Huntsville, Alabama
-57qg
5/

OUTLINE

Objectives

Current Status

Code Capabilities

Selected Results

Concluding Remarks

ACKNOWLEDGEMENTS

Code Development Done Under LeRC Contract NAS3-25644

--- Technical Monitor: R.C. Hendricks, LeRC

Program Manager: A. Liang, LeRC
MTI Manager: W. Shapiro, MTI

Technical Help, Encouragement from the Following People:

- Drs. A.K. Singhal, Y.G. Lai, CFDRC
-- Dr. B. Steinetz, LeRC
from Allison, UTRC, and GEAE

159
 OBJECTIVES (CFDRC)

Develop Verified CFD Code for Analyzing Seals

Required Features Include:
- Applicability to a Wide Variety of Seal Configurations
such as: Cylindrical, Labyrinth, Face, and Tip Seals
- Accuracy of Predicted Flow Fields and Dynamic Forces
- Efficiency (Economy) of Numerical Solutions
- Reliability (Verification) of Solutions
- Easeof-Use of the Code (Documentation, Training)
- Integration with KBS

SCIENTIFIC CODE DEVELOPMENT

Task 1: Develop a 3D CFD Code (SCISEAL) for

Cylindrical Seals

--- for Annular, Tapered, Stepped

Verification of Code Accuracy
Rotordynamic Coefficient Calculatlons

Task 2: Augmentation of SClSEAL

- Incorporation of Multi-Domain Capabilities

Note: Starting CFD Code = REFLEQS (developed by CFDRC

under a contract from NASA MSFC/ED32)

160
 STATUS 3RD SEALS WORKSHOP

Single Domain 30 Code

- colocated grids
--
-
high-order schemes
rotating and moving grid systems
turbulence models, including 24ayer model

Seal Specific
-- grid generation
pre-processing, geared for cylindrical seals

Rotordynamics
-- whirling rotor method for centered seals
perturbation method for centered as well as eccentric seals

Extensive Validation Effort

- long and short annular seals
-- labyrinth flows
entrance loss coefficients

AUGMENTATION IN SCISEAL
AND PRESENT STATUS

Major Item: Incorporationof Multi-Domain Formulation

-- flow treatment completed
all flow models extended to multi-domain
- additional: - 2D/3D switchable for efficiency in
planar/axisymmetric solutions
- conjugate heat transfer capability (completed in a
- separate contract)
local refinement multi-twne grid interfaces
Rotordynamics
- whirling rotor method transferred, extended to multi-domain grids
and user input grids
- perturbation model transferredfor cylindrical seals
Validation Effort Continued
- whirling seal flow solutions obtained and compared with
experiments (single domain code)
- several rim seal + disc cavity + labyrinth seal computations
performedand compared (multi-domaincode)

161
 main ~ e t h is
~ Fully
d
C~nse~ativ~

The ch has trated

a Va 2 0 and

nt to S
Flows

e Has or One-to-
Zon

ed Grid Match

162
 MULTI-DOMAIN INTERFACETREATMENT (Cont.)

Basic Interface Stencil

n
621 I I

Implicitness and Conservation

CURRENT CODE CAPABILITIES

Seals Code has:

Finite Volume, PressureBased Integration Scheme
Colocated Variables with Strong Conservation
Approach
-
High-Order Spatial Differencing up to Third-Order
Up to Second-Order Temporal Differencing
Comprehensive Set of Boundary Conditions
Variety of Turbulence Models (k-E, Low Re k-E,
Multiple Scale k-e, 2-Layer Model), Surface Roughness
Treatment
Multi-Domain Capability with Multi-to-One Interface
Treatment
Conjugate Heat Transfer
Moving Grid Formulation for Arbitrary Rotor Whirl
Rotordynamic Coefficient Calculation Methods, CFD
Based for Centered Seals Circular Whirl
Small Perturbation: Centered & Eccentric Seals

163
 SEAL SPECIFIC CAPABlLIT1ES

-
Preprocessor Geared for Seals Problems
Easy, Quick Geometry Definition and Grid Generation
Four Types of Cylindrical Seals:
- Annular, Axial Step-Down, Axial Step-up,
and Tapered
One Line Commands for

-- Automatic Grid Generation

Integrated Quantities: Rotor Loads, Torque, etc.
- Rotordynamic Coefficients

CODE VALIDATION AND DEMONSTRATION

Code has been Validated for a Large Number of

Benchmark Problems
- a list of 33 relevant problems

Extensive Validation Effort Conducted for Practical

Seals:
-- annular and tapered seals
labyrinth seals (stepped and straight)

Turbine Main Gas Path and Secondary Path Interaction

-- rim seal ingestion (UTRC)
large scale rig (UTRC)
- T-56 Stage 1-2 (Allison)

164
 VALIDATION CASES

1. Fully-developed flow in a pipe and channel.

2. Developing laminar flow in a narrow annulus between two cylinders.
Slug flow at inlet, fully-developed flow at outlet.
3. Laminar flow between rotating cylinders. Below critical Taylor number,
tangential flow only. -
4. Flow between two cylinders, rotating inner cylinder. Taylor vortex flow,
Laminar and turbulent.
5. 2-Ddriven cavity flow, Reynolds number up to 10,000. Comparisons with
numerical results by Ghia et.al.
6. 3-D driven cavity flow.
7. Couette flow under different pressure gradients. With and without heat
transfer.
8. Planar wedge flow in a slider bearing.
9. Laminar flow over a back step. Reattachment length comparison with
experiments by Arrnaly and Durst.
10. laminar flow in a square duct with a 90" bend. Comparison with
experimental data by Taylor et.al.
11. Shock reflection over a flat plate.
12. Turbulent flow in a plane channel. FulIy-developed solution at exit
compared with experiments by Laufer.
13. Turbulent flow induced by rotating disk in a cavity. Comparison with
experiments by Daily and Nece.
14. Centripetal flow in a stator-rotor configuration. Comparison with
experiments by Dibelius et.al.
15. Flow between stator and whirling rotor of a seal. 2-Dresults for 0,0.5, and
synchronous whirl frequencies

165
 VALIDATION CASES
16. Flow over a bank of tubes.
17. Turbulent flow in an annular seal. Comparison with experiments by
Morrison et.al.
18. Turbulent flow in a 7-cavity labyrinth seal. Comparison with
experiments by Morrison et.al.
19. Turbulent compressible flow and heat transfer in turbine disk cavities
Athavale etal.
20. 3-D driven cavity flow with lid clearance and axial pressure gradient.
Control of flow through vortex imposition.
21. Flow in cavities on a rotor for an electrical motor. Interaction of Taylor
vortices with driven cavity flow.
22. Flow in infinite and finite length bearings (without cavitation).
Comparison of calculated attitude angles with theory.
23. Flow and rotordynamic coefficient calculation for straight,
incompressible seals. Comparison with results from other numerical
and analytical solutions; Dietzen and Nordmann.
24. Flow and rotordynamic coefficients in tapered compressible flow seals.
Comparison with bulk-flow theory results; Nelson.
25. Rotordynamic Coefficients in a long annular incompressible flow seal.
Comparison with experimental data; Kanemori and Iwatsubo.
26. Calculation of entrance loss coefficients in the entrance region of a
generic seal. Effect of flow and geometry on the loss coefficient
values; Athavale et.al.
27. Flow coefficient and pressures in a 5 cavity, straight knife, look-through
labyrinth seal. Comparison with experimental data ;Witting et.al.
28. Flow coefficients and pressures in a 3 cavity, tapered knife, bok-
through labyrinth seal. Comparison with experimental data; Tipton
et.al.
29. Flow coeffients and pressures in a 2 cavity, straight-knife, stepperd
labyyrinth seal. Comparison with experimental data; Tipton et.al.

166
 VALIDATION CASES
30. Turbulent flow in a whirling annular seal, comparison with
experimental data, Morrison et. al.
31. Rim seal ingestion computation, comparison with experimental data,
Graber et. al.
32. Simulation of turbine disk cavity flow and interaction with main path
flow, Allison T-56 engine, turbine stage 1-2 cavity and labyrinth seal.
33. Flow interaction in a multiple disc cavity rim seal rig with intra-cavity
and secondary main path flow interactions. Comparisons with
experimental data from Daniels and Johnson.

SELECTEDRESULTS
Several Annular and Labyrinth Seal Results Presented
Earlier

Presented Here are Results for

-- whirling annular seal (Morrison, Texas A&M, 1992 & 1995)
rim seal gas ingestion (Graber et. al., UTRC, 1987)
-- Allison engine turbine cavity (Munson and Forry, Allison)
large scale rig UTRClMSFC (Daniels and Johnson,
UTRC, 1993)

167
WHIRLING ANNULAR SEAL - PROBLEM DESCRIPTION

R =82.5mm
Nominal Clearance C = 1.27 mm AY
I

E = 0.5 C

Seal Length L = 37.3 mm

Working Fluid: Water
Re = 24,000
Ta = 6,600 (rpm = 3,600)

U, = 7.49 mlsec
Row Rate = 4.83 literdsec

WHIRLING ANNULAR SEAL - COMPUTATIONAL MODEL

Rotating Frame of Reference

Grid = 40x20~15 wall

Incompressible + Standard k-E model

Inlet Boundary: seal entrance
(measured profiles)
Downstream Boundary: seal exit
(fixed p)
Periodic Boundary is assumed in the
circumferential direction

Results: Velocity Distributions and Pressure Along the Seal Length

168
 WHIRLING ANNULAR SEAL RESULTS

Contours of Ua,ial/U, at Several Axial Stations

Pressure side Suction side

(a> x/L = 0.00125 x/L = 0.0

03)x/L = 0.2125 X/L = 022

h'umerical Experimental

169
 WHIRLING ANNULAR SEAL RESULTS

Contours of Ua,ial/U, at Several Axial Stations

(c) x/L = 0.4875 x/L = 0.49

(d)x/L = 0.7625 x/L = 0.77

Numerical Experimental
170
 WHIRLING ANNULAR SEAL RESULTS
Contours of U,ial/U, at Several Axial Stations

(e) x/L = 0.9875 x/L = 0.99

b'umerical Expedimental

171
 WHIRLING ANNULAR SEAL RESULTS

Contours Of U,adial/Um at Several Axial Stations

(a) x/L = 0.00125 x/L = 0.0

(b)x/L = 02m x/L = 033

Numerical Experimental

172
 WHIRLING ANNULAR SEAL RESULTS
Contours of Uradial/Um at Several Axial Stations

(c> x/L = 0.4875 x/L = 0.49

(d) x/L = 0.7625 X L = 0.77

Numc-i cal Experimentd

173
 WHIRLING ANNULAR SEAL RESULTS

Contours of U,dial/U, at Several Axial Stations

(e) x/L = 0.9875 x/L = 0.99

Numerical Experimental

174
 WHIRLING ANNULAR SEAL RESULTS
Contours of ue/w,h at Several Axial Stations

(a) x/L=O.OOl25 x/L = 0.0

(b>x/L = 02m x/L = 0.22

Numerical Experimental
175
 WHIRLING ANNULAR SEAL RESULTS
Contours of u@!sh at Several Axial Stations

(c) x/L = 0.487s XfL = 0.49

(d)x/L = 0.7625 x/L = 0.77

Numerical Experimental

176
 WHIRLING ANNULAR SEAL RESULTS

Contours of u@sh at Several Axial Stations

(e) x/L = 0.9875 x/L = 0.99

Numerical Experimental

177
 4
OJO 0.5
See color plate on page 35 1.
 -
RIM SEAL PROBLEM DESCRIPTION

mo 5 co
t--- main gas path
mixture F1,

Stator
Disc
cavity Rotor f
R, 20% C02,80% N2

Cmean

mixture F2
5% C02,95% N2

qt = nondimensional purge mass flow = (Rf+)4'8

L Z R J
4 = cooling effectiveness = c-c,
c, - c o
S2R:
Ret = tangential Reynolds number =

RIM SEAL INGESTION COMPUTATIONS

Variation of Cooling Effectiveness due to Purge Mass Flow

Fixed Inlet Swirl Reynolds Number

Four ConfigurationsTested (reported by UTRC in AFWAL-

TR-87- 2050, Sept. 1987)

Each Configuration Tested at Two Purge Flow Rates

Ingestion Experiments Simulated Using Passive Scalar

Transport of Inert Species

179
 COMPUTATIONAL CONDITION

Two Purge Flow Rates qt = 1 x IOm3,8 x

-
Rotor Speed 2450 rpm, Ret = 5.0 x IO5

Inlet Swirl = Rotor Tangential Velocity

Axial Velocity Range 120450 m/s for the Four Configurations

Computational Grids: 50-60 cells in axial direction

60-70 cells in radial direction

180
 =. =. =.
C E E . Is
0

8
B

181
 Allison Engine Tubine Cavity
Stage 1-2 Disk Cavities MplM99 (M112Ln 0m Ln
Ibk T O mis T e n
M p 0.219
13x7 1009 320 1w
11 L: 1071 '
P
Satrld9 &73 1147 .M3 1147
3 .050 818 .Os1 818

Two labyrinthseal clearances are simulated (0.012in and 0,024 in)

Temperature
Stage 1-2 Risk Cavities

Labyrinth seal clearance = 8.012 in Lab seal clearance =

See color plate on page 348.

182
See color plate on page 349.

LE I

PT

platforms.

183
 I

184
 UTRC L N
Main patit Main path
I%

F1
Purge

0.3
0.2
0.1
0

See color plate on page 344.

185
See color plate on page 345.

See color
186
See color plate on page 347.

Regions I, I1 and Blade

See color plate on page 341.

187
 Regions 111, IV and Blade Shanks, Na. 202

See calor plate on page 342.

CONCLUDING REMARKS

0 A 3D CFD

* lti- On
Flow Geometries
 CONCLUDING REMARKS (Cont.)

Seal Specific Capabilities

- Rotor Loads, Torques, etc

Rotordynamice Coefficient Calculations

-- -
Full CFD Based Solutions Centered Seals
-
Small Perturbations Method Eccentric Seals

Extensive Validation Effort

189
 THERMOHYDRODYNAMIC ANALYSIS OF CRYOGENIC LIQUID TURBULENT
FLOW FLUID FILM BEARINGS

Luis San Andres

Mechanical Engineering Department
-
/2
j -37 I

Texas A&M University

College Station, Texas

Computational programs developed for the thermal analysis of t i l t i n g -

and flexure-pad hybrid bearings, and the unsteady flow and transient response
of a point mass rotor supported on f l u i d film bearings are described. The
mation of a cryogenic l i q u i d on the t h i n film annular region of a f l u i d film
bearing i s described by a set of mass and momentum conservation, and energy
transport equations for t h e turbulent bul k-flow velocities and pressure, and
accompanied by thermophysi cal state equations for evaluation of the f l u i d
material properties. Zeroth-order equations describe the f l u i d flow field for
a journal s t a t i c equilibrium position, while first-order (linear) equations
govern the f l u i d flow for small amplitude-journal center translational
motions. Solution t o the zeroth-order flow field equations provides the
bearing flow rate, load capacity, drag torque and temperature rise. Solution
t o the f i rst-order equations determines the rotordynami c force coefficients
due t o journal radial motions.
The hydroflext program calculates the static load and dynamic force
coefficients for the following bearing geometries:
1. hydrostatic bearings w i t h orifice compensation and rectangular recesses
(single row or two-parallel recess row),
2 . annular pressure seals (damper seals) (cylindrical and multilobe),
3. p l a i n cy1 i ndri cal hydrodynamic bearings (cy1 i ndri cal and mu1 t i 1obe) ,
4. fixed arc hydrodynamic bearings w i t h arbitrary preload,
5. t i 1t i ng-pad journal bearings ,
6. flexure-pad journal bearings (hydrostatic and hydrodynamic),
7 . cylindrical pad bearings w i t h a simple elastic matrix (ideal f o i l bearing).
Hydrofl ext i ncl udes the f o l l owing thermal models :
- adiabatic surfaces, i . e . insulated journal and bearing surfaces.
- isothermal journal a t specified temperature and insulated (adiabatic)
bear i ng .
- i sothermal bearing a t specified temperature and insulated (adi a b a t i c )
journal.
- i sothermal journal and bearing surfaces.
- isothermal journal and radial heat flow through bearing (stator).
- adiabatic journal and radial heat flow through bearing (stator).

Numerical computations and comparisons t o experimental results from the

open literature are detailed. The major features of the programs are also
descri bed. Some interesting results for herringbone journal bearings
operati ng a t 1arge journal eccentricities are presented and the 1 imi t a t i ons of
the Narrow Groove Theory are unveiled.

19 1
lution to the fit-order
andangular mOti0~.

192
Date: January 1, 1995 (latest release), January 1,1994 (originalrelease)
A

hydm#kd includes the following thennohydrodynamic models:

- adiabatic surfaces,i.e. insulatedjoumal and bearing surfaces.
- isothermaljournal at specified temperature and insulated (adiabatic)
bearing.
- isothermal bearing at specified temperature and insulated (adiabatic)
joumal.
- isothermaljournal and bearing surfaces.
- isothermal journal and radial heat flow through bearing (stator).
- adiabatic journal and radial heat flow through bearing (stator).
The extended computer program hydrotrun embodies hydroflext, and
calculates the dynamic (transient) force response of a rigid rotor
supported on fluid film bearings. The code calculates at each time step
of numerical integration the unsteady bulk-flow field and the fluid film
bearing forces due to prescribed timevarying external loads. The
transient analysis is restricted to i s o t h e d flows and rigid pad
bearings.
hydmfIext calculates
1) bearing flowrate or seal leakage
2) friction torque,power dissipation and temperature rise
3) load capacity (fluid film forces and moments) if journal eccentricity
is given, or journal equilibrium eccentricity components if the external loads
are given.
4) 16 complex impedancecoefficients due to dynamic journal center

193
 displacementsand journal axis rotations. The real and imaginary parts
of the impedancescorrespond to the stiffness and damping coefficients
evaluated at a specified excitationfrequency. Inertia coefficients are
also determined from evaluationof impedances at two whirl frequencies.
5 ) stability indicator or whirl frequency ratio for lateral journal
motions and equivalent stiffness at operating speed.
6) pssure and temperaturefields on the bearing surface, and density and
viscosity field variations, within ranges of fluid flow Reynolds
numbers and Mach numbers.
for
1) isothermal flow with bamtropic fluid,
2) themohydrodynamic adiabatic or radial heat flow and/or isothermal
joumal and bearing surfaces in the single phase flow regime.
with the following.fluids:
1) liquid hydrogen, 2) liquid nitrogen,
3) liquid oxygen, 4) liquid methane,
5 ) water, 6) oil,
7) air, 8) bamtropic liquid.

For cryogenic liquids, the fluid properties (density, viscosity and specific heat)
are calculated with the miprops program fromNBS Standard Database 12.

hydm@xt handles the following boundary cmditions at the bearing exit

planes:
1) periodic pressure asymmetry in the axial direction.
2) local discharge end seal effects via an orifice like model to
simulate wear-ring hydrostatic bearings or annular seals.
3) inlet specified circumferential pre-swirl velocity distribution.

The axial clearance functions included are of the type:

a) uniform, b) tapered,
c) stepped, or, d) arbitrary via spline interpolation.

Cylindricalbearings can be specified as multi-lobe geometries, and

bearing pads can have a specified assembly preload.
For tilting-pads and flexurqad bearings, pad mass moment of inertia and
flexure rotational stiffness and damping coefficientc are needed for full
specificationof the beaxing geometry.

194
Method:
The motion of a fluid on the thin film annular region of a fluid film
bearing is describedby a set of mass and momentum conservation,and
energy @amportequations for the turbulent bulk-flow velocities and
pressure, and accompaniedby themphysical state equationsfor evaluation
of the fluid material pperties. Zemth-order equations descn‘be the
fluid flow field for a journal static equilibrium position, while
fmt-order (linear) equations govern the fluid flow for small amplitude
journal center translationalmotions and journal axis angular motions.
Solution to the zeroth-order flowfield equations provides the bearing
flow rate, load capacity, drag torque and temperaturerise. Solution to
the f m t d e r equations determines the rotordynamic force coefficients
due to journal lateral and angular motions.
Features:
Fully Developed, Laminar and/or Turbulent Bulk-Flow Model
Liquid of Variable properties,functions of pressure and temperature
with realistic themphysical equations of state.
Governing Equations:
Mass conservationequation,
Bulk-Flow Momentum equations in circumferentialand axial directions
Energy Tramport equation for Mean Flow Temperature with radial heat
flow through the bearing sleeve and adiabaticjournal (shaft).
Turbulence Closure Model: Bulk-Flow with friction parameters based on
Moody’s friction factor equations for roughened surface conditions.

Numerical method of solution: Control volume - finite difference

(SIMPLEC) method.

Limitations and Restrictions: Does not include elastic deformations due to

thermal effects.

Documentation:
-Complete technical report (185 pp.)
- Tutorial and User’s Manual (60pp.)
-ExamplesManual (app. 250 pp.) :Test cases and output.
-On-line HELP available while running program in interactive mode.

Input:
- -
keyboard input menu driven with error checking.
-data files created by program

195
OUtRUt:
- monitor (screen) and ASCII data files
- dump files store all information on running session
- field data files for plotting pressure, film thickness and temperature as
Z=Z@,Y) surfaces.

Laneuage: FORTRAN77. Source code provided.

Hardware: programs developed for standard fortran in workstations and VAX

computers. PC version (486 or higher recommended)will soon be available.

-
Users: NASA Research Centers and contractors, Pratt & Whitney, Rockwell
International, and U.S. Air Force - Phillips Laboratories

Developers: Texas A W Tribology Group - Turbomachinery Laboratory

MechanicalEngineering Department
Texas A&M University
College Station, TX 77843-3123
phone: (409) 845-0160, fax :(409) 845-1835
e-mail: [email protected]

Available to U.S. companies from the Texas A&M University System

Technology Licensing Office.
Texas A W University
College Station, TX 77843-3 123
phone: (409)847-8682, fax : (409) 845-1402

Sponsors of Development:
NASA Lewis Research Center, Pratt and Whitney
Rockwell International, TAMU Turbomachinexy Research Consortium

Other Programs: Earlier versions of thennohydrodynamicprogram include:

hydroseult :hydrostatic bearings and annular seals (1994),
be& annular seals (1992),
hydmbed: hydrostatic bearings (1992)
same codes without (t) extension handle isothermal fluid Nm bearings.

196
 Y

hydrotran

force ~ ~ ~ e ~ a n c e

pliance bearing structure

197
 ynolds Nmbers
v

Shear Fluw R e ~ n Q #
~d ~ to
due
Ids # due to axfal flow.

LO2 126,28~ 55,112 201,00 0.10

51,178 Z36,O

LN2 9~,904 58,276 154.3 0.08

198
D()lDt = 6( + ul 6( )/&7 + u2 6( )/6m 4s the material derivative.
speed, Qs=heat to jo t and bearing, Vi=.f2.l?, Vz=O
%ir is journal surface shear stress, R: rotational speed.
f3: vo\umtric expansibn coefficient, KJ: turbulent shear factors.

199
200
 metries of bea

PS Pa

36Weg or pads

Plain JB JB with axial grooves

Etllptk JB 2 lobe JB 3 lab JB w or w h

w or wlo offset offset

~ ~ m p l i a JBs
nt

20 1
 Y

U 1
X

Hvdrostatic Bearings:
LO2 HJBs : Redecliff and Vohr (1969)
LH2 HJBs : Butner and Murply (1986)
H20 HJB: Chaomleffel and Nicholas (1988)
H20 HJBs: Childs et al. (1988-1 994), 36 test cases.
Oil HJB: Adams et ai. (1992) - laminar flow

Annular Pressure Seals:

H20, conv-div, Childs and Lindsey (1993)
H20, Kanki and Kawakami (1984)
air, Childs, Yang and Alexander (1994)

202
0 SA Rese ter

AF P L

Pratt and Wh , Un Te

8 Texas A&M Tur

prnent:

203
Pne lcmatic Ham mer
a) Limit recess axial length
reduce recess volume

204
205
206
 Y

s e d with low exit pressures

bearings may

!1
.(

Experimental results 2
?
are favorable. How to r n d i f j , i
bulk flow model ?
I : I

Models are qualitative a d highly apprcrxirnate,

Demonstrated tm low load capacity for primary
power,
WilL there euer be parameter
tests?

207
 THE ACOUSTIC INFLUENCE OF CELL DEPTH ON THE ROTORDYNAMIC
CHARACTERISTICS OF SMOOTH-ROTOFUHONEYCOMB-STATOR
ANNULAR GAS SEALS

Dara W. Childs 73-37

Mechanical Engineering Department
Texas A&M University
College Station, Texas

A two-control volume i s employed for honeycomb-stator/smooth-rotorseals, w i t h

a conventional control-volume used for the through flow and a "capacitance -
accumulator" model for the honeycomb cells. The control volume for the
honeycomb cells is shown t o cause a dramatic reduction i n the effective
acoustic velocity of the main flow, dropping the lowest acoustic frequency
i n t o the frequency range of interest for rotordynamics. In these circurn-
stances, the impedance functions for the seals can not be modeled w i t h
conventional (frequency-i ndependent 1 stiffness , damping , and mass coeffi -
cients. More general transfer fuctions are required t o account for the
reaction forces, and calculated here as a lead-lag term for the direct force
function and a l a g term for the cross-coupled function. These f i r s t order
functions are simple compared t o transfer functions for magnetic bearings or
foundations, For synchronous response t o imbal ance, they can be approxi mated
by running-speed-dependent stiffness and damping coefficients i n conventional
rotordynami c codes. Correct predictions for s t a b i 1 i t y and transient response
wi 11 requi re more general algorithms , pressumably using a state-space format.

209
 WALL SHEAR STRESS, WALL PRESSURE AND NEAR WALL VELOCITY
FIELD RELATIONSHIPS IN A WHIRLING ANNULAR SEAL

Gerald L. Morrison, Robert B. Winslow, and €3.Davis Thames 111

Turbomachinery Laboratory 31q-3 L7/
Texas A&M University
College Station, Texas 3 7956
ABSTRACT vv Radial velocity variance
The mean and phase averaged pressure and wall shear W Mean azimuthal velocity
stress distributions were measured on the stator wall of a 50% W* Nondimensional mean azimuthal velocity, W/W,
eccentric annular seal which was whirling in a circular orbit at the w* Rotor surface velocity
same speed as the shaft rotation. The shear stresses were measured ww Azimuthal velocity variance
using flush mounted hot-film probes. Four different operating Z Axial distance downstream of the seal entrance
conditions were considered consisting of Reynolds numbers of AP Pressure drop from seal entrance to exit
12,000and 24,000 and Taylor numbers of 3,300 and 6,600. At each 0 Shear stress direction, 0" = downstream direction,
of the operating conditions the axial distribution (from z/L=-0.2 to 90' = direction of rotor rotation
1.2) of the mean pressure, shear stress magnitude, and shear stress K Turbulence kinetic energy, (uu+vv+ww)12
directionon the stator wall were measured. Also measured were the K* Nondimensional turbulence kinetic energy, Knr:
phase averaged pressure and shear stress. These data were P Absolute viscosity, 7 . 8 4 ~ 1 0kglm-s
~
combined to calculate the force distributions along the seal length. P Density, 999 kg/m3
Integration of the force distributions result in the net forces and @R.M Direction of resultant axial moment
moments generated by the pressure and shear stresses. The flow @R.MXY Direction of resultant radial moment
field inside the seal operating at a Reynolds number of 24,OOO and Direction of radial force
@Rm
a Taylor number of 6,600 has been measured using a 3-D laser <> Phase averaged quantity
Doppler anemometer system. Phase averaged wall pressure and
wall shear stress are presented along with phase averaged mean INTRODUCTION
velocity and turbulence kinetic energy distributions located 0.16~ Noncontacting seals for rotating components in
from the stator wall where c is the seal clearance. The relationships turbomachines can only minimize leakage from high to low pressure
between the velocity, turbulence, wall pressure and wall shear regions; they can not prevent it. This is inherently so due to the seal
stress are very complex and do not follow simple bulk flow clearances between the rotating and stationary components that
predictions. comprise them. Two such seals commonly used in turbomachines
today are the labyrinth and annular seals.
NOMENCLATURE Labyrinth seals can produce less leakage than annular
Nominal clearance between rotor and stator, 1.27 mm seals for a given clearance and length, thereby, providing higher
Rotor diameter, 163 mm efficiencies. However, the higher leakage rates in annular seals can
Rotor eccentricity ratio be desirable for they result in more damping and hence increased
Radial resultant force, FLDAP rotordynamic stability (Allaire et al., 1978). Annular seals are
Axial resultant force geometrically similar to plain journal bearings but present a
Azimuthal resultant force different flow structure dominated by turbulence, fluid inertia
Rotor length, 35.6 mm effects and large axial flows. In fact, there is great potential for
Resultant moment about radial line, M/LDZAP them to be used as support elements in high speed cryogenic
Resultant moment about axial line turbopumps (Van Pragenau, 1990). Consequently, it is of great
Pressure interest to learn more about these seals.
Nondimensional Pressure, PUcAP A simple annular seal consists of a smooth shaft rotating
Time averaged pressure concentrically inside a stationary, smooth cylinder. Typically,
Nondimensional Mean Pressure p-,-Pm,,m )/AP however, annular seal applications are much more complex. For
Distance above centered rotor surface example, seal surface finishes may be hydraulically smooth but have
Reynolds number = 2pU-c/p high relative roughnesses compared to the small clearances present.
Wall shear stress Built in misalignments and/or misalignments resulting from
Nondimensional Stress, TdpW, dynamic or static loading during operation can cause a shaft to
Time averaged nondimensionalizedshear stress, rotate eccentrically inside its housing (Chen and Jackson, 1987).
T-C/PW* The destabilizing forces that cause the eccentricity can lead to a
Taylor number = (pW,c/p)(2dD)ln phenomenon called whirl in which the rotor center precesses around
Mean axial velocity the center of the stator.
Nondimensional mean axial velocity, UIU, Due to these flow field complexities the importance of
Average mean axial velocity, Q/xDc, 7.4 m/s having a solid understanding of both the fluid mechanics and
Axial velocity variance rotordynamics of annular seals becomes apparent. Especially since
Mean radial velocity they affect seal performance and, therefore, the efficiency and
Nondimensional mean radial velocity, VIU, reliability of the turbomachinery in which they are installed. Much
21 1
of the previous work on these seals has focussed primarily on machined to provide an eccentricity ratio of 50%. This results in the
rotordynamic effects. In recent years Morrison, et al. (1994) have rotor whirling in a circular orbit at the same speed as the shaft
performed measurements on the velocity field and turbulence rotation.
characteristics in the clearance of a whirling annular seal. They Brass instrument supports have been machined to fit into
completed an extensive 3-D laser Doppler anemometer investigation the optical window opening of the rig which is normally used for
of the flow field inside an annular seal with a whirl ratio of 1.00 and laser Doppler anemometer measurements. Four blocks were drilled
an eccentricity ratio of 0.50. They ran two Reynolds Number and tapped to house the pressure transducers (Figure 2), and four
conditions (12,000 and 24,000) with a Taylor Number of 6,600. similar blocks were fabricated to hold the shear stress probes. These
Prominent features found include: a peak axial velocity that begins blocks provide several axial measurement locations from
in the suction side of the seal at the inlet and rotates around to the approximately six clearances upstream to five clearances
pressure side of the seal at the exit; a vena contracta on the suction downstream of the seal resulting in axial distributions of the Wall
side of the rotor; a highly 3-dimensional flow with large radial pressures and shear stresses. Figures 3-5 illustrate the measurement
velocities near the inlet of the seal (made evident by recirculation grids used for the pressure, shear stress, and velocity measurements.
zones at the inlet); and, no significant increase in tangential velocity The block measurement surfaces were machined to fit flush with the
with whirling motion when compared to statically eccentric seals. stator wall so as not to disturb the flow.
In order to further characterize these clearance flows it is necessary Olivero-Bally et al. (1993) measured wall pressure
to investigate the hydrodynamic wall pressures and shear stresses on fluctuations in turbulent boundary layers. They studied noise
the stator wall that surrounds the seal. cancellation techniques and used pinhole mounted transducers to
The purpose of this study is to experimentally determine improve the spatial resolution of the wall pressure data in the low
mean and phase averaged characteristics of both the pressure and and high frequency range respectively. Pressure measurements were
shear stress along the stator wall of an annular seal using made in both air and water with piezoresistive transducers. In order
piezoresistive pressure transducers and hot-film anemometers to minimize spatial averaging effects, the transducers were mounted
respectively. This data will complement the existing flow field data in plexiglass housings with 0.3 mm diameter flush mounted
ofThames, (1992) and Momson, et al., (1994) while providing the pinholes. Their study indicated that by placing flush mount pressure
boundary conditions necessary to develop more accurate transducers behind a small pinhole the spatial resolution can be
Computational Fluid Dynamics (CFD) models and bulk flow increased without degrading the frequency response characteristics
models used for predicting annular seal flow fields. The further over certain frequency ranges. This technique was used in the
development of these models are necessary so they can be used to present study using mounting blocks as shown in Figure 2. Kulite
design new turbomachines with improved leakage characteristics flush mount piezoresistive transducers were mounted in the blocks.
based upon better understanding of rotordynamic stability. The large cavity ended 1 mm from the block surface and a 0.40mm
diameter hole was drilled through. The assembly was mounted in
FACILITIES the test section and the pressure transducers loosened in their
The experimental facility used for this investigation is mounts in order to purge any air from the cavity. This assured
comprised of a water storage and supply system, the seal rig test adequate frequency response (1 kHz) for the present study.
section for modelling the annular seal environment, and the The mean pressures were measured with both a
piqzoresistive pressure and hot-film anemometry instrumentation. ScaniValve system and by recording the DC voltages from the
These flow facilities have been previously described by Morrison, Kulite transducers mounted in the brass blocks. Both gave the same
et al. (1994). results. The pressure transducers were calibrated each day in situ
Water is pumped to the test rig from a 10 m3tank at a flow by connectinga pneumatic dead weight tester to the test section and
rateof 2.43 and 4.86 Us and at a constant inlet pressure of 138 kPa pressurizing the test section with known pressures. The calibrations
resulting in an axial Reynolds numbers of 12,000 and 24,000. The obtained were repeatable from day to day. Uncertainties on the
flow rate and pressure are controlled by ball and gate valves both pressure measurements are 1.5 P a .
upstream and downstream of the rig. Flow rates are monitored by Hanratty and Campbell (1983) completed an extensive
a turbine flow meter. A 70 kW rated heat exchanger connected to review of wall shear stress measurement techniques including the
the supply pump by-pass line provides cooling of the working fluid hot-film probe. According to the authors, the positioning of the
to maintain the operating temperature window of 30 2 4°C. All probe relative to the wall is one of the greatest sources of error in
water supplied to the test rig is filtered by a 10 micron filter bank using heated films. Pessoni (1974) found that the displacement of
near the discharge of the supply pump. a commercial probe from the test section surface by 0.1 mm in a 2.5
Figure 1 is a schematic of the seal rig that simulates the cm air tunnel resulted in 30 to 40% deviations in the calibrations.
seal environment and where the pressure and shear stress Pessoni suggests either calibrating the probe in siru or in a plug of
measurements are made. The test section is a stainless steel design larger area and transferring the entire plug to the test section as was
which comprises two main parts. The first is the inlet where a done in this study.
stainless steel plate with 3.2 mm holes, and a fine nylon mesh are The hot-film probes used in this study (TSI 1237W)
held. These pieces straighten and further filter the flow before it consist of a 3 mm diameter rod with the end cut off square. On this
reaches the plenum in front of the seal. The second main part of the end the sensing element (a small rectangular film) is mounted. This
test section is where the stator and rotor are mounted. The rotor rod is mounted in the brass blocks SO that the end of the rod is flush
(Da.163m) is mounted on an overhung shaft that is driven by an with the block surface. The probes were calibrated in a facility
electric motor controlled by a variable frequency drive operated at consisting of two parallel smooth plates which allows calibration of
1,800 and 3,600 rpm for this test case resulting in a Taylor numbers the probes while seated in the brass blocks. Pressure taps on the
of 3,300 and 6,600. The clearance between the rotor and stator has back side of the calibrator are used to measure the axial pressure
a nominal value of 1.27 mm (0.050 in.) and the rotor has a length of distribution which verifies that dP/dX is constant indicating that the
35.6 mm. The seal is mounted on a brass bushing that has been flow between the plates is fully developed. The mean shear stress

212
on the wall is calculated in a channel of height h for a pressure drop dimensionalization values produce non-dimensional fluctuating
AP along length L as follows (Haritonidis, 1989): pressures and shear stresses with comparable amplitudes. However,
as shown in Table I, the nondimensional pressures and shear
-
T =--APh stresses differ in dimensional value by the first being approximately
L 2 two orders of magnitude larger. The mean velocities and K were
nondimensionalizedby using the bulk flow velocity, U, (7.4 d s )
A Rapid Systems A/D converter possessing 4 channels, except for the azimuthal velocity which was nondimensionalizedby
each with 8 bit resolution was used to measure the AC component W,, (30.9 m/s). Further information about the test facility,
ofthe hot-film and pressure transducer signals at a sampling rate of measurement procedure, and data availability on MS-DOS disks and
2,000 sampledsecond with a low-pass filter of 1,000 Hz. The hardcopy format are available in Thames (1992), Winslow (1994)
external trigger was connected to an electronic proximitor probe and from G.L. Momson.
situated at the keyway of the main seaYrotor shaft between the
electric motor and the overhung shaft on which the seal is mounted. RESULTS
Thirty-two averages were necessary for the current study to obtain The magnitude and direction of the mean wall shear stress
repeatable phase averaged results. The DC value of the signal was along the length of the stator are shown in Figures 6 and 7
measured using an analog voltmeter. respectively. The shear stress magnitude plot shows at least an eight
Determination of the shear stress was complex since the fold rise in Pm just inside the seal. This is followed by a less rapid
shear stress calibration was highly non-linear. The fluctuating decay to a value of about half the maximum by Z b 0 . 2 . This rapid
voltages could not be multiplied by a constant to obtain T~ as they rise and fall in shear stress is due to a vena contructa present on the
were for the pressure conversions. Rather, the instantaneous seal rotor near the entrance (Morrison, et al.( 1994)) which forces
voltages were added to the mean voltages and then converted to the the fluid through a very small clearance which then increases
total instantaneous shear stress using the calibration curve. The downstream of the vena contractu. In addition, the rotor is
mean shear stress was then calculated and subtracted from the total accelerating the fluid tangentially, thus producing a very large wall
instantaneous shear stress to give the fluctuating shear stress shear stress at the entrance to the seal followed by the decrease as
component. Root-mean-squared (rms)values were calculated using the cross-sectional region available for flow increases. From
these fluctuating values and an averaging scheme was also Z / M . 2 to 0.7 the shear stress remains relatively constant with a
implemented to determine the phased average shear stresses and trend towards increasing caused by the tangential velocity
phase averaged rms values. increasing due to the rotor's viscous drag. The exception to this is
The flush mount hot-film probes used in this study had a for the high Reynolds number, low Taylor number case where the
rectangular sensing area which made them sensitive to the direction shear stress actually decreases in this region. This is due to the flow
of the shear stress much in the same way a single hot wire is becoming more fully developed in the axial direction which
sensitive to the velocity direction. To record the shear stress decreases the wall shear stress and the lack of rotor speed to
direction the probes were rotated, within specially designed probe generate an offsetting increase in wall shear stress by imparting
mounts which kept them flush, to find a minimum in voltage. When larger tangential velocities to the fluid. At the exit the wall shear
the long side of the hot-film is aligned with the mean shear stress stress increases to values equal to the maximum inlet shear stress.
vector a minimum in heat transfer from the film occurs (Ludwieg, Figure 7 shows that the direction of the shear stress is
1950). Hence, by slowly rotating the film in the flow, a minimum strongly dependent upon the Reynolds and Taylor numbers and their
in voltage is noticeable. To isolate this direction with reasonable ratio. This plot also indicates regions just before the entrance plane
accuracy, the probe was rotated to an identical larger voltage on (2n< 0.0) the shear stress, is in the upstream direction, marked by
each side of the minimum voltage. The angles for these two 9 angles greater than 90"for all cases except the one with the
voltages were then bisected to determine the minimum voltage largest R a a ratio (7.3). (Recall that 90" represents the rotor
angle. The probe was turned to this angle before the data direction, and 0" represents the downstream or axial direction.)
acquisition was initiated. Uncertainties on the shear stress Although it is not known exactly why there is a small degree of
measurements are 10%on the magnitude and 5" on the direction. backflow indicated by the inlet plenum shear stress directions,
The flow field measurements presented in this paper were research by Morrison et a1.(1994) may suggest that a backward
extracted from the data measured by Thames (1992) of which part flowing region is created upstream of the seal when the radially
was presented in Morrison, et al.( 1994). Descriptions of the laser accelerating flow is blocked from entering the clearance by large
anemometer system are included in those publications along with axial and azimuthal flows, and is thus deflected by the stator in the
uncertainty estimates. Those uncertainties are +0.001 for <IC*>and upstream direction.
0.015,0.058, and 0.013 for cU*>,<V*>, and <W*> respectively. For the largest R a a ratio, 24,000/3,300, the axial
All of the data presented have been non-dimensionalized. momentum is much larger than the tangential momentum, hence, the
The mean pressures were nondimensionalized by the pressure drop shear stress is oriented primarily in the downstream direction. There
from just at the seal entrance to the exit plane of the seal (values is a slight increase in 8 as the flow progresses through the seal due
given in Table I). This was selected based upon various seal to the increase in tangential momentum as the flow progresses
performance data available which non-dimensionalize well using through the seal. For the cases where the Taylor number
this parameter. The fluctuatingpressures were nondimensionalized contribution to the flow field development is greater, the direction
by cAPL The shear stresses were non-dimensionalizedby pW,/c of the mean shear stress vector changes significantly as the flow
based upon Stokes law of friction: enters the seal, from a direction mostly tangential in the direction of
the rotor, to one that is predominantly downstream. In the seal
interior the angles increase gradually as the rotor continues to
increase the azimuthal velocity by means of friction until the exit.
where the velocity gradients are represented by W,/c. These non- 0 increases into the exit plenum since the cross section increases

21 3
causing the axial velocity to decrease (and to be more evenly distributions on the stator wall as well as the resulting net forces and
distributed) compared to the azimuthal velocity. moments were calculated.
Once the flow enters the seal, the shear stress direction Axial and azimuthal forces were caiculated from the shear
becomes downstream with @=lo" for Re/Ta=3.6 and @=To for stress measurements while both the shear stress and pressure
RdTa=1.8. The 8 distribution for Re/Ta=3.6 is the same for both measurements were used to calculate radial or X-Y plane forces.
Reynolds and Taylor number combinations with 8 increasing from The axial or Z direction is considered positive in the downstream
10"at the inlet to 20" just upstream of the exit. Once the flow exits direction. The azimuthal or @ direction is considered positive in the
the seal, 8 increases to 40" as the axial velocity decreases due to the direction of seal rotation, where 0" represents the maximum
increased clearance. The 8 axial distribution for the Re/Ta=1.8 case clearance, - 180" to 0" represents the pressure side of the seal (the
is very similar with 8 being twice as large as the RefTa=3.6 case. clearance is decreasing in time from the maximum to the minimum),
Hence it appears that once the R f l a ratio is small enough the and 0" to 180" represents the suction side. The X axis points
product 8 ReITa produces a non-dimensional axial distribution. towards @=O" and the Y axis points towards @=90°. (See Figure
The non-dimensional mean pressure distribution is shown 19).
in Figure 8. The values of AP used to non-dimensionalize the mean Moments of the resultant pressure and radial shear stress
pressures are shown in Table I. The overall dimensional pressure forces are calculated about the axial midpoint of the seal (z/G-o.5),
drop increases with increasing Reynolds number and Taylor while moments created by the axial stresses are calculated about the
number. Doubling the Reynolds number results in a larger AP Z axis (seal center line). The angles of the moments are defined in
increase than doubling the Taylor number. The axial mean pressure the same manner as those of the forces.
distribution in Figure 8 shows a sudden decrease in pressure at the
seal inlet as the flow accelerates into the small clearance and FORCES, MOMENTS, AND DISTRIBUTED LOADS
encountersthe vena contractu. The percentage of overall pressure Figures 9- 12 represent the non-dimensional distributed
drop occumng at the seal entrance increases with increasing R f l a forces (per UL length) along the stator wall, and the directions of
ratio with the two Reynolds and Taylor number combinations these forces. Forces shown are in the axial, azimuthal and radial or
having a ratio of 3.6 producing the same non-dimensional axial X-Y planes.
pressure distribution however, the dimensional pressure drop for the Shear Stress Distributed Loads: The shear stress
high Reynolds number and Taylor number case is four times larger generated axial load distributions of Figure 9 show trends that are
than the low Reynolds and Taylor number case. From -0 to much like those of the mean shear stress plot, Figure 6. For the
0.25 the pressure remains constant for the two largest R f l a ratios Re=12,000, Ta=3,300 case there is a peak near the seal entrance
and increases for the smallest ratio. This result is due to the (0.027 ) before a gradual decay near ZL=O.2 (0.016). There is a
pressure recovery generated by the reduction in vena contractu size load decrease past W . 2 to a value of 0.014 by ZAA.6. before
over this region producing a larger cross-section for the axial flow rising gradually toward the exit plane where it peaks at 0.025 and
of fluid. This pressure recovery is offset by frictional headloss then decreases in the exit plenum (0.021). The resultant non-
generated by the wall shear stresses. For the R O a ratios of 3.6 and dimensional axial force for the 50% eccentricity sealis 0.0238. The
7.3, the pressure recovery and headloss are equal resulting in a resultant non-dimensional moment (due to the axial force
region of constant pressure. For the 1.8 R n a ratio, the pressure distribution and summed about the seal center line) and its direction
recpvery is larger resulting in a slight pressure increase. From in the radial plane is 0.4715 at -68".
ZKA.3 to the seal exit the pressure decreases at a relatively The Re=24,000, Ta=6,600 case exhibit almost identical
constant rate with the rate increasing with decreasing Refla ratio. behavior as the Re=12,000, Ta=3,300 case again showing the
The ratio of rotor speed to flow rate or TdRe appears to importance of R O a . There is a peak at the seal inlet (0.024) before
be a significant factor in the development of the annular seal flow a small decay to W . 2 5 (0.014). The remainder of the seal
field. This correlation implies that the influence of the rotor on the interior experiences relatively stable distributed forces (constant
flow depends strongly on the relative residence time of the flow in near 0.016) to near W . 7 . Here the value increases out to the exit
the seal clearance with respect to the rotor surface speed, W,. In (0.026). The resultant non-dimensional axial force is 0.0227. The
other words, the longer the length of time a given volume of fluid resultant non-dimensional moment and its direction in the radial
takes to pass through the seal, the more time the seal has to impart plane are 0.1753 at -64".
azimuthal momentum to that volume. When the low Reynolds number (12,000) is coupled with
This phenomenon is seen in many of the shear stress and the large Taylor number (6,600) the entrance and exit effects upon
pressure plots which show similar trends (in shape, but not the axial load distributionare significant. There is a large peak near
necessarily magnitude). The mean shear stress direction plot the seal entrance (0.057) which is about twice the value seen for the
(Figure 7) exhibits this correlation particularly well. The direction two cases already presented. This peak is followed by a rapid decay
curves for the cases with 3.6 RelTa fall nearly on top of each other near ZlLFO.2 to a value (0.020) almost equal to that observed
while the curve of the low RdTa ratio shows the largest angles and previously. the load remains fairly constant at 0.02 (just slightly
the curve of the large R f l a ratios show the smallest angles, once larger than the previous cases) until Z b O . 7 where it increases out
again suggesting that the ratio of the rotor speed to the flow rate is to the exit (0.045) reaching a value again almost twice the previous
proportional to the influence that the seal has on the flow field. cases. The resultant non-dimensional axial force is 0.0329. The
resultant non-dimensional moment and its direction are 0.3300 at -
FdRCE AND MOMENT CALCULATIONS 40".
A simple integration scheme was used to determine the The high Reynolds number (%000), low Taylor number
shear stress and pressure load distributions on the stator wall, (3,300) case develops the least axial load distribution of the four
calculated from the mean and phase averaged fluctuating shear cases. There is a very small increase near the seal entrance (0.019)
stresses and pressures. The axial, azimuthal and radial force followed by a decrease near ZYLs0.2 to 0.012. The seal interior
( W . 2 to 0.8) exhibits variations in the load distribution about a

214
value of 0.012. A slight increase at the exit plane (0.017) is at -88" and 0.264 at -98". The resultant non-dimensional moments
observed before decreasing into the exit plenum. The resultant non- and their angles are 0.0062 at -28" and 0.0066at -39" for the low
dimensional axial force is 0.01 80. The resultant non-dimensional and high Reynolds number cases respectively.
moment and its direction are 0.1 102 at 172". For the Re=12,000, Ta=6,600 case the pressure force
-F**Shear Stress Distributed Loads: The shear stress distribution shows a large increase in value from the inlet plenum
generated azimuthal load distribution magnitudes (Figure 10) follow to the entrance with a maximum of 1.5. The force distribution
the same trends as the axial loads with the high Reynolds number decreases until near ZlL=O.7 where the value rises from 0.25 to 0.9.
(24,000) and low Taylor number (3,300) case generating the least The force direction angle switches from the pressure side at the seal
load, the two cases with the same R f l a ratio (12,OOO/3,300 and entrance to the suction side at the exit although the switch is less
24,000/6,600) produce the same non-dimensionalized azimuthal severe than the other cases. The load angles range from -180"
force distribution which is larger than the first case, and the largest inside the inlet plenum, passes through 0" to 35" at the exit. The
non-dimensional forces being generated by the Re=l2,OOO, resultant non-dimensionalforce and its direction are 0.484 at -72".
Ta=6,600 case. The resultant non-dimensional moment and its angle are 0.0100 at
The Re=24,000, Ta=3,300 case does not possess any -49".
distinct trend other than the force increases through the seal The shear stress produced radial force distribution is not
beginning at less than 0.001 at the seal entrance. By Z k O . 2 5 the presented in this paper since it comprises less than 1% of the total
azimuthal force distribution has risen to a value of 0.0035. Along force and 6% of the moment. The data and a description are
the interior to W . 8 , the value falls to near 0.002. The value then presented in Winslow( 1994).
increases out past the exit to a value of 0.0055. The resultant non- Resultant Forces and Moments Tables I1 and I11
dimensional azimuthal load is 0.0033. provide listings of the total resultant forces, moments, and their
The two intermediate Refla cases (12,000/3,300 and respective directions for the various test cases and eccentricities.
24,000/6,600) exhibit the same shear stress azimuthal load The three force components are represented in Table 11: axial
distribution. After a relatively steep increase near the inlet to a (F*&, radial,)F,c( and azimuthal (FcR&. The radial component
maximum of 0.011 the azimuthal load distribution steadily comprises the sum of the shear stress and pressure components,
increases to 0.017 by W . 9 . At the exit the increasing trend while the axial and azimuthal components consist of the shear stress
becomes slightly steeper as value rise out into the exit plenum forces only. The direction for the total radial force is represented by
(0.024). The resultant non-dimensional azimuthal forces are 0.0166 QRm Two moments are represented in Table 111, the total moment
and 0.0145 for the low Reynolds number and high Reynolds due to the axial shear stresses calculated about the seal
number cases respectively. center line) and the total radial moment attributed to the radial shear
The low Reynolds number (12,000) and high Taylor and pressure stresses @I*-, calculated about 27JAl.5). The
number (6,600) case again possesses the largest non-dimensional directions for the total axial and radial moments are represented by
load distribution. After a sharp increase near the inlet to a @R.M and respectively.
maximum of 0.025, the azimuthal load decreases to 0.028 then
gradually increases to 0.04 at -0.7. At the exit this rising trend PHASE AVERAGED MEASUREMENTS
increases rapidly as the value rises to its maximum at the exit plane Figures 13-18 show contour plots of the phase averaged
(0.075). The resultant non-dimensional azimuthal loads is 0.0553. wall pressure, wall shear stress, axial velocity, azimuthal velocity,
E Pressure Distributed Loads and 0:The pressure radial velocity, and turbulence kinetic energy (IC) over one cycle
produced radial force distributions, Figure 1 1, show a large increase (rotation) of the seal along its entire length for Re=24,000 and
in value from the inlet plenum to the entrance with a maximum Ta=6,600. The measurement grids for these data have been
value of 0.76 for the Re=24,000, Ta=3,300 case. The force presented in figures 3-5. The velocity components and K were
decreasesuntil near ZIJA.7 where the value rises from 0.05 to 0.2. measured 0.16~ above the stator by Thames (1992) using a 3-D laser
The force direction angle (Figure 12) switches from the pressure Doppler anemometer system described in Morrison, et al. (1994)
side (-180" tom at the seal entrance to the suction side (0"to 180") and Thames (1992). The phase averaged wall pressure has
at the exit for all the cases. However, the high Re, low Ta seal relatively low amplitudes upstream of the seal. Upon entering the
switches sides in a direction opposite to that of the other cases. The seal, the phase averaged wall pressure increases in amplitude
angles change from-150" at ZkO.65, passes through -180" to 90" dramatically. The highest value (16) occurs from 30 to 50% of the
at W . 8 , a change of 120". The resultant non-dimensional force cycle on the pressure side of the seal and decreases to the minimum
and its direction are 0.212 at -139". The resultant non-dimensional value (-16) from 70 to 100% of a cycle on the suction side. The
moment and its angle are O.Oo40 at -28". locations of the maximum and minimum pressures remain the same
The pressure force distributions and load angles for the over the first half of the seal length with the magnitudes decreasing
ReJTa cases 12,000/3,300 and 24,000/6,600 are almost identical. from 16 to 6 by W . 5 . As the flow continues through the seal,
There is a large increase in value from the inlet plenum to the the amplitude of the phase averaged wall pressure decreases
entrance with maximum values reaching 0.96 and 1.1 for the low becoming more uniform around the seal with the pressure and
and high Reynolds numbers respectively. The load distribution suction regions switching near the seal exit.
decreases at a constant rate from zld-0.l to 0.65 reaching The phase averaged wall shear stress is shifted
minimums near 0.1. The values then rise from 0.1.to 0.45 at the approximately 114 of a cycle from the wall pressures with the
exit. The force direction angle switches from the pressure side at maximum (24) occurring at the seal entrance from 50% to 80%of
the seal entrance to the suction side at the exit. The switch a cycle and the minimum (-12) occurring from 0 to 30%. The
progresses from -1 10" at ZkO.65, passing through 0" and stopping positive value is located between the high and low pressure
at 55" at W . 8 . This progression is in the opposite direction locations with the pressure decreasing in the direction of the rotor
compared to the high Reynolds number, low Taylor number case. motion. This is a region of low azimuthal velocity with large
The resultant non-dimensional forces and their directions are 0.252 azimuthal and axial gradients in the axial velocity (Figures 15 &

215
16). The phase averaged wall shear stress becomes more uniform CONCLUSIONS
through the central portion of the seal with the maximum positive The stator wall pressure and shear stress distributions are
value of 10 and a minimum of -6.The amplitudes increase near the directly linked to the mean flow field and turbulence levels observed
s e d exit to values as large as at the inlet with the minimum shear 0.16c above the stator. The mean pressure decreases rapidly at the
stress shifting to 0% cycle and the maximum centering itself around seal inlet as the flow accelerates into the small channel. The rush of
50%of the cycle. fluid into the seal from the plenum causes a yenu contracta to form
The phase averaged axial velocity contours are presented immediately inside the seal inlet. The turbulence produced by the
in Figure 15. Bulk flow models predict maximum axial velocity at vena COnfnZCtQ and the large mean velocity gradients aSSOCiated
the location of maximum clearance (0%of a cycle). However, for with the sudden acceleration into the seal result in large levels of
this flow field the axial velocity has a magnitude of twice the bulk stator wall pressure, shear stress, and flow fluctuations. Contour
flow velocity and is located at 60-70% of a cycle in a region where plots of the phase averaged flow field show that at the seal inlet the
the phase averaged pressures are transitioning from maximum to pressure is high on the pressure side of the seal and low on the
minimum and have a value near zero. The minimum axial velocity suction side with the shear stress being approximately 90" out of
in the inlet is on the "pressure side" of the seal from 30-40% of a phase. As the flow progresses through the seal the level of pressure
cycle in a high pressure region but not at the maximum pressure and shear stress fluctuationsdecrease due to the turbulence levels in
location. As the flow progresses through the seal, the axial velocity the seal decreasing and the axial and tangential mean velocities
becomes almost uniform at z/L=0.5. However, the axial flow becoming more uniform around the seal circumference resulting in
begins to maximize on the "pressure side" of the seal around 30% smaller velocity gradients. At the seal exit the axial velocity has
of a cycle reaching a maximum for this azimuthal region at the seal become largest on the pressure side of the seal as compared to the
exit. Likewise, a region of minimum axial velocity is located on the suction side at the seal inlet. This results in the phase averaged
"suction side" of the seal. An explanation for this behavior is the pressure actually becoming negative on the pressure side and
flow will preferentially accelerate towards both the largest clearance positive on the suction side at the seal exit. It is the axial locations
as happens for a statically eccentric seal and for the minimum of maximum azimuthal variation in <U*>which corresponds to the
pressure. However, for this whirling case, the large minimum areas of maximum e*> and -33variation. Overall, the flow field
pressure is located on the suction side near the minimum clearance. is very complex with significant interplay between the mean
This low pressure causes the flow to deviate from the maximum velocity, turbulence, wall pressure, and wall shear stress. A simple
clearance location but not entirely into the low pressure region. assumption of no azimuthal variance of pressure or shear stress is
This is due to the small clearance present at the minimum pressure shown to be invalid as is the assumption that the axial velocity will
location not being able to accept all of the fluid. Thus the bulk of be maximum at the maximurn clearance location. This makes the
the leakage occurs between the two locations. modeling of the flow field inside whirling annular seals a very
Phase averaged azimuthal velocities (Figure 16) exhibit difficult task.
different behavior from the axial velocities. The minimum at the For all cases the mean pressure decreases rapidly at the
seal entrance occurs at the maximum clearance (0%) and the seal entrance plane as a large volume of fluid accelerates from the
maximum occurs just before the minimum clearance passes by at entrance plenum into the small clearance, causing a venu confracta
40% of a cycle. This does follows intuition since the minimum to form immediately inside the seal inlet. The turbulence produced
azimuthal velocities occur when the rotor is farthest away from the by the vena contracta and the large axial and azimuthal velocity
stator. There is a slight migration of the minimum value locations gradients associated with the sudden acceleration into the seal,
as the flow progresses downstream accompanied by an ever result in large levels of stator wall pressure and shear stress
increasing average value in <W*>. At the exit plane, the minimum fluctuations. At the seal entrances these shear stresses are typically
value is located around 15% of a cycle and the maximum azimuthal about 90" (lagging) out of phase with the pressures; this pattern
velocity is located near 60%. The maximum values of <W*> tend does not always hold downstream, but is highly dependent on the
to follow the locations of minimum <U*> throughout the seal. This test conditions.
is due to the lower values of <U*> resulting in longer residence The direction of the mean shear stress varies for the
times within the seal allowing the rotor induced tangential shear different flow cases. The high rotor speed turns the angle more in
stresses to cause increased azimuthal velocities. the direction of the seal rotation whereas the high flow rate tends to
Figure 17 presents the phase averaged radial velocities. turn the direction more downstream through the seal.
Other than the large values at the seal inlet caused by the flow A strong correlation between the ratio of the Reynolds
rushing into the seal, the radial velocities are very small. The phase number and the Taylor number is seen to exist as the pressure and
averaged turbulence kinetic energy, a*>, in Figure 18 maximize at shear stress characteristicsbehave similarly for test conditions with
the seal inlet where the axial velocity is maximum, the azimuthal identical ratios. This parameter, Rdl'a, apparently accounts for the
pressure variation is largest, and the vena contracts is present. This effects of rotor speed and flow rate on the annular seal environment
is due to the large velocity gradients resulting in lots of shearing that are encountered during seal tests.
effects generating vorticity and turbulence. The levels drop rapidly The flow field in a whirling annular seal tends to induce
once the flow passes over the vena contructa and both the axial and forces and moments on the stator wall (and hence the rotor) that are
azimuthal velocity gradients begin to decrease. <K*> becomes significant for the high eccentricity case studied, and can create
relatively uniform at the center of the seal. By the exit of the seal complications when operating these seals in turbomachines.
a*> has reached a minimum centered around 0% and has a local Pressure and shear stress induced forces and moments are shown to
maximum at 40% of a cycle. This increase in turbulence is due to exist on the stator that can push the seal away from areas of high
the axial velocity reestablishing peaks near the exit resulting in local pressure, and can tilt the seal away from alignment due to regions
velocity gradients. In fact CK*> maximized between the minimum of high and low pressure and shear Stresses that cause force
and maximum values of <U*>at the exit plane. imbalances around the seal. Such occurrences can lead to

216
undesirable contact of the seal on the stator wall, or to instabilities Master of Science Thesis, Texas A&M University, College Station,
in seal performance. TX, 77843.
In summary, the annular seal flow field is very complex
with significant interaction between the mean velocities, turbulence,
wall pressures, wall shear stresses, and forces. This makes the
modeling of the flow field inside whirling annular seals a very
difficult task.

ACKNOWLEDGEMENTS
This research was supported by the NASA Lewis
Research Center,the Turbomachinery Research Consortium at Texas
A&M University, and the Exxon Education Foundation.

REFERENCES

Allaire, P.E. Lee, C.C., and Gunter, E.J., 1978,"Dynamics of Short

Eccentric Plain Seals with High Axial Reynolds Numbers," Journal
ofspacecraft and Rockers, Vol. 15, pp. 341-347.

Chen, W.C., and Jackson, E.D., 1987, "A General Theory for
Eccentric and Misalignment Effects in High-pressure Annular
Seals," ASLE Transactions, Vol. 30,pp. 293-301.

Hanratty, T.J., and Campbell, J.A., 1983,"Measurement of Wall

Shear Stress," Fluid Mechanics Measurements, Hemisphere
Publishing Corporation, pp. 559-611.

Haritonidis, J.H., 1989,"The Measurement of Wall Shear Stress,"

Advances in Fluid Mechanics Measurements, Lecture Notes in
Engineering, Springer-Verlag, pp. 229-261.

Ludwieg, H., 1950,"Instrument for Measuring the Wall Shearing

Stress of Turbulent Boundary Layers," NACA TM 1284.

Morrison, G.L., DeOtte, R.E., and Thames, H.D., "Experirncntal

Study of the Flow Field Inside a Whirling Annular Seal," STLE
Tribology Transactions, Volume 37,1994,pp. 425-429.

Olivero-Bally, P., Forestier, B.E., Focquenoy, E., and Olivero, P.,

1993, "Wall-Pressure Fluctuations in Natural and Manipulated
Turbulent Boundary Layers in Air and Water," Flow Noise
Modeling, Measurement, and Control, ASME Publication, Winter
Annual Meeting, New Orleans, La., pp. 63-74.

Pessoni, D.H., 1974, "An Experimental Investigation into the

Effects of Wall Heat Flux on the Turbulence Structure of
Developing Boundary Layers at Moderately High Reynolds
Numbers," W.D. Dissertation, Mechanical and Industrial
Engineering Department, University of Illinois, Urbana.

Thames, H.D., 1992,"Mean Flow and Turbulence Characteristics

in Whirling A M U ~Seals," Master's Thesis, Texas A&M
University, College Station, TX, 77843.

Van Pragenau, G.L., 1990, "Damping Bearings for

Turbomachinery," NASA T P 3092, Vol. 11, Advanced Earth-to-
Orbit Propulsion Technology Conference,Huntsville, Ala., pp. 155-
162.

Winslow, Robert W., 1994."Dynamic Pressure and Shear Stress

Measurements on the Stator Wall of Whirling Annular Seals,"

217
Table I. Non-dimensionalizing parameters.
rh

12,000 3,300 20.6 3.7 15.5

12,000 6,600 32.8 3.7 30.9 Y (90")

1 24,000 I
124,000
3,300
6,600
I
1:
59.8 I 7.4

7.4
1 15.5

30.9

I 1 1 1 1:!21 1
APcL pWs,/c APcL
PWSdC
1 12,000 I 3,300 I 0.74 I 9.52 I 0.013

12,000 6,600 1.17 0.016

24,000 3,300 2.13 9.52 0.0044

24,000 6,600 2.89 0.0066

Table II. Summary of total non-dimensional forces and

(- 90.)
directions.

Figure 20. The directions of a set of resultant force and

moment vectors on the stator wall (at z/G-o.5), and the
X,Y, and Z axes are defined.

Table III. Summary of total non-dimensional moments and

directions.

218
 @
0 4 0 nn h.Pinhote Top
Texas ABM Seal Rig Composite Drawing
0-Ring Groove
P r e s s u r e Transducer

- 1-1- II--- Meosurenent

Plug Hount;ng
Surface Hole
LBrass Plug

Figure 2. Pressure transducer mounting block.

YLLLl

Figure 1. Seal test facility.

Pressure Measurement Grid Shear Stress Measurement Grid Flow Measurement Grid
,Re=24,000 Ta=6,600 e=50% LRe=24,000 Ta=6,600 e=50% Ta4.600 e S O %

1.o 1 .o 1.o

0.8
zn m zn
0.6
0.5 0.5
0.4

02

0.0 0.0 0.0

0.0 0.5 1 .o 0.0 0.5 1.o 0.0 0.5 1.o

Fmtion of Cycle Fraction of Cycle Frpction of Cycle
Figure 3. Phase averaged wall pressure Figure 4. Phase averaged wall shear stress Figure 5. Phase averaged velocity field
measurement grid. measurement grid. measurement grid.

Mean Shear Stress. *50% Mean Pressure, *SO%

Mean Shear Stress Direction, d 0 %

-A- llpumOT.Ium 140 -l

40-
120 -
30- 100 - 0.6
E -
8 so- E
F t P
20 - 60-
de4
6 40- 0.0
10 - -
0
1 . , u [ . [ . l . [ q , . [ r i
20

0-
-03

-0.4 J
-0.4 -03 0.0 0 2 0.4 0.6 O S 1.0 12 -20 . I . , . I . , * 1 . I . 1 I 1 -0.4 -03 0.0 0 2 0.4 0.6 O S 1.0 1.2
UL UL
Figure 6. Non-dimensional, mean shear
stress profiles for the 50% eccentricity
Seal.
 +
-, Re = 12,000 Ta = 3,300 ,+Re-
= 12,000 Ta = 3,300
Re = 12,000 Ta = 6,600 Re=12,000 Ta=6,600
Re = 24,000 Ta = 3,300 Re = 24,OOO T a = 3,300
-It- Re = 24,000 Ta = 6,600 Re = 24,000 Ta = 6,600

-8 3 0.08 7

ol Ti a

f
Y P
a O - 7 .
I
L
P &
v)
'eVI .I

Y
0.06-
a 4

w
P0 3
I

0.02 5
0
0.04 -
.E
I

-
2 I Y
L
Q
c" 0.00
-0.4 0.0 0.4 0.8 1.2 V
J
-0.4 0.0 0.4 0.8 1.2
zn UL
Figure 9. Shear stress axial load distribution. Figure 10. Shear stress azimuthal load distribution.

+
, Re = 12,000 T a = 3,300
Re = 12,000 Ta = 3 9 0
Re = 12,000 Ta = 6,600
Re = 12,000 Ta = 6,600
Re = 24,000 Ta = 3,300
Re=24,000 Ta=3,300
Re = 24,000 Ta = 6,600
Re=24,000 Ta=6,600
1.6 i
-
c)
a
180

Bv) 1.2
-
i? I

0.0 I -180
-0.4 0.0 0.4 0.8 1.2 -0.4 0.0 0.4 0.8 1.2
z/L z/L
Figure 11. Pressure radial load distribution. Figure 12. Pressure radial load angle of application.

220
 e A v e ~ e Pressure
d Phase A v e q e d Walt Shear Stress
,Re=24,000Tss;6,640 e&#% .a

CP, CP,
10 I .O 10
16.0 24.0 u
12.11 20.0 ia
0.8
9.6 16.0 08 1.6
UL A4 UL 12.0 UL IA
3.2 0.6 11.0 06 13
05 0.0 4.0 18
4.2 0.4 0.0 04 oa
d.0 0.6
6.1
4.0 0.4
-9.6 0.2 02
-12.0 03
-12A
-16.0 u.0
-16.0 00
00 00

-02
0.0 0.5 10 0.0
Fraction of Cycle
Figure 13. Phase averaged stator wall
pressure. stress.

Phe ial Vel0

0 &SO%

CWIS 4-2.
10 10 10
030 0.10 0.12
0.18 0.08 all
08 0.16 08 0.06 0.8 o. io
25 0.16 UL 0.04 m 0.09
06 0.12 06 0.02 06 8.08
0.10 -0.00 0.01
04 0.011 -0.02 04 a06
04
046 -0.04 0.05
8.04 -0.06 0.04
02 02 02
0.02 -0.08 0.03
0.00 4.10 0.02
00 00 00

radial
ve the uhon measured 0.16~
stator.

22 1
P&W CRYOGENIC FLUID-FILM BEARING AND SEAL TECHNOLOGY DEVELOPMENT
AND IMPLEMENTATION

Philip C. Pelfrey
Pratt & Whitney
West Palm Beach, Florida

This presentation will summarize Pratt & Whitney's past, present, and future
activities toward cryogenic fluid-film bearing and seal technology development
and implementation. The three major areas of focus for this technology are
analytical models and design tools, component testing, and technology
implementation. The analytical models and design tools area will include a
summary of current tools along with an overview of P&Ws new full 3-D Navier-
Stokes solution for hydrostatic bearings, HYDROB3D. P&Ws comprehensive
component test program, including teaming with the Air Force Phillips
Laboratory, NASA's Marshall Space Flight Center, and Carrier Corporation, will
be outlined. Component test programs consisting of material development and
testing, surface pattems/roughness, pocket and orifice geometry variations, and
static and dynamic performance of both journal and thrust bearings will be
summarized. Finally, the technology implementation area will show the benefits
and plans for P&W to incorporate this technology into products.

Overview

P&W's past, present, and future activities toward cryogenic

fluid-film bearing and seal technology development and
implementation. The three major areas of focus are:

analytical models and design tools

component testing

technology implementation

223
Analytical Models & Design Tools Past -
Reddecliff & Vohr (PN0317 & PN0345 - circa 1967)
- static and dynamic performance for hydrostatic journal
bearings (reynolds equation, finite difference)
- accounted for turbulence, compressibility, and inertia
- convergence problems at higher ecc. and clearances
HYDROB (Kocur)
- static and dynamic performance of hydrostatic journal
and thrust bearings with or without grooves (reynolds
equation, finite element)
- better turbulence, compressibility, and inertia modeling
- no convergence problems
HYDROBEART (San Andres)
- static and dynamic performance for hydrostaticjournal brgs
- Navier-Stokes, 2-D bulk-flow (finite difference)
- barotropic properties, thermohydrodynamic analysis

Analytical Models & Design Took - Present

HYDROSEAET (San Andres)

- adds multiple pad capabilities
- includes plain journal bearings
- calculates moment coefficients
- non-symmetrical vent pressures and circumferential
pressure gradients
- non-uniform pocket sizes and orifice diameters

RSR Software Library

- includes 2-D bulk-flow (incompressible) for thrust
bearings wavier-Stokes)

Note: Still use HYDROB for compressible thrust bearings

224
Analytical Models & Design Tools - Future

HYDROFLEX & HYDROTRAN (San Andres)

- adds transient analysis
- adds compliant pad capability
- 2-D bulk-flow

MTISealCodes

HYDROB3D (Braun/Dzodzo)
- full solution of 3-D Navier-Stokes equations
(finite difference/finite volume)
-journal and thrust bearing analysis
- accounts for compressibility,turbulence, and inertia
- does not assume uniform pocket pressure

Parallel Processing

500 engineering workstations working in parallel over our

computer network

Equivalent power of 50 Cray X M P super computers

(valued at 1.2 billion dollars)

500 workstations cost us only 8.5 million dollars

Overnight results to solutions which previously took weeks

225
Q
I
c3

226
 Sample 3-0 Hydrostatic Bearing Results

___-
3D) )I Print 11 15Apr
- 3dpl236.

'..
'.
I

.. * ..,

f
, .

. .

-" ..

227
Comparison of Results to Flow Visualization

228
Component Testing

Liquid hydrogen and HFC R134a material testing

(Phillips Laboratory/ARPA TRP)

Liquid oxygen material testing (NASA/MSFC)

Performance testing (P&W/ARPA TRP)

ARFATRP
- Dual-Use Hydrostatic Bearing Technology Program
- P&W (Lead), Phillips Lab, and Carrier Corporation
- Four year, 15 million dollar program
- Under contract with NASAIMSFC

229
230
Liquid Hydrogen Bearing

Liquid Oxygen Bearing

23 1
232
Technology Implementation

A rocket engine utilizing fluid-film bearing supported

turbopumps has the ultimate effect of lowering launch
costs as shown in the IHPRPT Vision:

Payoffs: Reduced Vehicle Cost & Weight

Increased Payload
350

300

250

Propellant Le’unch 200

Flowrate cost 0%
Engine (Millions) ,60
Weight
100

6o

0
0 10,oOO 2O.OOO 3O.OOO 40.000 a0.
m 8o.OOO
0.70 0.75 b.80 0:s 0.90 0.95 1.00
Vacuum Isp Low Earth Orbit - Payload (Pounds)
Maximum Theoretical Isp

233
Technology lmplementafion

Fluid-film bearings enable smallerllighter turbopumps

through:

No bearing DN limitation
No sub-critical rotor limitations

I------------ I

&=% Rolling Element

-/Subcriiicil

f lydrostatic / Siil)critical

I iyclrostatic / Sepercrilical

234
Technology implementation

Feature REB Turbooump FFB 'hrbopumr,

# unique parts 342 24

# total parts 1,960 65

Overall size (inches) 42 10

Weight (lbs) 1,000 90

Operating clearances 10's of mils mils

Speed (rpm) 35,000 much higher

Impellers shrouded unshrouded

Summary

Technology development activities at Pratt & Whitney are

focused on code development, component testing, and
technology implementation.

Analytical models and design tools are merging due to

parallel processing

Technology development is focused on end-product.

235
INTERSTAGE DISK-CAVITY/BRUSH SEAL NUMENCAL FLOW VISUALIZATION STUDY

V.V. Kudriavtsev and M.J. Braun

Department of Mechanical Engineering
The University of Akron
Akron, Ohio
5
35+/5-

ABSTRACT
This paper presents results and model developments that are directed
towards the computational fluid dynamics simulation of the secondary
flow system in a gas turbine engine. Numerical flow visualization results
of the 2-Daxisymmetric rotational fluid flow (Re=105) in a generic cross
section of an interstage turbine cavity are displayed. The core flow is
driven by an imposed pressure drop along the vane row of the main flow
path. Three interconnected disk cavities separated by a brush seal are
located along the secondary flow path. A new computational algorithm
was developed in order to predict the flow patterns and leakages for
different seal and cavity configurations. The code is based on the
numerical solution of the transient laminar Navier-Stokes equations,
written in primitive variables and approximated on a nonuniform
rectangular collocated grid. The program uses a mass and momentum
conservative formulation as well as a set of boundary conditions for
pressure and conservation of mass. The pressure solution uses the
Poisson equation under a direct implementation procedure.

OBJECTIVE

The main purpose of this study is to present an integrated brush

seaYdisk cavity flow model to demonstrate the effectiveness of this
approach and to emphasize the necessityfor further work in this
direction

237
INTRODUCTION

Recently, there has been a significant increase in the studies

concerning the detailed physics of rotational flows in disk cavities,
Przekwas et al., (1994), Virr et al., (1993), Athavale et al., (1993).

One of the main motivations for studying this problem is its direct
application t o gas turbine secondary flow systems management, in
particular the rim and interstage cavities.

Improved efficiency of the gas turbine performance has been

directly traced to improvements in the secondary flow path system,
which can be simulated both through improved experimental, and
novel computational methodologies.
A problem closely related t o the current study is the flow in a
cavity with rotating walls. This case has been extensively studied by
Chew(1985), Farthing et al., (1992) and Ong and Owen(1991).

Chew( 1985) treated some fundamental aspects of the heat transfer

in centrifugaly driven, steady-state free convection flow in a
cylindrical cavity with a non-uniform disk-temperature distribution.
In numerical and experimental studies performed by Owen and his
associates, a rotating cavity with axial cooling air throughflow was
used for the determination of heat transfer(0ng and Owen, 1991),
and flow structures(Farthing et a!., 1992).

Owen and Onur(1983) observed that a t certain rotational velocities

a non-axisymmetric vortex breakdown could occur and circulation
inside the cavity becomes weaker as the Rossby number is reduced.
The effect of the mainstream flow on the cavity flow has been
studied numerically by Vaughan and Turner(1987). They showed
that for certain conditions, a laminar 3-D solution gives a
sinusoidally varying ingress-egress cycle in the rim seal area.

238
KO et al., (1993) studied the axisymmetric subproblem of a
generic rotor-stator cavity. This work illuminated the presence of
the main flow induced recirculation zone in the cavity gap area.

Athavale et al., (1993) performed numerical calculations t o predict

the rotational cavity flows with long clearance sections. The
appearance of Taylor vortices was observed for supercritical
rotational Reynolds numbers.

In the secondary flow systems, cavities are usually separated from

the main stream, and each other by different types of seals, mostly
labyrinths and honeycombs. There has been a number of previous
studies that showed a marked improvement in sealing efficiency
when the labyrinth seal was replaced by a brush seal,

Carfile et ai., (1993) and Chupp and Dowler (1993).

Unfortunately, t o date, brush seal theory has not been developed

extensively, and a reliable numerical model that would allow
widespread parametric design is still missing. Very few of the
previous studies of this type of seal have involved theoretical, and
numerical work. Most of the existing studies are experimental in
nature. Chupp, et. al., (1991,1994) and Hendricks et. al., (1991)
pioneered a simple theoretical leakage flow model through the
brush seal.

However, simple models can not fully account for the effects of
changes in the details of the brush geometries. The only available
numerical simulations that analyze brush seal flows based on the
composition of the brush are those of Braun and Kudriavtsev(l992,
1993a,b), and Kudriavtsev (1993). In these works the brush seal
was simulated as a segment of densely packed cylinders, and the
Navier-Stokes equations were directly applied t o the flow upstream,

239
in-between, and downstream of these cylinders, Fig. 2.

These studies demonstrated that although fluid flow through the

brush can be successfully 'micro-modeled', such direct brush
modeling can be prohibitively expensive computationally,
Kudriavtsev and Braun (1994). These authors searched for
improved computational methodologies which eventually allowed
the development of several simplified numerical models, such as
brush partitioning(Kudriavtsev and Braun, 1994), and brush
representation by a variable porosity porous body.

All studies reviewed here were concerned with the cases when
either the disk cavity, or the seal were analyzed separately.
Hendricks (Przekwas et al., (1994)) proposed that these two
components be integrated in one computational algorithm, thus
increasing the physical accuracy of the simulation.

240
BRUSH - POROUS BODY

The brush body can be modeled approximately as a saturated

porous medium using Darcy's formula

and/or the Brinkman assumptions (Burmeister, 1993)

where e is the porosity of the porous media, k is the permeability,

U is the speed, and P is the pressure .

In the present study this approach is used t o investigate numerically

the flow through the brush.

The main purpose nf lbk studv is

t o present an integrated brush seal/disk cavity flow model,

to demonstrate the effectiveness of this approach, and t o

emphasize the necessity for further work in this direction.

24 1
PROBLEM FORMULATION AND DETAILS OF THE
SOLUTION PROCEDURE

The present study considers the fluid flow in a rotational interstage

cavity(1C). Such cavity consists of a combination of the main flow
path, front cavity with the rotating wall, stationary vane, and back
cavity with a rotating wall.

The cavities are separated by a brush seal affixed t o the lower axial
wall o f the vane, Fig. la.

The main flow is driven by a pressure drop imposed along the main
flow path. This pressure drop is also the driving cause for the
leakage flow through the brush seal that separates the high(front
cavity) and low(back cavity) pressure zones.

242
Governing Equations

The 3-0 Navier-Stokes equations(Schlichting( 1978)) are applied

here, while we are assuming rotational symmetry of the flow.

T o account for the pressure drop in the mainstream across the

vane row, we introduce a local flow resistance in the form of the
Darcy's law.

Thus, the permeability coefficient kX is calculated based on apriori

specified data regarding the pressure drop across the vane's blade
row, Eq. la.

Main Flow path and Cavity Space.

The governing equations for the fluid flow in the disk cavity and
along the mainstream flow path can be formulated in
dimensionless conservative form as

For such systems, the pressure equation was determined by

lvanushkin et al., (1980) using the momentum equation integration
technique first described by Welch et al., (1966) and then
exemplified by Roache( 1982) for Cartesian coordinates.

243
PRESSURE EQUATION:

were D is the dilation term

Zar+ " v+ a ua-

-D

6xy= 1 for the vane's blade row flow area, while tixy= 0 for
the rest of the computational domain.

The Bmsh Reaion.

For this portion of the flow additional modifications were

introduced into the main system of governing equations.

244
Momentum equations for the brush region

In a view of the modifications of the momentum equations the

pressure equation takes now the form

a2( uvl.2)
axar + 2 T )
r

245
Pressure equation for the brush renion

where D is a slightly modified dilation term

where kU are the three nondimensional permeability coefficients

(Darcy iuhbers) of the brush seal with respect t o the different
directions. On inspection of Eqs. 6 through 10 one notices that IC,
also called the Darcy number, plays quite a prominent role
principally characterizing the sealing capability of the brush.
Accuracy in its determination is important, and can be established
via three different methods.

246
DETERMINATION OF THE DARCY NUMBER

The Darcy number (k) , plays quite a prominent role principally

characterizing the sealing capability of the brush.
Accuracy in its determination is important, and can be established
via three different methods.

The Numerical Approach. This approach requires numerical

modeling of the Navier-Stokes equations through the brush
segment, Fig. 2, following the methodology described by Braun and
Kudriavtsev (l992,1993a,b), obtaining values of the average
pressure drops, and finally calculating permeability using Eq. la.

The Theoretical Approach.

Brush permeability can be calculated using Ergun's( 1952) formula
k= d2e3
A( b e ) *
. .

where A=150.
Kuwahara et al., (1994) performed numerical studies of a matrix of
10x10 inclined cylinders and squares. It was found that the
calculated permeability is almost identical to the (linear term)
variable k in Ergun's equations (Eqs.12, 13).
The authors determined that Ad53 when € E (0.08-0.84) for the
square rods, and A d 4 3 when e~(0.5-0.75),for the circular rods.
It was found that two-dimensional and three-dimensional models
lead to almost identical expressions for permeability( Kuwahara et.
al (1994)).

247
The format of the empirical Ergun’s expression( 1952) allows to
account also for the variation of a brush packing density under real
engine conditions, Le. variable pressure difference across the seal.

From Eq. 13 one can see that for constant bristle diameter (d), the
Darcian leakage velocity UD is a direct function of the average
porosity and pressure differential. The change in the pressure
differential is linked with the change in average porosity due t o the
deflection of the bristles under the pressure. Thus, we can account
for the potential movement of the bristles through the change in
porosity e.

The Experimental Approach.

Experimental leakage flow data can be used t o obtain brush seal

effective permeability using Eq. la. In this case dP/dX is an
experimental pressure gradient that is function of the brush
thickness. However, in this case we will obtain “effective” seal
permeability, since a large component of the leakage flow rate
comes through the clearance between the brush and rubrunner,
while the backing and front plates impose an additional flow
resistance. Materials presented in the paper of Carlile et al.,
(1993) provide enough information for estimation of a brush seal
permeability coefficient ku.

248
Boundary condiions, 1

In Fig. l a one can find a sketch of the computational domain with

specified boundary conditions. The main flow enters the
computational domain with zero radial velocity v, zero
circumferential velocity w( preswirl), and an axial velocity u=l. The
normalizing reference velocity is Uo=lOO m/s. The maximum
rotational velocity at the disk tip is w = l where again, the
normalizing velocity is Uo=lOO m/s. The rotational velocity along
the radial disk walls is a linear function of the radial coordinate

w(r)= r/ rm * wtop (where wtop=utop=l)

The rotational velocity at the shaft surface(bottom of a cavity) is

constant

where rm is cavity (disk) tip radius and rs ft is shaft radius. On

the surface of the stationary vane u=v=w=%! A t the main stream
outflow, one can impose g=g=O, which

accounts for developed flow conditions, and the axial velocity

gradient is calculated through the mass balance of the continuity
equation

The pressure boundary conditions on the stationary walls are

determined from the momentum equations, assuming that
u=v=w=O. Thus

249
Boundary Conditions, 2

Since at the interstage cavity inlet(Fig. l a ) u=l=const and

v=O, one can write

At the cavity exit(Fig. l a )

--
aP
ax--"=
aU rn' 1
-urn+=(
a2u
-yf+uzT
ar
1au a2u
+l?XT )
Thus the pressure conditions are calculated along all the boundaries
of the computational domain using Neumann type boundary
conditions.
In the gas turbine the flow along the main flowpath is driven
due to the pressure difference between the nozzle guide vane and
the turbine exhaust. For the interstage cavity under consideration
the most part of the pressure drop in the area where the vane
blade row is located. In the 2 0 formulation it is not possible to
account for this effect. We used a distributed flow resistance (in
the Darcy law format) t o account for this pressure drop along the
main flow path.

250
Solution Parameters and Computational Grid

For this study we had chosen constant permeability k p 6 . 0 x10-13.

One can obtain this value by using Ergun Equation with c=0.216
and d=0.075 mm.

A nonuniform grid 207x145 (Fig. IC)

was superimposed over the
computational domain of Fig. la.

In order to provide enhanced flow resolution the grid had an

increased density near the cavity walls and in the area adjacent t o
the brush seal.

Numerical experiments indicated that this grid size constituted the

minimum requirement for the accurate resolution of the complex
vortex separation processes that are taking place within the
interstage cavity.

The overall grid contained 23 separate regions along the axial

direction, and 14 such regions along the radial direction. Inside
each one of these regions the grid steps are uniform, but different
from region t o region.

Figure IC presents the computational mesh used for all the flow
simulations discussed in this paper. The sketch of the grid imposed
over the domain of the brush seal is shown in Fig. lb.

25 1
TABLE 1
Physical Characteristics and Parameters used in the Numerical
computations

Brush seal Darcy number

Rotational ratio R=wmax/Uin=l.

Main flow Reynolds number Rem=1.2 x 105

Vane flow Darcy number Da=kx= 10' 4

Computational domain dimensions:

Normalizing length Lo=l mm,
Nondimensional Axial size 1,=28.1
Nondimensional Radial size I,= 34.48
Nondimensional Shaft Radius =160.0

Nondimensional Tip Cavity Radius rm=183.48

Brush seal characteristics:

Brush permeabilities e= k/, k,= 0.O1

Permeability (k,) kErgun=6.0 x 10 -13

- 13
keT=
[ kexp calculated from Carlile, Hendricks & Yo er(1993)
5=0 lodata].

Bristle diameter d = 0.075 mm

Brush seal porosity c=O. 216

Brush seal clearance 1,=0.08

-- mm

252
RESULTS AND DISCUSSION, 1

In Figs. 3a, and 3b one can find the global flow patterns that occur in the interstage
cavities presented in Fig. 1. Figure 3a shows the flow when the side disk walls of the
cavity are rotating, while Fig. 3b presents the version of the same flow when the
velocity of the side walls is zero. These flows, whether they contain or not the disks
rotational effects, are generally extremely complicated even for the simplified geometry
used here. One can see four interacting vortices of different sizes in the front cavity,
three major vortices in the seal cavity, several more vortices and a shear layer in the
downstream cavity. The three vortices in the vane cavity are generated by the shear
layer that fills the smaller downstream cavity. A small gap induced recirculation
zone(GR2) similar to the one presented by KO et al., (1993) and Guo et al., (1994) is
present in the area where the vane cavity is bound by the main flow stream. Generation
of this zone provides the mechanism by which flow leaves the back cavity. A
comparison between the flow patterns of Fig. 3a and Fig. 3b indicates a slight increase
in the vortex structure complexity when rotation is not present; also the corner vortices
that are located near the rotating walls are much stronger. Thus, it appears that for the
geometry and flow parameters that were considered in this study, the rotational effects
have limited influence on the global interstage cavity flow structure. To further support
this conclusion we present in Fig. 3c the contour images of the flow field when the side
walls are rotating. The dark area zones are indicative of regions that have not been
penetrated by the effects of the rotating side disks, and thus the w-component of the
velocity is still close to zero. One can also see that since the mainstream w-velocity field
was set equal to zero(b0undary conditions), the main stream ingress into the interstage
cavity areas prevents the development of strong rotation effects. Note that rotational
components are strong only in the lower corners of the interstage cavity. It is also worth
mentioning that in the area surrounding the brush seal, the flow rotational component
is considerably diminished in value.
In Figs. 4a, and 4b one can see details that reveal the flow structure in the brush
seal area for the cases with rotating(Fig. 4a), and respectively non-rotating side
walls(Fig. 4b). One can again appreciate the complex flow formations that are taking
place in the area where the brush seal is located. Because of the brush high resistance to
flow passage, part of the secondary flow is reversing its direction towards the rotating
wall. This effect becomes especially strong when the front cavity wall is rotating. It can

253
RESULTS AND DISCUSSION, 2

be seen that the flow entering the area between the front plate and the shaft is skewed
and generally non-uniform. A major part of this flow is heading towards the clearance,
where the velocity magnitude is several times higher than elsewhere. Part of the flow
enters the brush(porous medium), and starts a slow motion along the brush walls. First
downward and inside, and then upward, outside of the brush backing plate(BP). From
Figs.4a, and 4b we can clearly see that the radial component of the brush flow is
comparable in size to the axial component. Thus the assumption that the radial
component is negligible and the brush flow can be represented by the flow pattern of an
axial cross-section, clearly does not hold. If one studies further the area near the BP and
the gap between the BP and the shaft, one finds very complex flow structures, probably
real, that could never be visualized if a set of simplified calculations(uniform inflow and
specified outflow boundary conditions), or bulk calculations were made. We refer in
particular to the secondary vortex that is splitting behind the BP, with part of it
entering the gap area, meeting the leakage flow and reversing its direction.
The powerful leakage flow stream, is instrumental in shaping the flow pattern
that fills the seal cavity, Figs. 5 and 6. This flow stream remains attached to the shaft
surface, whether the shaft rotates or not. Note that the cavities’ lateral walls are
attached to the shaft, and when the shaft rotates so do the side walls. Figures 5a, and
5b demonstrate different flow patterns that appear in the seal cavity as the process
progresses in time. The flow is completely past the transient developmental stage(from
rest to steady state). However, its shape keeps evolving in time, proving that while a
steady state is not existent, a quasi steady-state, somewhat periodic behavior has set in.
When the shaft is rotating, Fig. 5a, we can clearly observe three secondary vortices of
continuous and periodic changing flow patterns, which exchange momentum by means
of the appearance and disappearance of the s m d middle vortex. In the case of the
stationary shaft, Fig. 5b, one can distinguish two main secondary vortices. The second
one of the two exhibits an elongated backward moving finger that interposes between
the first main vortex and the leakage flow stream emerging from under the brush. The
rotation of the shaft seems to stimulate the development of the two main vortices, and
suppresses the minor vortices and their effects.
Extremely interesting is the flow formation that appears under high shaft
rotation, and small flow leakage from the brush. Figure 5c presents just such a case.
The decrease in -1 b b a r from a strong reduction in seal clearance or

254
RESULTS AND DISCUSSION, 3

decreased brush permeability. The disappearance of the strong axial flow along the shaft
allows the formation of Taylor vortices, between the inner rotating shaft and outer
stationary cylinder(vane). This form of fluid instability creates three almost equal
length vortices that fill the entire cavity between the vane and the rotating shaft. In
effect, the apparition of the Taylor formations indicate the domination of the rotational
effects upon the axial flow shear layer that contains the leakage fluid. Again, it is
proven that the integrated cavity/seal approach allows increased insight in the many
complicated flow features that are influencing the global flow field.
Figure 6 delivers a very intuitive insight into the effect of the rotating shaft on
the patterns and penetration of the rotational effect into the flow of the seal cavity.
The periodic nature of the flow can be observed again, together with the fact that the
zone of higher rotational component(white color) penetrates inside the cavity, increasing
its size, Figs. 6.1, 6.2, and 6.3, only to start returning to its initial size and location, as
the flow cycles (Figs. 6.4, 6.5, 6.6, and 6.7) .
Quantitative mapping of the pressure development in the front and back
cavities as well as in the seal cavity, Figs. 7a, 7b, and 7c support our qualitative
conclusion that flow in the interstage cavity is periodic in nature. Thus, Fig. 7a shows
variations of the pressure drop across the brush seal and in-between the front and back
cavities. One can clearly assess the periodic variation of the pressure that is associated
with a corresponding variation in the flow patterns. Figure 7b demonstrates the periodic
behavior of the maximum velocity in the brush seal clearance area and at the seal
cavity outflow boundary. In Fig. 7c the periodic nature of the rotational component in
the front cavity is on display. Corroborated, all these data allows the important
conclusion, that for high main flow Reynolds numbers(always the case in the gas
turbine flow path) the fluid flow has a periodic oscillatory nature that require time-
dependent terms in the governing equations. These terms are routinely omitted by some
researchers (KOet al., (1993), Iaicovidies and Launder(l991), Guo et al., (1994) ).

255
CONCLUSIONS

The present paper presents work-in-progress results and

model developments that are directed towards the computational
fluid dynamics simulation of the secondary flow system of the gas
turbine engine.

A computational algorithm and computer program (FLOCON) has

been developed t o model complex interstage gas turbine cavity flow
with the brush seal inserted between adjacent cavities. A
combination of two flow models, the first, for the brush seal(porous
media assumption), and the second, for the interstage
cavity(rotationa1 axisymmetric flow) was successfully implemented
into a single computational entity.

.The core flow is driven by an imposed pressure drop along the

vane row of the main flow path.

.Three interconnected disk cavities separated by a brush seal are

located along the secondary flow path.

.A new computational algorithm was developed in order to

predict the flow patterns and leakages for different seal and cavity
configurations.

.The code is based on the numerical solution of the transient

laminar Navier-Stokes equations, written in primitive variables and
approximated on a nonuniform rectangular collocated grid.

.The program uses a mass and momentum conservative

formulation as well as a set of boundary conditions for pressure
and conservation of mass. The pressure solution uses the Poisson
equation under a direct implementation procedure.

256
CONCLUSIONS

A time dependent Alternating Direction Irnplicit(AD1) method is

also used for the solution of all primitive variables.

This integrated computational approach allows the simulation of

the rotating disk cavity vortex flow periodic features, as well as
their impact on leakage flow and sealing effectiveness. In the
present study this approach is used to numerically investigate the
flow through the brush.

The proposed integrated approach allows solutions for complex flow

features that otherwise would not be accounted for. The results of
the present study indicate that for some conditions rotational
effects may be of minor effect for the brush seal flow and the
global flow pattern. It was found that the flow in the brush seal
penetrates inside the porous brush body and has stronger near
wall( backing and front plates) absolute velocities.
Also, the axial velocity component is comparable with the radial
velocity component due to lower flow resistance along the bristles.
This fact brings up the conclusion that brush seal modeling when
the brush is only represented by its axial (horizontal) cross-section
is not self supporting.
In general, it was found that flow in the cavity has periodic
nature, for both pressure, velocity and vortex formations.

257
REFERENCES, 1

Athavale, M.M., Przekwas, A.J., Hendricks, R.C., 1993, "Driven Cavity Simulation
of Turbomachinery Blade Flows with Vortex Control", AIAA-93-0390, 31st Aerospace
Sciences Meeting & Exhibit, Jan 11-14, 1993, Reno, NV
Braun, M.J., Kudriavtsev, V.V., 1993a, "A Numerical Simulation of a Brush Seal
Section and Some Experimental Results", Int. Gas Turbine and AeroEngine Congress,
Cincinnati, 1993, ASME Paper 93-GT-398, 1-12, (also Transactions of the ASME,
Journal of Turbomachinery, Vol. 117, No.1, Jan. 1995, pp. 190-202)
Braun, M.J., Kudriavtsev, V.V., 1993b, "Numerical Visualization of Flow Structures
in Dense Banks of Cylinders Located in a Channel", Third Int. Symposium on
Experimental and Numerical Flow Visualization, 1993 ASME Winter Annual Meeting,
New Orleans, Louisiana, Nov. 28 - Dec. 3, 1993/Eds. B. Khalighi, M.J. Braun, C.F.
Freitas, D.H.Fruman, p . 1-8, (accepted for the publication - ASME Journal of Fluids
Engineering, March-19957
Braun M.J., Kudriavtsev, V.V., 1992, "Experimental and Analytical Investigation of
Brush Seals", Seals Flow Code Development-92, NASA CP-10124, Proceedings of a
workshop held at NASA Lewis Research Center, August, 5-6, 1992, pp. 181-195.
Burmeister, L.C., 1993, "Convective Heat Transfer, John Wiley & Sons, pp.44-51.
Carlile, J.A., Hendricks, R.C., Yoder, D.A., 1993, "Brush Seal Leakage Performance
with Gaseous Working Fluids at Static and Low Rotor Speed Conditions", 1993, Journal
of Engineering for Gas Turbines and Power, Vol. 115, pp. 397-403
Chew, J.W., 1985, "Computation of Convective Laminar Flow in Rotating Cavities",
J. Fluid Mech., Vo1.153, p.339
Chupp, R.E., Dowler, C.A., "Performance Characteristics of Brush Seals for Limited-
Life Engines", 1993, Journal of Engineering for Gas Turbines and Power, Apr. 1993,
Vol. 115, pp. 390-396
Chupp, R.E., Holle, G.F., Dowler, C.A., 1991, "Simple Leakage Flow Model for
Brush Seals", AIAA Paper No. 91-1913
Chupp, R.E., Holle, G.F., 1994, "Generalizing Circular Brush Seal Leakage Through a
Randomly Distributed Bristle Bed", ASME Paper 94-GT-71, International Gas Turbine
and AeroEngine Congress & Exposition, The Hague, Netherlands, June 13-16,1994
&gun, S., 1952, "Fluid Flow Through Packed Columns", Chem. Eng. Prog., Vo1.48,
N0.2, pp. 89-94
Farthing, P.R.,Long, C.A., Owen, J.M., Pincombe J.R., 1992, "Rotating Cavity with
Axial Throughflow of Cooling Air: Flow Structure", ASME Journal of Turbomachinery,
VO~.114, pp. 237-246
Guo, Z., R.hode, D.L., Davis, F.M., 1994, "Computed Eccentricity Effects on Turbine
Rim Seals at Engine Conditions with a Mainstream", ASME94-GT-31, Int. Gas
Turbine and Aeroengine Congress & Exposition, The Hague, Netherlands, June 13-16,
1994

258
REFERENCES, 2

Hendricks, R.C., Schlumberger, J., Braun, M.,J., Choy, F.C., Mullen, R.L., 1991, "A
Bulk Flow Model of a Brush Seal System", ASME Paper No. 91 GT-325, 36th ASME
International Gas Turbine and Aeroengine Congress & Exposition, Orlando, FL, June 3-
6
Iacovides, H., Launder, B.E., 1991, "Parametric and Numerical Study of Fully
Developed Flow and Heat Transfer in Rotating Rectangular Ducts", Journal of
Turbomachinery, Jul. 1991, Vol. 113, pp. 331-338
Ivanushkin, S.G., Kim, L.V., Kondrashov, V.I., Tomilov, V.E., 1980, "Internal
Transient Problems of Convective Heat Transfer" , Tomsk University Publishers,
Tomsk, pp. 1-150 (in Russian)
KO, S.H.,Rhode, D.L., Guo, Z., 1993, "Computed Effects of Rim Seal Clearance and
Cavity Width on Thermal Distributions" , ASME Paper 93-GT-419
Kudriavtsev, V.V., 1993, "Numerical Analysis of the Transient Fluid Flows in the
Brush Seal Elements of the Aerospace Turbomachinery", Ph.D. Dissertation, Moscow
Aviation Institute (in Russian)
Kudriavtsev, V.V., Braun, M.J., 1994, "Advances in Brush Seal Numerical
Modeling" , SAE Aerospace Atlantic '94 Conference and Exposition, SAETP-941208,
/Advanced Dynamic Seals, Dayton, Ohio, pp.1-13 (Also AIAA Journal of Propulsion
and Power, to be published in 1996).
Kuwahara, F., Nakayama, A., Koyama, H., 1994, "Numerical Modelling of Heat and
Fluid Flow in a Porous Medium", Heat Transfer, 1994, V01.5.~Proceedings of the Tenth
International Heat Transfer Conference, Brighton, UK, Paper-ll-PP-18, pp. 309-314,
Ong, C.L., Owen, J.M., 1991, "Prediction of Heat Transfer in a Rotating Cavity with
a Radial Outflow", ASME Journal of Turbomachinery, Vol. 133, pp.115-122
Owen, J.M., Onur, H.S.,1983, "Convective Heat Transfer in a Rotating Cavity",
ASME Journal of Engineering for Power, v. 105, pp. 265-271
Przekwas, A.J., Athavale, M.M., Hendricks, R.C., 1994, "Progress in Advanced
Modeling of Turbine Engine Seal Flows", AIAA-94-2803, 30th Joint Propulsion
Conference
Roache, P.L., 1982, Computational Fluid Mechanics, Hermosa Publishers.
Schlichting, H., 1978, Boundary-layer Theory, N.Y., McGraw-Hill
Torrance, K.E., 1968, "Comparison of Finite-Difference Computations of Natural
Convection", Journal Research NBS: Mathematical Sciences, N72B, 281.
Vaughan, C.M., Turner, A.B., 1987, "Numerical Prediction of Axisymmetric Flow in a
Rotor-Stator System with an External Mainstream Flow", Proceedings 5th Int. Conf. on
Numerical Methods in Laminar and Turbulent Flows, Swansea, UK.
Virr, G.P., Chew, J.W., Coupland, J., 1993, "Application of Computational Fluid
Dynamics to Turbine Disk Cavities", ASME Paper No. 93-GT-89.
Welch, E.J., Harlow, F.H., Shannon, J.P., Daly, B.J., 1966, "The MAC Method: A
Computing Technique for Solving Viscous, Incompressible, Transient Fluid Flow
Problems Involving Free Surfaces", Technical Report, Los Alamos Scientific Laboratory,
LA-3425,UC-32, TID-4500

259
 T20ne with imposed
main flowpath flow resistance
- S
I

i outflow
.-
I I

/
rotating
front vane 'rotating
disk cavity disk
#Ivity'

/brushsea'
cavity

.
+ s'cal clearance
T**

number of nodes
in the grid interval

B) GRID INTERVALS AROUND THE BRUSH SEAL

FIGURE 1 (A,B,c). SIMULATED INTERSTAGE CAVITY

260
 190

185

180

175

170

165

160 I
0 10 20 30

C) COMPUTATIONAL GRID (209 ~ 1 4 4 )

F1GUR.E l(A,B,C). SIMULATED INTERSTAGE CAVITY

FIGURE 2 FLOW THROUGH THE SIMULATED BRUSH SEGMENT

LOCATED IN THE CHANNEL

26 1
 185

170

A) ROTATING CAVITY
165

icn
IV U
0

B)NONROTATING CAVITY

FIGURE 3 A$). FLOW PATTERNS IN THE INTERSTAGE SEAL C A W

262
 190

W
185
0.5
0.464286
0.428571
180

0.321429
1 0285714

0.178571
1

5
0

160 A
0 20 30

SIN ERST c

entering gap area

759 15
4 5 6 7 8 9 1 0 1 1 1 2 4 s 6 7 8 9 10 11 12

TA C TING CAVITY

FIG IN AREA U G THE AL

263
162 162

160 160 _- I , , , , , , ,

tn
.- 1s 20
I
i
162 162

160 160

162
162

160
160

162
162

160

162
:62

160
160
1 ” I , , . . ‘ I * ’
10 15 20

’ 62

,60

264
 =2

i . . . . , I . < . . * . . ,
70 15 20

162
SEAL
160 4 deaeaaeoftheKUVA
I . . . . I . . . . I ' . '
10 1s t 20

&rowingof the HRVA

10 s . 15 1 20

265
 A)PRESSTJRE DROPS ACROSS THE BRUSH SEAL (1) and
BlETWEEN THE FRONT AND BACK CAVITlES(2)

- h-4

B)AXZAL VELOCITY COMPONENT U

J
lmooE5

(1) BRUSH SEAL CLEARANCE

(2) SEAL C A W OUTFLOW

a0

-1El
;a
8d
69owE4 ammE4 1.oawEs

C) ROTATIONAL COMPONENT W (center of the h n t cavity)

FIGURE T(A,B,C) PERlODIC FEATURES OF THE FLUID FLOW

266
 TWO-DIMENSIONAL DYNAMIC SIMULATION OF A CONTINUOUS FOIL BEARING

M.J. Braun, F.K. Choy, M. Dzodzo, and J. Hsu

Department of Mechanical Engineering
The University of Akron
Akron, Ohio
s/7-37
3 q p z -9 .i

1 SCOPEOFWORK
In this paper, the two dimensional(radial and circumferential) transient Navier-Stokes equations
are used to solve the hydrodymmc problem in conjunction with the time dependent motion of the
journal, and the deformable, spring supported foil, Fig. 1A. The elastic deformation of the foil and its
supports are simulated by a finite element model. The timedependent Navier-Stokes formulation is used
to solve for the interaction between the fluid lubricant, the motion of the journal and the deformable foil
boundary. The steady state, the quasi-transient and the full transient dynamic simulation of the foil-fluid-
journal interaction are examined on a comparative basis. For the steadv state simulation, the fluid
lubricant pressures are evaluated for a particular journal position, by means of an iterative scheme until
convergence is achieved in both the fluid pressures and the corresponding foil deformation. For the
Quasi-transient case, the transient motion of the journal is calculated using a numerical integration
scheme for the velocity and displacement of the journal. The deformation of the foil is evaluated through
numerical iteration in feedback mode with the fluid film pressure generated by the journal motion until
convergence at every time step is achieved. For the full transient simulation, a parallel real-time
integration scheme is used to evaluate simultanausly the new journal position and the new deformed
shape of the foil at each time step. The pressure of the fluid lubricant is iterated jointly with the
correspondingjournal position and the deformed foil geometry until convergence is achieved. A variable
time-stepping Newmark-Beta[11 integration procedure is used to evaluate the transient dynamics at each
time step of the bearing(Choy et al.[2,3], Braun et al.[4]).

For simplicity in the numerical solution, an infinitely long bearing(tw0 dimensional in the radial and the
circumferentialdirections) is used for the purpose of parametric demonstration.

2 NOMENCLATURE

h characteristic length; h=RJlOOO

J Jacobian of the coordinate transformation; y’=y’(xJ)
F;Ie Matrix m e s s : foil and its support in global coordinates
Wf] Mass matrix of the foil
M, Massofthejournal
P dimensionless pressure; +p/p~(v/h)~
R, shaftradius
S sourceterm
t time
Ui dimensionless Cartesian velocity vector component;Ui=uihlv
Ui dimensionlesscovariant velocity components(normal to
xJ curvilinearcoordinate
XJ dimensionless general curvilinear coordinate; xJ=xJ/h
y!” displacement of the journal
yJ Cartesian coordinate

267
 dimensionlessCartesian coordinate; yJ=yJ/h
cofactor of the i-th row and j-th column of the Jacobian matrix
kinematic viscosity
density
dimensionlesstime; 7‘tv/h2
displacement of the foil

THE ANALYTICAL MODEL

The transient fluid lubricant behavior is modeled by the joint solution of the continuity equation

where 6 represents the cofactor of the i-th row and j-th column in the Jacobian matrix J of the
coordinate transformaton yi= y’(xJ); y’ is the reference Cartesian coordinate system. The finite volume
method based on the finite difference method was applied for the numerical simulation which uses
nonorthogonal coordinates(Fig. 1B) and a collocated grid. The SIMPLE procedure of Patankar and
Spalding[S] was used for solving the set of equations. The code developed by Dzodzo[6] has been used
after imposing modifications for the boundary conditions. The collocated control volume cells,
corresponding to the arbitrary coordinates x1,x2 and the reference Cartesian coordinate y’ ,y” system are
presented in Fig. IB. For the steady state and the quasi-transient cases, assuming the foil deformation
follows closely the effects of the j o d motion, the deformation of the foil is modeled by a combination
of bending elements supported by external linear springs, Fig. 1A. ”he foil structural model is developed
by installing the localized bending elements stiffness matrix into the global matrix. The deformation of
the foil is calculated by solving the force equations as

where SIgis the global &€ness matrix of the foil structure, and Fy1 and F,,z are the y’ and y’ direction
forces acting on the foil through the fluid lubricant. These forces are evaluated by integrating the fluid
pressure at the foil boundary, for the i-th location, as
ei
Fy;= ( 112) (Jp cos0 d8 + T’p cos8 d8 ) (4)
i-l @i
and
@i
F? = ( 112) (Jp sin0 dB + p sine de)
i-l ei

The transient motion of the journal in the y’ and y” directions can be represented with the following
eqUtiOnS
2n
y;=(imj)(0J P ~seae+fu1(t)) (6)
and
2n
yf = ( l/Mj)( J P sin 8 dB + fu2(t)
0
) (7)

268
In Eqs.6 and 7, f,l(t) and fu2(t) are the forces due to mass imbalance in the Y’and directions. The
accelerations yj and y! are then integrated numerically to evaluate the velocity and displacement of the
journal at each time steps. For the full-transient case, the transient motion of the foil is evaluated in
parallel with the transient motion of the journal.

Both foil and journal transient equations, Eqs. 6 and 7 and the foil Eq. 8 are solved simultaneously.
Taking into account the load forces and initial conditions of both the journal and the foil, the transient
motion of the journal and the foil were calculated independently by using the Newmark Beta integration
procedure.

4 DISCUSSION OF RESULTS

To demonstrate the above developed procedure, a 4in(10.16cm) diameter bearing was used as an
example. A thin circular foil of O.Olin(0.254mm) thickness is supported at 60 equally spaced locations
around the outer circumference, Fig. lA, by radial linear springs with a stiffness of
50,000l&/in(8.756*lO6N/m).The radial clearance between the foil and the journal is 0.003in
(0.0762mm) and the fluid lubricant has a dynamic viscosity of 1.26*1O4lb&n2 (8.68*lO”kg/s-m), and
density of 2.846*10-21b,,,/in3 (787.7kg/m3). A vertical load of 5001& was applied to the journal operating
at rotational speeds ranging from 10oO to 4000 rpm.

Figure 3 shows, for various operational speeds, the comparison between the steady state journal
equilibrium positions for a rigid journal bearing and the same dimension compliant foil bearing. Note
that in both cases, the equilibrium positions shift towards the bearing center as speed increases(Ch0y et
al.[2], Braun et al.[4]). However, for the foil bearing due to the deformation of the foil, Fig. 3, the
equilibrium position of the journal is shifted in the direction of the occurrence of the deformation. This
shift, appears to occur in order to maintain a certain minimal radial clearance, necessary to the build-up
of the self acting pressure wedge(similar to the pressure distribution mechanism in the rigid journal
bearing). Figures 4A and 4B depict the fluid film pressure distribution in the bearing clearance for both
the foil and rigid journal bearings. Note that in Fig. 4 4 for the foil bearing, when the journal speed
decreases to lOOOrpm, the positive pressure recedes circumferentially and achieves a higher peak
pressure than that of the rigid bearing, Fig. 4B. The explanation can be found only when one considers
Figs. 3 and 4 together. By inspection, one can see that as the lower speeds are characterized by steady
state shaft positions further away from the geometric center, the smaller minimum clearances exhibited
by the compliant foil bearing will generate the higher peaks observed in fig. 4A. Figure 5A presents the
stiffness characteristics of both the rigid journal and foil bearings at various operating speeds. Note that
at the lower operating speed of 1000 rpm, the cross-coupled stiffness, K.&$,I,,z) for the foil bearing is
substautially lower(40%) than that of the rigid bearing. As the speed increases, to 4000rpm, the two
curves run with almost parallel slopes, and thus thedifference between their magnitudes remains
relatively constant. However, their absolute magnitude has now increased to 1.2*1061b/in(2.1*108N/m).
Under these circumstances the reduction in the K&$J,.z) of the foil bearing represents only 10%
difference from that of the rigid bearing. At the same time, the direct stiffness, K n G l y l ) and
K,,@,,z,,z), remain relatively the same. This reduction in the cross coupling stiffness of the foil bearing
shows that this device offers improved stability when compared to the rigid bearing. For damping, Fig.
5B shows a reverse trend, as the foil bearing exhibits lower direct damping characteristics, (&(C$y~)
and C,&+p), than those of the rigid bearing. However, this trend is not as pronounced as the effects
observed in the stiffness characteristics of Fig. 5A. This phenomenon is further corroborated by the
stability characteristics exhibited by the logarithmic decrement graph shown in Fig. 6. A study of this
figure shows that the foil bearing provides a better stability at lower speed, and its advantage decreases
with respect to the rigid bearing, as speed increases.

269
In order to veri@ the linear steady state analysis performed in the previous example, a time transient
analysis was carried out for the same foil bearing. Figure 7A depicts &e ami-transient
simulation(with0ut the inertia effects of the foil) of the journal motion with a mass imbalance eccentricity
of h4.005 in. at a rotational speed of 1,000rpm. During the transient motion, the journal moves
quickly into its steady state elliptical orbit as predicted by the linear stability analysis, Fig. 6, and proves
the system to be highly stable. Figure 7B shows the journal motion for the same case using the full
transient simulation, i.e., including the foil inertia effects. Note that the full transient progresses quickly
into the steady state orbit, in thesame manner as the quasi-transient, but with a relatively smaller orbit
amplitude. The most likely explanation is, that due to the transient motion of the foil, a higher damping
effect is being generated in the system, which in turns,results in smaller vibration amplitudes.

In order to examine the nonlinearities of the foil bearing, the same analysis is performed for a journal
rotational speed of 4,0OOrpm, where the linear analysis predicts an unstable behavior. During this
simulation, the quasi-transient procedure fails to converge to a reasonable result as the calculated
deformed shape of the foil changes considerably from one time step to another in an irrational and
discontinuous manner. This phenomenon is possibly due to the fact that during the quasi-transient
analysis, the foil deformed position is iterated together with the fluid pressure, and forced to converge at
the same time. Such a procedure implicitly assumes that the foil will move instautly to its deformed
position following the motion of the journal and pressure changes without any delay. However, due to the
presence of the mechanical inertia of the fluid in the gap between the journal and the foil, and due to
foil's own mechanical inertia, such motion is practically impossible, especially at higher speeds. To
confirm this theory, the full transient analysis was performed at 4,000 rpm. Figure 8A depicts the orbit of
the journal motion and the corresponding foil deformation at consecutive time steps. The change in the
lubricant pressure curve during this transient motion is shown in Fig. 8B. Note that now, both the
resulting foil deformation and the corresponding pressure are continuous and follow in a logical manner.
Thus one may conclude from this exercise that, when higher speed andor nonlinearity behavior are
involved, a full transient simulation is required to provide an adequate prediction of the performance of
the system.

Figure 9 shows the journal motion for the full transient case at 4,000 rpm. Note that, due to the
instability of the system, the journal motion goes into a "limit cycle" type orbit. Note that the deformed
shape of the foil is smoothly continuous and even though we are witnessing a limit cycle behavior, the
clearance between the shaft and the foil shows that the system has overcome instability and it can
function without damage in this state. This outcome is probably due to the compliance of the foil. Due to
this large deformation, the converging and diverging zones of the fluid are substantially Merent than
those of a rigid journal bearing, and the existence of the large deformation in the foil further confirms
the necessity of a flexible foil transient model to accurately predict the performance of the system.

5 CONCLUSIONS

A numerical model to simulate the steady state and transient dynamics of a foil bearing was developed.
Results from the foil bearing analysis are compared to those from a rigid journal bearing. The
conclusions d
ram from this study are as follows:

1. The dynamics of the fluid film can be effectively simulated using the Navier-Stokes equations, and any
limitations concerning inertia effects have been eliminated. The procedure allows a natural calculation of
the damping coefficientsand is compatible with the dynamic formulation of the problem.
2. The foil bearing can provide better system stability than a rigid journal bearing.
3. The location of the journal equilibrium position can change substantially, on comparison with the
rigid bearing, due to the possible large deformation of the foil. The peak pressure in a foil bearing will be
higher than of its rigid counterpart for the lower velocities(1000rpm), while for the higher onesFigs. 41,

270
the pressure generating the load Canying capacity, and the pressure peaks will be lower and distributed
more evenly around the circumferenceof the bearing.
4. For cases without large foil deformation or journal eccentricity, the linear steady state analysis is
accurate, and sufficient.
5. When higher rotational speed andor nonlinearity are involved, a full transient simulation(foil-fluid-
rotor) is required in order to provide aq adequate prediction of the performance of the system. The quasi-
transient(with0ut foil inertia) or the linear analysis will not be able to provide an accurate simulation of
the event.

6 ACKNOWLEDGMENT

The work presented in this paper was supported in part by NASA Lewis Research Center, Cleveland,
Ohio under grant NAG3-1366. The continuous support of NASA Technical Project monitor, Mr.Jim
Walker, is greatly appreciated.

7 REFERENCES

1. N.M. Newmark, "A Method for Structural Dynam~cs",J. of the Engineering Mechanics Division,
Tran~.ASCE, EM3, July 1959, pp. 67-94.
-
2. F.K. Choy, M.J. Braun, and Y. Hu, "NonlinearEffects in a Plain Journal Bearing Part 1 : Analytical
Study" ASME Journal Of Tribology, Vol. 113, N0.3,pp.555-562, July 1991.
3. F.K. Choy, M.J. Braun, and Y. Hu,"Nonlinear Transient and Frequency Response Analysis of a
Hydtodynamic Journal Bearing", ASME Journal of Tribology, Vol. 114, No. 3, July 1992, pp. 448-454.
4. M.J. Bra- F.K. Choy, and Y. Hu, "Nonlinear Effects in a Plain Journal Bearing - Part 2 : Results"
ASME Journal OfTribOl~g~, Vol. 113, N0.3, pp.563-570, July 1991.
5. S.V. Patankar, and D.B. Spalding, "A Calculation Procedure for Heat, Mass and Momentum Transfer
in Three-Dimensional Parabolic Flows", Int. J. Heat Mass Transfer, Vol. 15, p. 1787.
6. M. Dzodzo, "Laminar Natural Convection in Some Enclosures of Arbitrary Cross Section", Ph'D
Thesis, Faculty of Mechanical Eng., UGversity of Belgrade

27 1
 y2L Yl
FIGURJ3 1. Analytical Model of the Foil Bearing System:
(A) The Foil Bearing; B) The Fluid Lubricant Model

272
 STEADY STATE ANALYSIS TRANSIENT ANALYSIS

I
Iteration of Fluid
Pressure and Foil
Deformation

Time=Time
+ AT
canvergence of
Fluid Pressure
I I and Foil Defomuation
L - ...-. . -. . , ..... _.. ---

J
I

f f
1
’
\
Quasi-Transient

FIGURE 2. Flow Chart for the Steady-State, Quasi-Transient,

and the Transient Numerical Simulations

273
2.204 i i i

2.202

2.200

2.198

2.196
2.196 2.198 2.200 2.202 2.204

FIGURE 3. Equilibrium Position of the Journal for the

Rigid and the Compliant Foil Bearings

274
 200 - I I

180 -
160 -

140 -
n

2 120 -
W

E! 100 -
E
1 8 0 -
E
60 -
40 -
2o
- @ I I
0

200

180

160

140
n

E 120
U
4000 xpm

E 100

y
v1
80
pc
60

40

20
I I I I I ! \J
0
0 60 120 180 240 300 360
CIRCUMFERENTIAL 0 (")

FIGURE 4.Prcssurc Profile with Various Operating Speeds:

(A) Compliant Foil Bearing; (B)Rigid B b g

275
 1000 2000 3000 4000
ROTATIONAL SPEED (rpm)

1000 2000 3000 4000

ROTATIONAL SPEED (rpm)

FIGURE 5. Bearing Characteristics for both the Rigid and the

Compliant Foil Bearings. (A) Sti€€ness; (B)Damping

276
 FOIL B.
/JOURNAL B.
0.5
0.0
1000 2000
3000 4000
ROTATIONAL SPEED (rpm)

FIGURE 6. Logarithmic Decrement of the Jouxnal Stability

for the Rigid and the Compliant Foil Bearings

2.204
IP 0
2.202

2.200

2.196
2.196 2.198 2.200 2.202 2.204

2.204 I 1 1 1

2.202

2.200

2.198

1 1 1
2.196
2.196 2.198 2.200 2.202 2.204

FIGURE 7. Transient Trajectories of the Journal Center at

1000 rpm. (A) Quasi-Trausimt; (B) Full Transient

277
 CONSQUENT DEFORMED FOIL SHAPE
AT TIME STEP 4580 , 4620, AND 4660

2.204
9 Y2 (in)
RPM
.005 IMBALANCE
4000

2.200

2.196 I I
2.196 2.200 2.204
4000 RPM
.005 IMBALANCE
DT = 1.E-05
I I I I I I I I

4580

210 240 270 300 330 360 3 9 0 420

DEGREE

FIGURE 8. Full Transient Model at 4000rpm. (A) Foil

Dcformation and Journal Motion; (B) Fluid Pressure Profile

278
 2.204
I Y2
I I I I

2.202

2.200 I I I 1 - I It

2.198

2.196
r 1
2.194 I I I , Yl
2.194 2.196 2.198 2.200 2.202 2.204

FIGURE 9. Full Transient Orbit of the J o d at 4000rpm

279
 THREE DIMENSIONAL FLOW AND PRESSURE PATTERNS IN A HYDROSTATIC
JOURNAL BEARING

M.J. Braun and M.B. Dzodzo

Department of Mechanical Engineering
--’i3-3
3 >I-’/
ii
60
The University of Akron J

Akron, Ohio

The flow in a hydrostatic journal bearing (HJB) is described by a mathematical

model that uses the three dimensional non-orthogonal form of the Navier-Stokes
equations. Using the u, v, w, and p, as primary variables, a conservative formulation, finite
volume multi-block method is applied through a collocated, body fitted grid. The HJB has
four shallow pockets with a deptMength ratio of 0.067 . This paper represents a natural
extension to the two and three dimensional studies undertaken prior to this project (see
literature review, Braun, and Dzodzo,1995).

The geometry of the HJB codguration can be seen in Fig. 1. The figure presents
the pockets positioned symmetrically in the axial direction, while the adjacent lands are
connected circumferentially. The present approach takes into account the finite width of
the pockets and bearing lands, thus requiring a three dimensional treatise of the flow for
the entire bearing. The process is steady-state, the fluid is Newtonian, and incompressible.

The governing equations of motion were cast in a dimensionless form. The

characteristic length @), is of the order of magnitude of the clearance and has been chosen
to be O.OOlin(0.0254mm). The equations are written in a general form for a non-
orthogonal body fitted coordinate system in accordance with a formulation first presented
by Peric(1985). The three dimensional approach to the problem required the application of
the multi-block technique, in order to accommodate the geometry of the bearing. A
literature review, as well as some details about the concept and application of the multi-
block approach are presented by Rizzi et al.(1993). The individual blocks contain
collocated control volume cells anchored in a system of corresponding local arbitrary
coordinates. The local system of coordinates is then referenced to a global Cartesian
system.

The case presented here is that of a hydrostatic jet dominated flow which generates
a flattened central vortical cell(CVC) in the downstream part of the pocket, while a fluid
turn-around zone(TA2) occupies all the upstream space, Fig. 2, thus preventing any exit
flow through the upstream port. Figure 3 presents the non-dimensional pressure
distribution(P=plp(p / h)’) in a bearing unwrapped format. Most significantly, the
pressure in the pocket varies strongly across its circumferential direction, while axially its
variation is less pronounced. These results are validated by experimental evidence. It is
worth noting that the present day industrial codes consider the pressure in the pocket
constant, fact that influences considerably the stif€ness and damping predictions.

28 1
 Braun, M.J., Dzodzo, M., 1995, "Effects of the Feedline and Hydrostatic Pocket
Depth on The Flow Patterns arid Pressure Distribution", ASME Journal of Tribology, Vol.
117, NO.2, pp. 224-233.
Peric, M., 1985, "A Finite Volume Method for the Prediction of Three-Dimensional
Fluid Flow in Complex Ducts", Ph.D. Thesis, University of London.
Rizzi A., Eliasson P., Lindblad I., Hirsch C., Lacor C., Haeuser J., 1993, "The
Engineering of Multiblock/Multigrid Software for Navier-Stokes Flows on Structured
Meshes," Computers Fluids, Vol. 22, No. 2/3, pp. 341-367.

282
 :tor

Figure 1 Three Dimensional Computational Grid

TAZ cvc
- -

Flo& In
I
Figure 2 Flow at the Mid-Cross Section of a Hydrostatic Pocket

283
 /
Jet Impingement
Zone A

A
c---
-\
i-

I A-A Pressureat

Pocket 2 Pocket 3 Pocket 4

3 Pressure Map in the Hydrostatic Journal Bearing

284
 THREE DIMENSIONAL FLOW AND PRESSURE PATTERNS IN A SINGLE POCKET
OF A HYDROSTATIC JOURNAL BEARING

Department
M.J. Braun and M.B. Dzodzo
of Mechanical Engineering 97-37
The University of Akron
Akron, Ohio 3 /
d G+f&i
ABSTRACT
The flow in a hydrostatic pocket is described by a mathematical model that uses the three dimensional
Navier-Stokes equations written in terms of the primary variables, u, v, w, and p. Using a conservative
formulation, a finite volume multi-block method is applied through a collocated, body fitted grid. The
flow is simulated in a shallow pocket with a depthllength ratio of 0.02.
The flow structures obtained and described by the authors in their previous two dimensional models
are made visible in their three dimensional aspect for the Couette flow.
It has been found that the flow regimes fonned central and secondary vortical cells with three
dimensional corkscrew-like structures that lead the fluid on an outward bound path in the axial direction
of the pocket. The position of the central vortical cell center is at the exit region of the capillary restrictor
feedline. It has also been determined,that a fluid turn around zone occupies all the upstream space
between the floor of the pocket and the runner, thus preventing any flow exit through the upstream port.
The corresponding pressure distribution under the shaft presented as well. It was clearly established
that for the Couette dominated case the pressure varies significantly in the pocket in the circumferential
direction, while its variation is less pronounced axially.

FORMULATION OF THE MATHEMATICAL MODEL

Geometry. The typical geometry of a shallow hydrostatic pocket incorporated in a single pocket
HJB-like configuration can be Seen in Fig. 1. The figure presents the pocket positioned symmetrically in
the axial direction, while the adjacent lands are connected circumferentially. A cross section A-A
through this pocket is presented in Fig. 2, where we show only a portion of the lands. The figure shows
the relative relationship between the pocket, restrictor, and the lands. At the bottom of the pocket there
is a capillary feedline of length and diameter B. The pocket projected footprint is of length L. The
most influential physical parameter of such a configuration is its aspect ratio DL. The physical
dimensions of the pocket and its environment are presented in Table 1. In earlier papers presented by
Braun and Dzodzo(l994a,b,c, 1995a,b), and Dzodzo et al.(1994) the geometry introduced the
assumption that the cavity width is much larger than its length or depth, thus reducing the problem to a
two dimensional case of coordinates (YI,Yz). The present approach takes into account the finite width of
the pocket, thus requiring a three dimensional treatise of the flow. The process is steady-state, the fluid is
Newtonian, with constant properties, and incompressible.

Dimensionless Equations for the Fluid Model. To obtain a certain degree of generality in the
discussion of the results, the governing equations of motion are cast in a dimensionless form. The
characteristic length has been chosen to be h=0.001in( 0.0254mm). The equations are Written in a
general form for a non-orthogonal body fitted coordinate system in accordance with a formulation first
presented by Peric(1985).
The three dimensional approach to the problem required the application of the multi-block
technique, Fig. 3, in order to accommodate the geometry of Fig. 1. A literature review, as well as some
details about the concept and application of the multi-block software are presented by Rizzi et al.
(1993). The individual blocks contain collocated control volume cells anchored in a system of
corresponding local arbitrary coordinates X',Xz,X3 that are presented in Fig. 3. The local system of
coordinates is then referenced to the global Cartesian system represented by Yl,Yz,Y3.

285
 Using these notations, and the dimensionless variables defined in the nomenclature, the
continuity equation takesthe form

where Uj=Ul P!+U&+U3/3!

The momentum equations for Cartes& velocity components Ui(i=1,2,3) can be written as

where 4,represents the cofactor of the i-th row and the j-th column in the Jacobian matrix J. This
matrix performs the necessary coordinate transformation Yi=Yi@J) where Yi represents the YI,Y~,Y~
reference Cartesian system of coordinates. The advantage of this formulation is its readiness to naturally
interface with the irregular geometry of the pocket as presented by Braun and Dzodzo(1995a,b).
Boundary Conditions. The boundary conditions on the lands, at their open ends are:

The boundary conditions were imposed by using the explicit step-space extrapoIation(Shyy (1985)). The
velocities U1, U2 and U3 at Y3=0 and Y3=1000for the current iteration, were assumed to be equal to the
corresponding velocities of the neighboring interior cells (in Y3 direction) from the previous iteration.
By using these boundary conditions even the recirculation zones at the outlet cross sections can be
predicted. The pressure in both the circumferential and axial directions at the open boundaries of the
lands were assigned to a reference environment pressure,

p-0; Y3=0 and Y3=1000,05 X2 5 2 (4)

The boundaq pressure condition at the inlet of the restrictor(Y2=0) was assigned to be F247.3
for the case of the Couette dominated flow and -1235.8 for the case of the jet dominated flow. The
dimensional pressure and velocity correspondents are given in Table 2. The velocity boundary conditions
at the inlet of the restrictor (Y2=0) are:

3 = a vaY,
aY* 2 =i%+);y2=o
aY, (5)
The explicit step-space extrapolation (Shyy, 1985) was used, in the same way as for the lands' outlet
boundaries, but applied now in the Y2 direction. The resulting velocity distribution at the inlet of the
restrictor was an almost parabolic fully developed velocity profile. The tangential velocity at the shaft
s u r f 4 = 2 ) was U,,=Re=8. This condition resulted in boundary conditions at the rotating shaft surface
of the form

Ul =-U,*sina; U&J,*cosa; U34, at Y~=Y~,+R*cosQ

and Y2=Yzc+R*sina (6)

where Ylc.zcrepresent the position of the center of the shaft, and the angle Q can be determined from
Eqs.6. The rest of the boundary conditions along the stationary walls are set to Ul=U2=U3=0.
RESULTS AND DISCUSSION
Due to the interaction between the penetrating jet coming through the restrictor and the shear
layer carried by the shaft, the flow is eminently three dimensional. Its complicated structure is the
resultant of the mergence between Poiseuille and Couette driven flows, as well as the influence of the
axial boundary conditions. The Poiseuille portion of the flow is caused by the pressure differential
between the inlet of the restrictor and the pocket bottom, while the Couette motion is engendered by the
shear layer generated by the motion of the shaft. Depending on the relatiomhip between the axial

286
boundary conditions (e.g either symmetric pressure, or different pressures at the axial ends), significant
differences may appear in the structure of the axial flow and the pressure distribution on the lands.
Figure 5 presents a two dimensional cross section in the circumferential direction through the
three dimensional flow in the pocket(Figs. 1 , 2 and 3). This cross section is located at the middistance
between the two axial ends of the pocket/shaft assembly. The physical situation involves a flow structure
dominated by a high shear Couette flow interacting with a relatively low pressure grad1entp243.7 at
the restrictor inlet) Poiseuille flow. The figure presents the central portion of the shallow pocket at the
location where the restrictor merges with the pocket floor. It can be seen that the diameter of the
restrictor forms a large gap in the pocket at the point of enjoinment. The shear layer is moving almost as
a rigid body with the shaft and creates two recirculation zones at the location where it merges with the
incoming jet. This central portion presents a modified vortical cell(MOVC), similar to the central
vortical cell in a deep pocket presented by Braun and Dzadzo (1995a,b), but now completely shifted from
the downstream portion of the pocket to the exit region of the capillary restrictor feedline. This change
in location causes the effective diameter of the r d c t o r to be severely diminished, and thus strongly
influences the local flow discharge coefficient. The downstream of the MOVC contains a flattened
recirculation pillow as evidenced by Fig. 6, and the correspondingDetail A.
The detail of the SVC can be better observed in Fig. 7. The dominance of the Couette flow
causes a very remarkable and somewhat little expected effect. Thus a study of Detail A shows that at the
upstream exit of the pocket no fluid exits the pocket in the circumferential direction, but is rather turned
around by the shear layer which occupies the entire clearance, forming a so called turn around
zone(TA2). This situation is entirely responsible for the birth of the SVC. In fact SVC serves to satisfy
flow continuity and redirects the flow in the axial direction of the pocket(just like MOVC).
Figure 8 was obtained by introducing tracer particles at an arbitrary location of the capillary
restrictor entry, and then follow their path as they move towards the exit of the pocket. Figure 8a
presents the three dimensional trajectory of a particle as it moves and gets trapped first in the MOVC,
then leaves it, and enters the SVC. Once the particle reaches the SVC core, it spirals outwards axially,
and leaves the confines of the pocket. Another example of spiral flow moving axially out can be seen in
Fig. 8b. One can conclude that for the Couette dominated flow the fluid is generally trapped in the
pocket and than it axially corkscrews its way out.
A three dimensional picture of the pressure formed under the shaft in the case of the Couette
dominated flow can be seen in Fig. 9. The Rayleigh effect at the downstream exit of the pocket is clearly
visible in the (A-A) cross section of Fig. 9b. The pressure increases all along the circumferential length
of the pocket, starting at the upstream end, and reaching a level approximately 33% higher at its
downstream end. This figure presents clear proof that previous assumptions of constant pressure across
the pocket, or that the pressure is constant in the first half of the pocket are not correct.

CONCLUSIONS
An analytical formulation using 3-D Navier-Stokes equations is used for the determination of
the flow patterns in shallow hydrostatic pocket with adjoined lands, Fig. 1. The laminar, steady-state,
constant properties, Navier-Stokes equations are cast in a formulation that permits a body fitted
coordinates numerical implementation. The results confirm the findings of the 2-D models of Braun et
aL(1993, 1994b) and Braun and Dzodzo (1994a,b,c,d and 1995a,b) and extends the numerical
simulation to the case of the three dimensional flow. The findings of the numerical simulation of a
shallow hydrostatic pocket can be summaflzed
* as follows
0 in the Couette dominated flow the TAZ occupies all the space between the floor of the
pocket and the runner with the consequencethat no flow exits through the upstream port
0 the three dimensional flow shows that the flow trapped either in the MOVC or SVC

eventually finds its way out in the axial direction. In the Couette dominated flow the fluid is
trapped for long periods in the pocket area as it axially corkscrews its way out of the
pocket(Fig. 8). Thus the fluid tracers are zig-zagging several times from the downstream to
the upstream edge of the pocket while moving slowly towards the axial edges of the pocket.

287
 athe pressure under the runner in the case of the Couette dominated flow increases
significantly from the upstream to the downstream ends of the pocket, with a strong spike in
the region of the Rayleigh step end. For the circumferential speeds used in this paper(U=S),
the inertial pressure drops are more pronounced on the axial edges of the pockets as a
consequence of the dominant axial outflow from the pockets.

REFERENCES
Braun,M.J., Dzodzo, M., 1994%Numerical Predictions of Discharge Coefficients in a Hydrostastic
Bearing," Advanced Earth-to-Orbit Propulsion Technology - 1994, ed. RJ. Riclunond RJ. and Wu S.T.,
Alabama, May 17-19, NASA CP-3282, Vol. II, pp. 412-420
Braun, M.J., Dzodzo, M., 1994b, "Effects of Hydrostatic Pocket Shape on The Flow Patterns and
Pressure Distribution" 5th International Symposium on Transport Phenomena and Dynamics of Rotating
Machinery(ISROMAC-5), Vol. A, pp. 88-106, May 8-12, Maui, Hawaii.
Braun,M.J., Dzodto, M., 1994c, "Effects of the Feedline and Hydrostatic Pocket Depth on the Flow
Pattern and Pressure Distribution", Paper No. 94-Trib-27, ASWSTLE Tribology Conference, Maui,
Hawaii, Ocober 16-19.
Braun, M.J., Dzodzo, M., 1995% "Effects of the Feedline and Hydrostatic Pocket Depth on The
Flow Patterns and Pressure Distribution", ASME Journal of Tribologv,Vol. 117, No. 2, pp. 224-233.
Braun,M.J., Dzodzo, M., 1995b, "Effects of Hydrostatic Pocket Shape on The Flow Patterns and
Pressure Distribution", International Journal of Rotating Machinery, Vol. 1, No. 3, (in press)
Dzodzo M., Braun, M.J., Hendricks RC., 1994, "Pressure and Flow Characteristics in a Shallow
Hydrostatic Pocket with Rounded Pocket/J.,and Joints", Proceedings, 4th International Tribology
Conference, AUSTRIB'94, G.W. Stachowiak, ed.,Uniprint the University of Western Australia, Perth,
Australia, Vol. 2, pp. 587-592.
Peric, M.,1985, "A Finite Volume Method for the Prediction of Three-Dimensional Fluid Flow in
Complex Ducts", Ph.D. Thesis, University of London.
Rizzi A., EIiasson P., Lindblad I., Hirsch C.,Lacor C.,Haeuser J., 1993, "The Engineering of
MultiblocIdMdtigrid Software for Navier-Stokes Flows on Structured Meshes," Computers Fluids, Vol.
22, NO. 213, pp. 341-367.
Shyy, W., 1985, "NumericalOutflow Boundary Condition for Navier-Stokes Flow Calculations by a
Line Iterative Method'', AIM Journal, Vol. 23, No. 12, December, pp. 1847-1848.

TABLE 1. The geometrical characteristic of TABLE 2. The runner velocities and the
the pockets restrictor inlet pressure
(for p 9 0 0 kg/m3,
D I ess in m p=9*10-2kg/ms(v=p/p=100*106m2/s) and
L 448.6 0.4486 0.01139444 h=25.4*104m)
D 9 0.009 0.0002286
C 2 0.002 0.0000508 RFU0 8
B 50 0.050 0.00127 Linear speed 31.46 m/s
Lz 292 0.292 0.0074168 Rot. speed 7885 rpm
R 1500 1.5 0.0381
WP 350 0.350 0.00889 Inlet pressure
WL 325 0.325 0.008255 F247.3 34.5*105 Pa(500 psi)
h 1 0.001 0.0000254 F1235.8 172.4*105 Pa(2500
Psi)

288
Fig. 1 Shallow hydrostatic pocket with adjoined lands

289
 - Uo=8

1
L=Q48.6

R=1500

P=247.3

Fig. 2 Circumferential cross section through the shallow hydrostatic pocket (see also
Figs. 1 and 3, cross section A-A)

290
 Y3=0

7”
CLEAJLANCE

LANDS

Fig. 3 Three dimensional view of the block structure of the hydrostatic pocket and a
portion of the lands

29 1
 'DETAIL B
Restrictor-pocket Interface t

Restrictor Inlet

Fig. 4 Cell distribution inside the blocks of Fig. 3

292
Effective cross section /
of the flow
(strangulation of the restrictor)

Fig. 5 Flow structure in the central portion of the pocket. Modified oblong (MOVC)
and secondary vortical cells (SVC) for Couette dominated flow

293
 Core flow attached
to\the runner

Fig. 6 Structure of the modified oblong vortical cell in the dowmtream section of the
Couette dominated flow

294
 Core flow attached
/ t o the runna

Fig. 7 Structure of the secondary vortical cell in the upstream section of the Couette
dominated flow

295
Fig. 8 (a) Partide streakline v h d i z h g the t h e dimensional MOVC, the SVC and
their interaction
(b) Partide streaklines visualizing the spiral flow moving axially out of the W e t
for Couette dominated flow

296
I

Rayleigh step
-
L
\

(b) -20 -10 0 10 20

CIRCUMFERENTIAL

Fig. 9 (a) Pressure distribution formed under the shaft in the case of the Couette
dominated flow
(b) Pressure distribution in the A-A (Fig. 1 and 3) cross section
297
 LIQUID ANNULAR SEAL RESEARCH
/ 2 O -3
-i 7
Alan B. Palazzolo, Balaji Venkataraman, and Sathya S. Padavala
Department of MechanicalEngineering
Texas A&M University
3F y a - - -
College Station, Texas

and

Steve Ryan, Pat Vallely, and Kerry Funston

NASA George C. Marshall Space Flight Center
Marshall Space Flight Center, Alabama

ABSTRACT
This paper highlights the accomplishments on a joint eflort between NASA - Marshall
Space Flight Center and Texas ABM University t o develop accurate seal analysis software
f o r use in rocket turbopump design, design audits and trouble shooting. Results for arbitrary
clearance profile, transient simulation, therrnal eflects solution and flexible seal wall model
are presented. A new solution f o r eccentric seals based on cubic spline interpolation and
ordinary diflerential equation integration is also presented.

299
 INTRODUCTION

Development of the SSME-ATD-HPOTP required accurate simulation for vibration con-

trol. Component modeling of liquid annular seals is a critical step in accomplishing this task.

This is particularly true for the intentionally “roughened bore” annular seals which provide
damping and reduce cross-coupled stiffness. Seal modeling software was developed which
included the following features to more realistically simulate the fluid-structure interaction
forces:
e Arbitrary clearance profile in the axial and circumferential directions
e Fluid compressibility for LH2 and LOX modeling
e Hir’s and Moody’s Friction Factor modeling
0 Eccentric Operation
0 Thermehydrodynamic coupled solution for concentric and eccentric operations
e Transient motion
0 Flexible wall modeling
Bulk flow continuity, momentum and energy equations are solved in a coupled man-
ner. Constitutive thermophysical relations for LH2 and LOX are obtained fiom the MST
Database, MIPROPS. A simple and efficient dependence in the eccentric problem is devel-
oped using cubic spline interpolation. The resulting nth order two point boundary value
problem consists of ordinary B e r e n t i d equations for the zeroth order (steady state) and
fist order (perturbation) cases. Leakage, torque, horse power loss, stiffness, damping and
inertia dynamic coefficients are calculated by the software. Both direct and cross-coupled
coefficients are provided.

300
 GOVERNING EQUATIONS

The 2-Dbulk flow mass continuity, momentum and the energy equations for the seal
are given by,
Mass continuity:

Axial momentum :

Circumferential momentum :

Energy equation :

RESULTS

Figure 1shows the general solution path employed. A structural FE model of the entire
LOX pump was developed at NASA-MSFC and predicted the operating clearance distribu-
tion in Figure 2. The variation of clearance in the axial and circumferential directions raised
concerns related to effects on the rotordynamic co&cients. An arbitrary clearance profile
version of the code was developed, patterned on the geometry shown in Figure 3. Subsequent
simulations revealed only slight deviations of the dynamic coefficients from their values for

an average clearance model, as typified in Figure 4. Further concerns on seal distortion led
to examination of the elliptical clearance profiles in Figure 5. The results in Figure 6 again

show only slight influence on dynamic coefficients except for high "ellipticities".

301
 Past investigations of bearing flexibility effects on direct damping have shown consid-
erable reductions. Consequently, efforts were made to analyze seal flexibility effects. Figure
7 shows a diagram of the general 2-D,axisymmetric, 4-node, isoparametric FE model em-
ployed to obtain seal deformations. Figure 8 contains a computation flow diagram for the
determination of dynamic coefficients with flexibility effects.

Figures 9, 10, 11 show comparisons of direct and cross coupled stiffnesses and direct
damping between Iwatsubo’s published results and the current approach. The results show
some significant changes with the inclusion of flexibility. Figures 12-15 show the effects
of flexible, rigid, constant properties and thermo-hydrodynamic modeling on the dynamic
coefficients of an SSMEATD HPOTP damping seal. Modeling assumptions are seen to
exhibit significant iduence on the dynamic coefficients.
Sudden transient vibrations were observed in the development stages of the SSMEATD
HPOTP. The seal code was updated to provide forces due to transient shaft motions. The
forces were determined by direct solution of the momentum and continuity equations and by
integrating the spatially varying dynamic coefficients, multiplied by increments in position,
velocity and acceleration. The dynamic coefficients were obtained at any spatial position by
interpolating from a radial/circumferential grid. The latter, approximate approach was found
to be nearly as accurate and many times more efficient that the direct solution approach.
Spatial variation of the direct stiffness is illustrated in Figure 16. Transient forces were

applied to the single mass (Jeffcott) rotor model in Figure 17. Numerical integration was
used to predict the locus of the rotor in response to ramp and ramp/sinusoidal type external
loads. Figures 18-21 show that

(4 A model employing concentric linear coefficients may produce significant error

for large motions.

(b) The ”integrated” dynamic coefficient approach yields responses that are nearly
identical with direct solution of the continuity and momentum equations.

302
 SUMMARY

This manuscript has highlighted the development of advanced software for the prediction
of dynamic forces and force coefficients in liquid annular seals. Effects of variable clearance
profile, seal flexibility, therm-hydrodynamic modeling and transient motions have been

included. Future work includes CFD modeling of seal and impeller related fluid forces.

ACKNOWLEDGEMENTS

The authors gratefully acknowledge NASA’s Marshall Space Flight Center and the
NASA/ASEE Summer Faculty Program for sponsoring this research.

303
 REFERENCES

Iwatsubo, T., Yang,B.S., 1987, “The Effects of Elastic Deformation on Seal Dynamics,”

The 1987 ASME Design Technology Conferences - 11th Biennial Conference on Mechanical

Vibration and Noise, Sep.27-30, Boston, Massachusetts, pp. 435440.

Venkataraman, B., and Pdazzolo, A.B., 1995, (‘Thermo-hdrodynamic Analysis of Tur-
bulent Cryogenic Annular Seals with Large Eccentric Motion,” presented at the 1995 Society
of Tribologists and Lubrication Engineers (STLE) Annual Meeting in Chicago, Illinois, May
1419, 1995 and accepted for publication in the STLE Tribology Transactions.
Venkataraman, B., and Palazzolo, A.B., 1995, “Effects of W
all Flexibility on the Rotor-
dynamic Coefficients of Turbulent Cryogenic Annular Seals,” accepted for publication in the
ASME JOURNAL OF TRIBOLOGY and to be presented at the 1995 Joint ASME/STLE
Tribology Conference, Kissimmee, Florida, October 8-11, 1995.

304
 Governing Equations

Perturbation
Analysis

I
Zeroth Order First Order
Solution
Solution Solution Fo

First Order
Seal Forces Solution

Coefficients

Figure 1 Flow Chart of Solution Proccdm

Figure 2 Predicted Clearance Profile for Sed, Unit 3-01

305
 Figure 3 Grid for Numerid Integration

Mamat Load (load angle 290) (N)

Figure 4s D& StiEness for Dirtorted scsl, Unit 3-01

306
 $7 *I elliptical seal

- Kxx - quadratk ptoflla

--*--
K y y - Ymar noRk

0.0 0.3 02 oa 0.4 0.5 0.6 0.7 0.1

Ellipticity. 6

0.0 0.1 02 0.3 0.4 0.5 0.6 0.7 0.1

Ellipticity. 6

Figure 6b N o d e d C m s Coupled StiEncss for Elliptical Sed (Curved)

307
j'th Element

r"
-C-X

Assume infinite seal stiffness

1
Solve 2-point BVP to get pressure distribution. P(Z)

I Assume E =
,
EFeaikd I

r 1
Update clearances of seal based on hd.f(Z)
1
I
lset P(Z)= P A q
Solve for new pressure distribution, P-(Z)
1

1 Obtain first order pressure solution, ~~

ft,f~,f& 1
Assume E = EFc-

Use Ad-symmetric FEM and ji,fr,js,f4

to obtain rt,rz,r~,r4
1
f(z)= f-(zl
Update first order clearances based on rl,rl,rs.ra
] \Set
1
Obtain new first order pressure distribution, f,,..fa,
1
1YES
I Solve for Rotordvnamic coefecients 1
(END)
Figun8 Flaw diagram for t h t numerical solution

308
 0 4.1e8

s
x
38H8
p 3.1H8
d 2dH8
m
:2.1W +Current (rapid)

L
-
e

rn
ISH8
u 1.IH8
-&-)wabubo(Rsrbk)

5 5.5H7
5.0m
2 3 4 5 6

Pressure Ratio (PinlPex)

-~ ~~~~~ ~ ~

Figure 9 Comparison ofDirea Stiffnesswith Iwatsubo and Yang (1987).

2
i
m

w
-n
'0
a?

S
0
x
m
2
0 I I
I
I
2 3 4 5 8

Pressure Ratb (PtnlPex)

Figure 10 Comparison ofcross coupled Stiffnus with Iwatsubo and Yang (1987).

309
Pig

FW 15 Comouisan of dia inmi. b e n m n flexible and rigid mod& far UNIT 341 S u i .

3 10
 210

190

170
n
E
4,
z 150
E.
$ 130
m
m

==
rn
110

-8
tj

L
90
E
70

50

30

Figure 16 Variation of K, for Unit 3-02 as a Function of Rotor Position

Fx.Fy: external loads

Figure 17 Rotor-Seal Model Used for Simulation

31 1
 40
linear- model (e:O)
20
new method 1
0 new method 2

-20

40
n

f
v
-80

% -80

-100

-120

-140

-160

-180
0 10 20 30 40 50 60 70 80

time (ms)

Figure 18

10000

SO00

8000 new method 2

7000
bulk flow model

6000

5000

4000

3000

2000

1000

-1000

time (ms)

Figure 19 Gradndy Applied Load, 8900 N (2000 Ib), Disp. (y), Seal Force (Fy)

312
 225

200
new method 1
175 new method 2

150

h
125
E
=t
W
100

%
75

so

25

-25
0 5 10 15 20 25 30 35 40 45 50

time (ms)

Figure 20

4500

3000

1500

-1500

-3000

-4500

-6000

-7500

-9000

-10500

I 1 I I I I I I I I

0 5 10 15 20 25 30 35 40 45 50

time (ms)

Figure 21 High Fkequency Load (Eccentric), 4450N at 500Hz,Disp. (y), Force (Fy)

313
 SEALS RESEARCH AT ALLIEDSIGNAL J- - 3 7
a
M. mat Ullah
AlliedSignal Engines
AlliedSignal, Inc.
Phoenix, Arizona

A consortium has been formed t o address seal problems i n the Aerospace

sector of AlliedSignal , Inc. The consortium is represented by makers of
Propul s i on Engi nes , Auxi 1 i ary Power Units , Gas Turbine Starters, etc. The
goal is t o improve Face Seal reliability, since Face Seals have become
reliability drivers i n many o f our product lines. Several research programs
are being imp1 emented simultaneously this year. They include: Face Seal
Modeling & Analysis Methodology: Oil Cooling of Seals: Seal Tracking
Dynamics : Coking Formati on & Prevention: and Seal Re1 i abi 1 i t y Methods.

Air/Oil Seals Must Be Improved,

Now!

a Consortium formed across Aerospace

a Products: Propulsion, APU, Starters etc.
0 Focus: Face Seals
Goal: lox Life Improvement
a Strategy: Address Root Causes

315
 Root Causes & Their Effects

0 Causes:
)) High Temperature in Seals
)) Incorrect Tracking Dynamics
)) Coking Propensity of Oil
0 Effects:
)’Leakage (oil & air)
- Smell of Oil or Smoke in Cabin
- High Oil Consumption, etc.
)) Wear

3 16
/

317
 pr
=%a
I

i
7

318
 I

i
319
 Activities This Year
0 Programs On-going:
>> Thermal ModVAna Methods
>> Oil Cooling Heat Transfer
>> Tracking Dynamics ModVAna Methods
>> Coking Chemistry, Causes & Resolution
>> Reliability Methods

0 Most Pgms have 1995 Year End

Deadlines...Future Recomms to follow

Thermal Modl/ Ana Methods

0 Predictive Tools Dev: Temperatures,

Deformations, Stresses, Leakages
0 Steady-State in 1995
0 Includes:
>> Selection of Seal Codes
>> Experimental Verification of Models

320
 - c ---
THERMAL MODELING/ANALYTICALMETHODS ARE BEING DEVELOPED
AND REFINED THAT WILL PREDICT SEAL PERFORMANCE

Oil Cooling Heat Transfer

0 Experimental Heat Trans (AZSt. Univ.)
0 Overall Heat Transfer Coeffs in
1995
0 T o Be Used in Seal Models & Analyses
0 Includes:
>> Oil Jets
>> Oil Splash (conventional method)

>> Proprietary Cooling Scheme

32 1
 EXPERIMENTAL HEAT TRANSFER WORK IS UNDERWAY TO MEASURE
COOLING EFFECTIVENESS, AND TO DEVELOP DESIGN DATA

Tracking Dyn. Modl/ Ana Methods

0 Predictive Tools Dev: Tracking

Response, Face Loads
0 Steady-State in 1995
0 Includes:
B) Selection of Tracking Models
)) Experimental Verification of Models

322
RIG TESTS ARE BEING PERFORMED TO CALIBRAlWCHARACTERIZE
MODELS THAT WILL PREDICT SEAL/ROTOR SEPARATION AS A
FUNCTION OF THE DESIGN PARAMETERS

Coking Chemistry: Causes & Resolution

0 Understand Coking Mechanisms, Precursors
0 Develop Tests to Measure Coking Propensity
0 Develop Practical Solutions
0 Scope Includes (With UDRI):
>> Test Methods Dev
>> Oil Components Analysis
>> Recommend Additive Packages (proprietary)
>> Smart Filter (proprietary)

323
 Reliability Methods
0 Predictive Tools Dev: Seal Life, Failure
Modes
0 Baseline Today’s Seals...Compare Effects of
Future Improvements
0 Weibull Methods
e Scope Includes:
>> Reliability Database Development & Cleanup
)) Process Definition (RawData -to- Life Pred)

Future Research Needs ...

0 Partial List....
>> Transient Behavior of Seals (Temperature
Spiking Problem)
>> Friction & Wear Measurements With
Realistic Geometries
>> Efficient Oil Scavenging from Seal
Compartments

324
 THIS IS AN EXAMPLE OF A FUTURE RESEARCH NEED AT ALLIEDSIGNAL
(MANY SUCH NEEDS HAVE BEEN IDENTIFIED)

A TRANSIENT TEMPERATURE SPIKING PROBLEM HAS BEEN OBSERVED IN

AWOIL FACE SEALS. CAUSES UNKNOWN SO FAR. WHAT CAUSES IT AND
HOW TO PREVENT THIS POTENTIALLYDAMAGING PHENOMENON?

325
INNOVATIVE SEALING TECHNOLOGY FOR NEXT GENERATION SUBSONIC ENGINES

John Munson
Allison Engine Company
Indianapolis,Indiana

and

Bruce Steinetz
NASA Lewis Research Center
Cleveland, Ohio

The next generation of subsonic engines can be expected t o continue the

historical trend towards increased thrust t o weight (T/W) and decreased
speci f i c fuel consumpti on (SFC) . Devel opment programs currently underway
throughout the gas turbine industry such as DoD’s Integrated High Performance
Turbine Engi ne Techno1ogy ( IHPTET) , and more recently NASA’s Advanced
Subsonic Transport (AST) programs, have altered these trends i n both pace
and magnitude. Advanced seals and sealing technologies have become a
prominent part of these efforts due t o the large potential performance gains
which can be realized.
Allison has recently completed a s t u d y for NASA the goal of which was t o
quantize the potential performance benefits which might accrue through the
use of advanced seals i n future subsonic gas turbine engines. For the study
a two engines where analyzed, a small turboshaft and a larger turbofan engine
t o help asses the effect of engine size on the results. Engines were analyzed
stage by stage w i t h the most sensitive areas highlighted. Leakage charac-
teri sti cs for advanced seals were t h e n substituted i n t o secondary a i rfl ow
model s , and the 1eakage reducti ons documented. These 1ea kage reducti ons were
then converted t o changes i n performance, i .e. increased range, decreased
takeoff gross weight, etc. and presented. I t was found t h a t the development
and use of a realtively few advanced seals, less t h a n 5. could for example
reduce SFC by 10% or more.

327
 Improved Sealing Technology Provides the Best
Avenue for Large Improvements in Performance
Supportive of NASA - AST Program Goals
+8-I0% Fuel efficiency improvements in future subsonic
commercial transports (EIS 2005)
+ Reduced emissions
Supportive of IHPTET goals
Relatively low cost technology ( large benefit / cost ratio )
Close correlation between leakage reduction and
% performance improvement

Program Objectives
*Define sensitivity between performance gain and leakage
reduction
.Considered 2 engines, small turboshaft & regional turbofan
.Baseline seal and engine technology defined by current
Allison AE-3007, and LHTEC T801
.Analytically demonstrate the effect on engine performance
provided by incorporation of advanced sealing concepts
.Identify technology hurdles and propose development plan to
mature sealing technologies for 2005 EIS date

328
 Analytical Procedure

G n G \ J
Flow Svstem Models +
-
-
yz~ mbustion
01
Compressor Aero
Efficiency / Emissions
Surge Margin
Aero
/ Turbine
Efficiency

Thpst ( p i h t
Specific ue on umption

Definitions
Selected 2 modem engines to define baseline, both contain
latest seal technologies
+Brush Seals I I I
+Abradable / Abrasive blade tip systems A I A d,
+ Mechanical / Carbon sump seals
What is Leakage ?
+Overboard airflows
+Any airflow which exits or enters
from the engine main power
stream
+Blade & vane cooling flows
specifically exempted

Assumption - cooling flows are exactly correct

329
 Results Reported Based on Simplified Duty Cycle

Regional Turboshaft
Point Condition Duration -
Point Condition Duration

Idle SLS 103°F Day 34% Idle SL Stand Day 5%

Cruise 0.7M, 20K Stand 33% 50% SL I'

I1
40%
MCP
Max SL, ISO+18"F 33% MCP SL I'
U
46%

IRP SL I1
9%

Advanced Engines
Regional 20 - 24,000pound thrust class
+ - Increased cycle pressure ratio 40:l -
+ - Increased turbine inlet temperature - 2900°F
+ - Increased bypass ratio - 1 5 1
Small turboshaft
+ - Compressor unchanged
- Rebladed gasifier / power turbine
+ - Increased power turbine mechanical speed
+ - Increased turbine inlet temperature - 2900°F

330
 Leakage in Regional Engine Concentrated at Relatively
Few Locations

50% of leakage through 2 LukrgoLoulion

Yirdon
Weighted
Weighted
XdToW
Weighted
X
turbine blade / vane gaps Rank Leakage CoreFiow

Frord sump 0 rad Wer 0.00

Top 5% leaks account for honts~npintwrhaftseal
Front sump II sod
75% of total leakage flow Front sump Useal M f o r
CWnpd~CWJOd
0.09
0.12
0
Renriler&afbowd 0.18

Effort at few locations can hWWllp#5d

,cenfersumP#6seal
0.00
0.00 0.01
0.00
0.00
yield large returns Ccmkutablslnardoinor 0.12 247 0.18

1.38
2hdWb3rdnorde ' 0.37 9.13 0.65
3rdmrdob9rdblade ' 0.33 8.34 0.59
3rdWlo41hnonle ' 0.02 0.58 0.06
4lhthnanleb4hblado 0.15 Z.19 027
4lhWeb5lhnonk ' 0.04 0.90 0.07
51h noplo b S1h Made ' 0.11 2.64 0.19
61hbhdebrlb 0.04 1.06 0.W
Cankurtab lslnozdeoulor 0.05 1.05 0.08
lrlnozzb b Itttlade 0.03 0.75 0.05
lslbladolo2ndnordo ' 021 521 0.38
2ndnonleb2ndbbde * 0.15 3.72 0.27
a'ldblsdobjrdnonk ' 0.01 __0.29 0.02
JldnonlebMblrQ ' 0.w 0.00 0.00
3idMadeto4hnonle 0.00 0.00 0.w
4lhnordob4hblade ' 0.00 0.00 0.00
4thbladeb5lh,throzde * 0.00 0.00 0.00
5lhthnonlob6hblade ' 0.00 0.00 0.00
pthbladobrlb 0.01 0.30 0.02

Total Lodm @lad0 tip Exdud) 100.00 7.160285

331
 Smaller Turboshaft Uses Proportionally More Air

0 Leakage is still relatively Ranking % Weighted

Leakage Location Mission Mission %
concentrated Weighted Weighted Core Flow

Top 5 leaks account for Front sump #1 seal buffer 0.08 1.66 0.27
Front sump vent 0.06 1.13 0.18
50% of leakage Front sump #2 seal buffer 0.04 0.91 0.15
Intershaft seal 0.06 1.23 0.20
Leakage somewhat more Comp 1st stage 0.26 5.20 0.85
Combustor to 1st nozzle inner 0.16 3.30 0.54
distributed throughout 1st nozzle to 1st blade ' 0.55 11.10 1.81
engine
2nd I)lade to 3rd nozzle '
2nd blade to 3rd nozzle * 0.55 11.21 1-83
3rd nozzle to 3rd blade ' 0.39 8.13 1.34
3rd blade to 4th nozzle ' 0.33 6.78 1.10
4th nozzle to 4th blade * 0.09 1.80 0.29
4th blade overboard 0.21 4.36 0.72
1st n o d e b 1st blade 0.33 6.80 1.1 1
1st nozzle/ 1st blade trk 0.07 1.45 0.24
1st blade to 2nd nozzle 0.31 6.31 1.03
2nd nozzle / 2nd bid trk 0.22 4.49 0.73

Total Leab (Bladetip Excludq 100.00 16.35

332
Engine Performance Improvements Impact Entire Aircraft
Regional based on 120 - 150 seat aircraft optimized for
minimum takeoff gross weight
+Twin engine
4 1600 NM range
I

+ McDonnell-Douglas MD-80 used as starting point for model

AH-64 Apache used as model for helicopter airframe
+Helicopter equiped with 1016 Lb payload
+Fuel for 260 NM mission + 30 min reserve
+Takeoff gross weight 14612 Lbs

Smaller Engine Shows Greater Sensitivity to Leakage

Reaional Turboshaft
ASK: AFdWgt ATOGW AMFB A
SK AHP ATOG- Ahel A- Ahge
W Bum load
l4.5 N!?I %Et %m ~ S h l E %I a ! P~I % & I !%@I
1.5 -2.8 0.4 2.0
0.7 -2.1 0.2 1.1 1.530 -3.280 0.99 2.497 -39.7 -20.1
0.3 -1.6 0.2 0.6
0.870 -1.770 0.55 1.410 -22.8 -3.78
0.984 -1.667 0.58 1.560 -20.6 -10.4
1.o -1.9 02 1.3 0.732 -1.309 0.44 1.170 -16.1 -8.20
0.8 -0.9 0.1 1.o 1.624 -2.446 0.93 2.554 -30.6 -15.6
0.8 -0.9 0.1 1.o 1.167 -1.174 0.60 1.784 I -15.4 -8.00

Small coefficients per %Increase in bleed flows, or %Decrease in component

efficiency

333
Blade Tip Clearances Strongly Effect Turbine Performance

Reeional Turboshaft
oms Qo3w

40250
F
j 0.010 0-

0E om50

1 0.005 oam
i 0x1060

O m O.oo00
4.015 QolO om5 4.01~~ amo m o.ocso 411025 0 . m
-Tb-@-=1 -Tb-mW

Small turboshaft shows almost 2X sensitivity

/

Similar analysis performed for centrifugal (turboshaft), and

axial (regional) compressors

Seal Technology Requirements

Must be near term to meet EIS 2005 goal

Currently under development somewhere
Required materials currently available
Risk / Cost considerations
Size / Space / Physical constraints limit applicability
Desire more than one approach, if possible

334
 Seal / Seal Technology Strategy

hioritv sealim Locati0m

s - 1
compressor (axial) Seal TechnoloAes

Blade tiDs
compressor (axial) Active Clearance Control

turbine (unshrouded)
turbine (shrouded) -
Demonstration of Potential Regional Performance Gains
-
AFlOW -
ASFC AFnNVq ATOGW -
AMFB
FR Rim -2.50% -1.80% 5.43% -0.50% -2.8%
Camp Int - - -1.1 0% 3.40% -0.40% -
-1.4%
Tot’l Rim -250% -290% 8.83% -0.90% -42%
Comp End -0.20% -0.33% 0.61% 0.09% -0.4%
Preswid -0.32% -0.49% -
0.91 Yo -0.1 3% -0.6%
Tot‘l Face -052% -0.82% 1.52% 422% -1.O%
Tot’l FR -3.02% -3.72% 10.35% -1.lPkl -5.2%
HPT Tip - I -1 60%
L 3.36% 4.48% -
-2.1%
Grand Tot -3.02% -5.32% 13.71% -1.6Vh -7.3%

335
 Demonstration of Potential Small Engine Gains

-
AFlow -
ASFC -
AHP APAYLD -
ARNG
FR Rim -3.27% -5.00% 10.7% 130% 65.7%
Intershaft 4.32% -0.28% 0.56% 7.30% 3.70%
HPT Tip -1.78% 2.67% 33.5% 17.1%
Corn? Tip .
- -3.96% 6.90% 85.3% 43.3%
Tot'l Tip I -5.74% 9.57?h 118.8% 50.4%

Grand Tot -3.60% -11.O% 20.8% 256% 120%

Conclusions
Large gains possible through development & use of advanced
seal technology
Large gains are attainable through the use of only a few
advanced seals
Even larger gains then demonstrated are possible with
"clean sheet" approach to secondary flow system design
No matter what else is done with engine cycle, etc. identified
losses due to sealing remain
Better seals are an inexpensive way to substantially boost
engine performance

336
 FRACTIONAL WHIRL MOTION IN WAVE JOURNAL BEARINGS

Florin Dimofte and Robert C. Hendricks

NASA Lewis Research Center
Cleveland, Ohio
423 - -37
ABSTRACT

Unloaded gas, plain journal bearings experience sub-synchronous whirl motion due to fluid film instabilities and
wall contact usually occurs immediately after the onset of the whirl motion. An alternative is the wave journal
bearing which significantly improves bearing stability. The predicted threshold where the sub-synchronous whirl
motion starts was well confirmed by the experimental observation. In addition, both a two-wave and a three-wave
journal bearing can operate free of sub-synchronous whirl motion over a large range in speeds. When the sub-
synchronous whirl motion occurs, both the two-wave and three-wave bearing can run in a whirl orbit well within
the bearing clearance. At large clearances and wave amplitudes a two-wave bearing, unlike other bearings, can
exhibit a sub-synchronous whirl movement at both low and high speeds, but can run extremely stable and without
whirl at intermediate speeds. Moreover, in these cases, the whirl frequencies are close to a quarter of the
synchronous speed. The three-wave bearing can exhibit sub-synchronouswhirl motion only after a specific threshold
when the speed increases and the whirl frequencies are close to half of the synchronous speed.

INTRODUCTION

The whirl motion of the shaft inside a gas journal bearing is a result of an unstable condition of the lubricant fluid
film. The frequency of this motion is, in most cases, equal or less than
one-half of shaft frequency and is called “Sub-SynchronousFrequency
Whirl” ( S S J W ) . Unloaded plain journal bearings are very susceptible to
the SSFW. Plain journal bearings experiencing SSFW usually develop
an unstable motion and wall contact occurs immediately after the onset
of the whirl motion resulting in bearing failure. This phenomena was
observed soon after gas bearing applications were developed starting in
the middle 1950’s. Due to its importance for bearing life the SSFW was
also well analyzed as a fluid film instability condition. Among others,
Castelli and Elrod [l] and Constantinescu [2] contributed work that
theoretically established when SSFW occurs.

Unlike the plain journal bearing (Fig. 11, a wave journal bearing (Fig.2 ’ FIG.1 PLAIN J O U R N ~BEARING
shows a three-wave bearing) reduces the journal bearing sensitivity to
SSFW. A wave journal bearing [3] is a bearing with a slight, but precise
variation in the circular profile of a journal bearing such that a waved
profile is circumscribed on the inner bearing diameter and having a
wave amplitude equal to a fraction of the bearing clearance. Fig. 2
shows a three-wave bearing. The clearance and the wave and the wave’s
amplitude are greatly exaggerated in Fig. 1 and 2 so that the concept may
be visualized.

BEARING STABILITY

In a bearing stability calculation, as a first approach, the rotor can be I 1

337
considered rigid and

as:

I
(2, - M
zyx
v2) (2, -zqM v2) ][f]-~

and the corresponding

mass is less than

TEST RIG

SSFW and provides

how the bearing behaves under

RESULTS AND DISCUSSION

lain journal bearing is very s

movement o

Then, both a three- and a two-wave ja

338
 Fig. 6. When the radial clearance was reduced to 0.015 mm the three-wave bearing
ran free of SSFW up to a speed of 11900 RPM. The wave amplitude ratio was,in
this case, only 0.168. These experimentally observed thresholds of SSFW were in
very good agreement with the theoretically prediction as can be seen in Fig. 7 [ 6 ] .
Moreover, as speed increases, the whirl orbit is stable keeping a safe range weil
within the bearing clearance (see bottom screen in Fig. 8). The three-wave bearing
profile limits the maximum amplitude of the shaft center (upper screen of Fig. 8)
that modifies the shape of the orbit from circular to a three side orbit (bottom
screen of Fig. 8).

The test of a two-wave bearing with 0.038 mm radial clearance and a wave
amplitude ratio of 0.442 shows that the bearing operates very stable at intermediary
speeds such as loo00 RPM (Fig. 9) while experiencing SSFW at both low and high
speeds such as 600 and 24000 RPM respectively (Fig. 10 and Fig. 11). When the
SSFW occurs the bearing runs stable in a manner similar to the three wave
bearing. However the dominant sub-synchronous whirl frequency, in this case, is
close to a quarter of the synchronous shaft speed. In addition the half frequency
and moreover the synchronous frequency are also present as can be better seen in
the upper part of Fig. 11. When the wave amplitude ratio was decreased to 0.214,
the whirl frequency increases close to half of the synchronous frequency like in
both plain and three wave journal bearings.
FIG. 5 Ssmv OF A PLAIN
B
C

1. Ili, V. Proble
BEARINGS. 360 Deg ai of

339
Engin Trans.

579-588, (1965).

340
Streamfunction Plot ..

Regions I, I1 and Blade Shanks, Run No. 202

Main F5
C
Seal 1
Seal 2
L

Region II

Rotor 1
L ~ e g i o nI

1 Purge F1

Figure3 - ( a ) Streamline pattern in regions I, II and the connecting blade shank region. (b) Streamline
pattern in regions Ill, IV, connecting blade shank region and the slot in the stator support. (c) Flow
detail at seals 1, 2, and blade shank region to illustrate the complex vortical structure and Mainpath
flow ingestion.
Streamfunction Plot - - - -

Regions 111, IV and Blade Shanks, Run No. 202

Main F7 Main F6

- 0.00744 -
-0.01 57-
Region Ill -0 024 -
-0.0323-
Region IV -0,0406-
Rotor 2
L -0.0488-
7 -0.0571-
-0.0654-

Figure 3 - ( a ) Streamline pattern in regions I, I1 and the connecting blade shank region. @) Streamline
pattern in regions Ill, IV, connecting blade shank region and the slot in the stator support. (c) Flow
detail at seals 1,2, and blade shank region to illustrate the complex vortical structure and Mainpath
flow ingestion. .
Streamfunction Plot
Detail in Blade Shanks on Rotor 1, Run No. 202

Main F5
1 4
-
Seal 2 ,

Region I I Blade shanks

Rotor 1

Region I

Figure 3 --Concluded. (c)Flow detail at seal 1,2, and blade shank region to illustrate the complex vortical
stmcture and main-path flow ingestion.
Mixture Fraction (Concentrations)
Computed Values of F1 Concentrations, Run No. 202
Main F5 Main F4
Main F7 Main F6 a-

Rotor 2 Rotor 1
Mixture Fraction (Concentrations)
Computed Values of F2 Concentrations, Run No. 202
Main F5
Main F7 Main F6
4 4- Seal 2
,,a1 3

Rotor 1
Rotor 2

'urge F2

Figus3 Ingestion of main-path flow in multiple

gas-turbine disc cavities
UTRC/MSFC Large Scale Rig
Mixture Fraction (Concentrations')
Computed Values of F4 Concentrations, Run No. 202
Main F4
Main F5
Main F7 Main F6 1.4- 7 - 4 -
I
Seal 1
Seal 2
Seal 3
n
Rotor 2 Rotor I

+
Purge F2

Figure3 Ingestion of mairr-path flow in multiple

gas-turbine disc cavities
UTRClMSFC Large Scale Rig
Mixture Fraction (Concentrations) --

Computed Values of F5 Concentrations, RunNo. 202

Main F5
Main F7 Main F6

Rotor 2 Rotor 1

Figure3 Ingestion of main-path flow in multiple

I gas-turbi ne disc cavities
UTRCIMSFC Large Scale Rig
Temperature
Stage 1-2 Disk Cavities

Labyrinth seal clearance = 0.012 in Labyrinth seal clearance = 0.024 in

1 F i g u ~6
I
I Stream Function
-
Stage 1-2 Disk Cavities
-

- 0 . ~ 1 7 ~
Labyrinth seal clearance = 0.012 in Labyrinth seal clearance = 0.024 in
Note: The magnitudeis greater than 0.I in the free stream region (red color with no contours)

Figure 7
Pressure
Stage 1-2 Disk Cavities

Labyrinth seal clearance = 0.012 in Labyrinth seal clearance = 0.024 in

Pressures on the stator wall
Direction of rotor motion in absolute frarm
1.0

Axial
distilnw-
ZFL

0.5 1.0
a. Numerical results

Axial I
distance
z,rL

i.a
Fraction of time
 O.O Time Fraction 0.5 -
12a Experimental data

Direction of rotor q i n md w k l

-8.42
- I i13
- 14.2-
12b. Smdard k-s Model Results
I
Time Eractlon

. .
12c. L ow-ReModd Results
0.5 Time Fraction 1

Figorell. Stator Wall Preonues as aFuncdon of the Axial md Tangentid Diamce. TimeFraction
0.5 i s at Ivlimmm Cleamce, 0.0 t o 0.5 is PreonueSide, 0.5 to 1.0 isSucdon Side,
1 352
 Form Approbed
REPORT DOCUMENTATION PAGE I
OMB No. 07044188
'ublic reporting burden for this WllecriOn Of informationis estimated to average 1 hour per response. including the lime for reviewing instruuions. SearChirg existing data sourms.
ath he ring and maintaining the data need+ and mnpleting and reviewing the mllection of information. Send mmments regarding thk burden estimate or any other aspect of this
m l W i d information. including suggestions for reducing this burden. to Washington Headquarters Services. Direclorate for Information Operations and Repom. 1215 Jefferson
h i s HQhway. Suite 1204, Arlington. VA 222024302, and to the office Of Management and Budget. Paperwork Redunion Project (0704-0188). Washington. Dc 20503.

I. AGENCY USE ONLY (Leaveblank) 3. REPORT TYPE AND DATES COVERED

1
March 1996 I
Conference Publication
1. TITLE AND SUBTITLE 5. FUNDING NUMBERS

Seals Code Development Workshop

WU-242-2046
i. AUTHOR(S)

Robert C. Hendricks and Anita D. Liang, compilers

r. PERFORMING ORGANIZATION NAME(S) AND ADDRESqES) 8. PERFORMING ORGANIZATION

REPORT NUMBER
National Aeronautics and Space Administration
Lewis Research Center E-9945
Cleveland, Ohio 44135-3191

k. SPONSORlNGlMONlTORlNG AGENCY NAME(S) AND ADDRESS(ES) 10. SPONSORlNWMONITORlNG

AGENCY REPORT NUMBER

National Aeronautics and Space Administration

Washington, D.C. 20546-0001 NASA CP-10181

11. SUPPLEMENTARY NOTES

Responsible person, Robert C. Hendricks, organization code 5300, (216) 977-7507

12a. DlSTRlBUTlONlAVAlLABlLrrYSTATEMENT 12b. DISTRIBUTIONCODE

This publication is available from the NASA Center for Aerospace Information, (301) 6214390. I
13. ABSTRACT (Maximum200 words)

Seals Workshop of 1995 industrial code (INDSEAL)release include ICYL, GCYLT, IFACE, GFACE, SPIRALG,
SPIRALI, DYSEAL, and KTK. The scientific code (SCISEAL)release includes conjugate heat transfer and multidomain
with rotordynamic capability. Several seals and bearings codes (e.g., HYDROFLEX, HYDROTRAN, HYDROB3D,
FLOWCONl, FLOWCON2) are presented and results compared. Current computational and experimental emphasis
includes multiple connected cavity flows with goals of reducing parasitic losses and gas ingestion. Labyrinth seals
continue to play a significant role in sealing with face, honeycomb, and new sealing concepts under investigation for
advanced engine concepts in view of smct environmental constraints. The clean sheet approach to engine design is
advocated with program directions and anticipated percentage SFC reductions cited. Future activities center on engine
applications with coupled seal/power/secondaryflow streams.

14. SUBJECT TERMS 15. NUMBER OF PAGES

370
Seals; Rotordynamics;Bearings; Tribology; Secondary flow 16. PRICE CODE
A16
17. SECURITY CLASSIFICATION 18. SECURITY CLASSIFICATION 19. SECURITY CLASSIFICATION 20. LlMlTATlON OF ABSTRACT
OF REPORT OF THIS PAGE OF ABSTRACT
Unclassified Unclassified Unclassified