Table Of ContentDevelopment and Validation of High Performance Unshrouded Cetrifugal
Impeller
Wei-Chung Chen, M.Williams, John K. Paris, G.H.Prueger
(The Boeing Company, Rocketdyne Propulsion_ Power)
Canoga Park, CA 91309-7922
R.Williams (NASA/MSFC)
Huntsville, AL.
Abstract water-test rig. The experimental data will
be compared with the analytical predictions
The feasibility of using a two stage and presented in another paper 7. The
unshrouded impeller turbopump, Figure 1, to experimental data provides validation data
replace the current three-stage RLV engine for the numerical design and analysis
shrouded impeller hydrogen pump has been methodology. The validated numerical
evaluated from the standpoint of turbopump methodology can be used to help design
weight reduction and overall payload different unshrouded impeller
improvement. These advantages are a by- configurations.
product of the higher tip speeds that an
unshrouded impeller can sustain. The issues Introduction
associated with the effect of unshrouded
impeller tip clearance on pump efficienc Rocket engine weight reduction is constantly}
and head have been evaluated with one- being pursued in order to place more
dimensional tools and full three-dimensional payload into orbit at lower cost. Turbopump
CFD analysis. Unshrouded impeller weight is typically 25% to 30% of the gross
rotordynamic fluid reaction forces and engine weight and thus is a good candidate
coefficients have been established through for weight reductions. Turbopump weight
can be broken down into rotor assembly,
time dependent CFD simulation of the
whole 360 degree impeller with different 20%, and housing assembly 80%. The key
rotor eccentricities and whirling ratios. item to note is that, although, the housing
Unlike the shrouded impeller, the makes up the greatest portion of the
unshrouded impeller forces are evaluated as turbopump weight it is driven by the
the sum of the pressure forces on the blade turbopump rotor element envelope, i.e.
and the pressure forces on the hub using the diameter and axial length, Figure 1.
Reduction in rotor element diameter can be
CFD results. The turbopump axial thrust
control has been optimized by adjusting the done through an increase intip speed
1st stage impeller backend wear ring seal capability. This can be accomplished
diameter and diverting the 2nd stage through an increase in material capability or
backend balance piston flow to the proper removal of the impeller shroud to reduce
location. The structural integrity associated blade stresses. NASA Monograph SP-8109
with the high tip speed has been checked b indicates that for a hydrogen pump using
analyzing a 3D-Finite Element Model at forged titanium impellers an increase in tip
maximum design conditions (6 % higher speed from 2000 feet/sec to 2500 feet/sec is
than the design speed). This impeller was achievable with removal of the shroud. This
fabricated and tested in the NASA/MSFC relates to a 25% reduction in diameter to
achievethe sameimpeller stagepressure clearanceon these turbopumps indicates that
rise.
impeller design parameters can impact the
efficiency defect. In 1997 Johannes Lauer,
et. al. 3describes an experimental study of 14
semi-open impellers of different design.
The results were not conclusive, but
indicated that the blade number, and exit
angle had the largest impacts on tip
clearance sensitivity.
Rotordynamic Coefficient Prediction
Stable turbomachinery operation depends on
Figure 1:Unshrouded Impeller Turbopump the damping of the rotor motion 4. Currently,
Layout
rotordynamic stability parameters are
Reduction in rotor assembly axial length is estimated by using bulk flow theories and
best achieved through elimination of a stage small perturbation (quasi-steady)
or rotor element. This can be accomplished assumptions. A well-established experience
by increasing the head achieved by one rotor base with unshrouded impeller
element through an increase in the head rotordynamics does not exist.
coefficient or through increased tip speed.
The use of high head unshrouded impeller To help understand the unshrouded
technology has the potential to significantl impeller' s rotordynamic performance,
reduce turbopump weight by allowing the Enigma's computational rotordynamic
use of higher tip speeds and decrease of a methodology 5'6 was applied to the
rotor stage. unshrouded impeller. This method directl_
simulates the rotor whirling motion (no
Blade Desiqn to Minimize Clearance impacts
quasi-steady assumptions) and can be, in
principle, applied to large eccentricity whirl
A number of papers have been published on
problems.
the impact of tip clearance on unshrouded
(semi-open) impellers for compressors and
For Navier-Stokes based rotordynamic
pumps. In 1972 Rocketdyne completed an
calculations, the impeller shaft/hub moves
evaluation of shrouded and unshrouded
with an imposed whirling harmonic motion,
impeller performance using the J2 liquid
Figure 2, and the flow equations are
oxygen turbopump. Water tests were
integrated time-accurately until reaction
completed with a shrouded impeller and an
force time periodicity is observed. The fluid
impeller with the shroud removed at various
reaction force vector time history is
tip clearances. The performance impact of a
calculated; the force history can then be
10% increase in the tip clearance resulted in
post-processed and decomposed into normal
a 12% decrease in efficienct_, I. Y. Senoo 2
and tangential components. Because of the
wrote in 1987 that a tip clearance change
direct simulation of the moving hub, the
from 0 (shrouded)to 10% of the impeller
flow model must consist of the complete
exit width decreases efficiency by 4% for
three-dimensional geometry (full 360
compressors. The different impact of tip
degrees in circumference).
2
By comparing these coefficients to those
y
predicted for the SSME HPFTP impellers,
which are comparable in size to the RLV
turbopump impellers, a quantitative
FT
assessment can be made regarding how
significant these impeller forces may be for
an unshrouded impeller. Table 1 lists this
=rotor eccen_rlCl_
O=whirl frequency comparison.
F,=normal force(atminimumclearance)
FT=tangentialforce (atminimumclearance) Table 1: Rotordynamic coefficient comparison,
unshrouded/shrouded impellers
Figure 2: Whirling impeller rotor RLV SSME HPFTP
unshrouded shrouded
Historically, impeller coefficients have
Kxx -50,330 lb/in -20,144 Ib/in
rarely been a design driver from a
rotordynamics stand point in liquid Kxy 95,877 lb/in 8,774 lb/in
hydrogen turbomachinery, for two primary Cxx 15 lb-sec/in 6 lb-sec/in
reasons. First, the magnitude of the impeller
coefficients is proportional to fluid density,
These data, Table l, show that the
and are therefore typically predicted to be
unshrouded impeller direct stiffness (Kxx)
quite low relative to the other coupling
and direct damping coefficients (Cxx) are
elements (in fact, sometimes ignored) in
about 2.5 times larger than the same
liquid hydrogen. And second, Rocketdyne
coefficients currently used to simulate the
has usually incorporated shrouded impellers
SSME HPFTP rotordynamic performance.
into their turbomachinery designs, which
However the cross-coupled stiffness
intuitively would be expected to have lower
coefficient (Kxy), which affects turbopump
rotordynamic coefficients than comparabl_
stability, for the unshrouded impeller is
sized unshrouded impellers. The validity 6f
almost 11 times the same coefficient for the
the second assumption above can be
SSME HPFTP. Not only is this predicted
assessed by examining the normal and
coefficient now so large such that it cannot
tangential force data predicted by the CFD
be ignored, but it is, in fact, over twice as
analysis. By applying a second degree
large as the cross-coupled stiffness attributed
polynomial curve fit to the data in Figure 3,
to the turbine Alford force, which must
rotordynamic coefficients for the subject
always be considered in a rotordynamic
unshrouded impeller can be extracted.
stability assessment. While the increased
--_Tangentlal Force
damping coefficient would partially offset
.................................... -g'_5 ......... [ .-e- Normal Force
_v the destabilizing effect of this large cross-
coupling term, it's net effect would certainl
iv
be much more destabilizing than the
comparably combined effects for a shrouded
o.s 1 _ 5 impeller (stability representing a balance of
whirl drivers, Kxy, and whirl dampers,
WhirlingRatl Cxx).
Figure 3: Computed hub+blades normal and Another point of qualification needs to be
tangential forces
made regarding this comparison. The
3
current best estimateimpeller coefficients
for theSSMEturbopumpsarecalculatedb,_ DCros,_ove=r .45 x Dr., 2 stage
pelter
scalingempirical data derivedfrom rig i'27 x Dim petter 1 stage
.65 x Dim pc.,, 3 stage
testingatCalTech. Analytically predicted
impeller coefficients,from eitherbulkflow
or CFD models, havehistorically not r2xsl ,lai e
comparedwell with empiricaldata,andin
..,:i40x s,age
fact havetypically tendedto significantl
under-predict the measured coefficients.
Therefore, it is possible (perhaps even
[48X_ Dc_4"_" stage
likely)that the difference between the
rotordynamiccoefficientsof shroudedand
..."../2"4x '°
unshroudedimpellersisevenlargerthanthat
suggestedherein, if empirical data were
availableforboth.
In summary,it appearsvery likely thatall The crossover or diffuser diameter, which
the rotordynamic coefficients associated
sets the housing major dimension, is
with unshrouded impellers couldbe dependent on the impeller diameter through
significantlylargerthanthoseofcomparable an increase of the diffusion requirement for
shrouded impellers. Furthermore,and a given head coefficient. The scaling
consistentwith what would be intuitivel_ parameters of 1.27, 1.45, and 1.65 are based
expectedt,he potentialdestabilizingeffects_ on the fact that the more stage pump the
inparticularcouldbemuchmoresignificant. impeller discharge is less tangential and
This makes it imperativethat impeller more radial flow therefore require bigger
coefficientsbeaccuratelyaccountedforin diffuser to diffuse the flow. The pump
the rotordynamicmodelof any turbopump
length is set based on the number of stages
utilizingunshroudedimpellers,andprobabl6_ where the scaling parameters of 32, 40, and
makesthepumpstability issue,whichcan
48 are based on the RS68 fuel turbopump.
beproblematic forhighperformance
Weight should then vary with the diameter
hydrogenpumpsin any case,that much
and length to give a cubic relation to
morechallenging.
volume. Rocketdyne experience has
determined that the relationship is closer to a
With regardtothe current numerical
2.75 power law and this was used.
procedurefor unshroudedimpellers, more
validation is needed to completely assess the
Figure 4 shows the result of the weight trade
impact on the presented layout.
completed to define the impeller tip speed
and staging. The RLV 3 stage baseline is
Pump weight study shown at approximately 2500 pounds using
Wei.qht Calculation Assumptions
the prescribed weight assumptions and
32,000 RPM. Prior to review of the
The following equations were usedto
operating speed to take into account turbine
define the turbopump weight based on stage
constraints, a design point was selected at
number considerations and diameter:
42,000 RPM and an impeller tip speed of
2600 feet/sec. This would have shown a
pump weight improvement of 1086 pounds.
4
Dueto constraintsin turbinematerialsthe design which had a reasonable head
speedwaslimitedto 32,000RPM. With a coefficient(re) and eye-to-tip ratio for
fixed pump speedand a fixed discharge operating flow range considerations. This is
pressureincreasing tip speed increases the 2-stage baseline design point. This
impeller diameterandresultsin aheavier design shows a 626 pound decrease in
turbopumpeven with a reductioninone turbopump weight from the RLV 3 stage
pumpstage.Takingintoaccounttheweight baseline design and a 147 pound decrease
correlationandturbineconstraintonRPM from a 3 stage RLV design with equivalent
resultedin a 2200 feet/sectip speedto head coefficient.
minimize weight andarriveat an impeller
3500
CMC 2stage Metal Alloy 2stage
RLV 3 stage baseline turbine limi / turbine limi
3000 Ut = 1980 FP_ , /
v=o._ \ ta
eye/tip=0.56 \ N
D = 14.2 inch '_1 \
2500
W
n
m
__N. DUets=ig2n6001:2FPsStage
_.r 2000
t.- -- 3stage
Ut = 1800 FPS _ I ¥ =0.38
,iO
¥ =0.53 _ _ eye/tip =0,59
eye/tip =0.61 _ D=14.2 inch 3stage
D. 1500 --D =12.95 inch _ Ut =1980 FPS --
E
eye/lip --0.78
O=
_ _D==100..4648 inch
IOO0
2 stage baseline __
Ut =2200 FPS
¥ = 0.53 --0-2 stage, tipspeed =2600 FPS
500 -- eye/tip =0.5 ---t,--2stage, tipspeed = 2400 FPS _ __ 2utSta2g2e0=0 FPS _--
D=15.75 inch --x-- 3stage, tipspeed =1960 FPS ¥ =0.53
---e-- 2stage, tipspeed =2200 FPS eye/tip =0.695
--IK-- 3sytage, tipspeed =1800 FPS D =11.97 inch
0 h I i i i t i i J
25000 30000 35000 40000 45000 50000 55000
Pump Speed, RP
Figure 4: Weight trade study for design point selection
Axial thrust calculation inch diameter. The total unbalanced force
that a balance piston must counteract is
A concern with unshrouded impellers is the 614,606 pounds. An evaluation was made
balance of axial thrust. Shrouded impellers as to the routing of the balance piston sump
have similar k-factors in the front and rear to achieve this requirement, Figure 7. The
shrouds, whereas an unshrouded impeller initial evaluation was to determine if rotor
does not. This imbalance has direct impact balance could be achieved with routing the
on the position of wear rings and the balance balance piston secondary flow across the
piston design. Figure 5 shows the rotor force second stage impeller alone. This is
distribution. The first stage impeller is represented in the curve with the square
balanced with a wear ring located at a 14.2 symbols. As can be seen, the balance piston
would be operating in a closed position with The use of antivortex ribs increases has a net
the high pressure orifice and there is no result of increasing the balance piston flow
force margin available to account for load rate by 20% over the configuration without
0lbf 558,559lbf antivortex ribs. This increase in flow rate
2664 lbf _,_6 Ibf 15,831lbf would result in a 1.23% drop in pump
50,240 Ibf efficiency. At this point in the design it was
felt that the decrease in performance was not
worth the increase in balance piston
capability. As the design matures this
decision would have to be revisited. It
should be noted that the baseline
Figure 5: Rotor Force Balance
configuration with balance piston sump
uncertainties. Two other cases were then return to the inlet of the first stage without
run with routing of the balance piston sump antivortex ribs has only a 0.8% negative
to the inlet of the first stage impeller. This effect on pump efficiency. This balance
adds some complexity to the mechanical piston shows adequate-operating range with
arrangement but as can be seen in Figure 6 the unbalanced force at approximately 50%
dramatically increases the balance piston of the force range. This sort of mechanical
capability. The circle symbol and triangle arrangement was successfully shown in the
symbol curves-both represent return to the Rocketdyne MK-49 hydrogen pump for an
inlet of the first stage. The triangle symbol orbital transfer vehicle system.
curves shows larger balance piston
2-stage unshroud impeller balance piston flow
capability through the use of antivortex ribs
16
in the balance piston cavity. This effectivel
14
reduces the swirl in this cavity and decreases
the pressure drop increasing the net force. 12
10
.._
2.-_ _ impellerE_iarx_l:Islo'_Force_lity _8
900000 a
,'," 6
80O0O0
4
j 'p"
_ 600000_
s:xJooo
ID _ngpant 0 0.035 0.01 0.015 0.02 0.025 0.03 0.035
0 rotortravel,inches
_d --=-B.P.Flowto1ststage,'N'oanlJ-vorte0d(bs(baseline) Figure 7: Balance piston flow rate
L:K_O00- -4_-B.P.Rowto 2ndstage,v_thantl-vort¢<dbs
Structural Analysis
100000 B.P.Rowto1ststage,_gi ar_-vorte_ribs
-*- _ force
0 r , i A structural assessment was performed on
0 0.0(35 0.01 0.015 0.02 0.025 0.03 0.035 the unshrouded impeller to demonstrate that
rotortravd,inct_
the structural factors of safety and calculated
Figure 6: Balance piston configuration study
fatigue life meet the required structural
criteria (factor of safety). Hub bust analysis
The negative consequence of the antivortex
was done using two-dimensional
ribs is an increase in balance piston
axisymmetric finite element models (FEM)
secondary flow. Figure 7 shows this impact.
following the Rocketdyne burst
methodology. Three hub thickness values Conclusions
ranging from 0.200 to 0.400 inch were
evaluated; the results of which are presented The test data for this particular impeller was
in Table 2. not available yet at this paper publication,
however based on the existing data and
Table 2: Burst Analysis Results some CFD predictions the following
Hub Thickness conclusions can be made:
0.200 in. 0.300 in. 0.400 in.
FSl_rst 2.02 1.90 1.77 (1) Unshrouded impellers can be used to
N*_lo,,(rpm 41,385 39,490 38,187 reduce the pump overall weight and increase
Vtlp allow 2,996 2,859 2,764 the overall payload.
(ft/sec)
(2) CFD predicts that unshrouded impellers
Three-dimensional FEMs were used to create unfavorable rotordynamic coefficients
determine the yield, ultimate, and high cycle in the pump. How this impacts on the pump
fatigue (HCF) factors of safety. Results are rotordynamic stability needs more study.
shown below in Table 3.
(3) Unshrouded impeller axial thrust balance
Table 3: Blade Factors of Safety
is similar to that of a shrouded impeller;
Hub Thickness (inches) Required need to identify the proper location for the
Safety
0.200 0.300 0.400 Factor balance piston sump.
Yield 1.92 1.95 1.89 1.10
Ultimate 1.79 1.82 1.77 1.4 (4) Structural analysis usually favors the
HCF 2.40 3.26 3.33 1.4 unshrouded impeller because the common
high stresses located near the L.E. regions of
The initial design employed a hub thickness a shrouded impeller no longer exist.
of 0.200 inch. High stresses were obtained
at the blade tip with pressure loading on the (5) Impeller efficiency is lower for
blades and hub. Hub bending caused the unshrouded impellers. The efficiency loss
high stresses, which also existed for the can be minimized by reducing the shroud
0.300 inch hub. The stresses obtained using clearance through pump development
the 0.400 inch hub were satisfactory. program.
Optimization of the hub for weight or other
References
considerations would require a more detailed
analysis.
J R.K. Hoshide and C.E. Nieison, "Study of Blade
Clearance Effects on Centrifugal Pumps," Contract #:
The items evaluated met the required
NAS3-13311, Report #: NASA CR-120815, NASA-
structural criteria. It was determined that the Lewis Research Center
blade thickness could be reduced by 5% and 2 y. Senoo and M. Ishida, "Deterioration of
Compressor Performance Due to Tip Clearance of
still maintain adequate margins on the
Centrifugal Impellers," Journal of Turbomachinery,
factors of safety. The maximum allowable
January 1987, Vol. 109, pp. 55-61
tip speed, taking hub burst and blade life 3 Johannes Lauer, et. al., "Tip Clearance Sensitivit
into account was determined to be of Centrifugal Pumps with Semi-Open Impeller,"
approximately 2700 ft/s. 1997 ASME Fluids Engineering Division Summer
Meeting, FEDSM97-3366.
4
WilliamsM, ., "A HelmholtPzressurEequation
Method for the Calculationof Unstead
IncompressiVbliescouFslows",Int.J.NumeMr.eth.
FluidsV,ol.14,1992p,p.l-12.
5 M.Williams,W.Chen,L. Brozowskai ndA.
Eastland,"Three-DimensionFainl ite Difference
MethodforRotordynamFicluidForcesonSeals,"
AIAAJ.,vol.35,No.8,pp.1417-142109,97.
6Hiwata Akira, Tsujimoto Yoshinobu , "Theoretical
Analysis of Fluid Forces on an Open-Type
Centrifugal Impeller in Whirling Motion" Forth-
International Symposium on Pumping Machinery,
May 29, 2001 New Orleans, LA.
7 R. Williams, "Comparison of Unshrouded Impeller
Analysis and Experiment" 37th
AIAA/ASME/SAE/ASEE Joint Propulsion
Conference, Salt Lake City, Utah. 8-11 July, 2001.
