Document
Proceedings of IGTI: ASME TURBO EXPO 2002 3-6 June 2002 Amsterdam RAI International Exhibition & Congress Center Amsterdam, The Netherlands
GT-2002-30004
AN ENGINE RESEARCH PROGRAM FOCUSED ON LOW PRESSURE TURBINE AERODYNAMIC PERFORMANCE Santo Chiappetta Raymond Castner Pratt & Whitney Canada NASA Glenn Research Center Dr. John Adamczyk John Wyzykowski NASA Glenn Research Center Pratt & Whitney Canada A number of studies (Halstead, et al. (1997); Hodson, ABSTRACT (1990), LaGraff and Ashpis, (1997) suggest that the A comprehensive test program was performed in the boundary layers on LPT blading transitions towards a Propulsion Systems Laboratory at the NASA Glenn laminar flow state as Reynolds number is reduced. Thus Research Center, Cleveland Ohio using a highly for a fixed level of aerodynamic loading a reduction in instrumented Pratt and Whitney Canada PW 545 turbofan Reynolds number can result in flow separation. If the engine. A key objective of this program was the separated flow regions are large the efficiency of the LPT development of a high-altitude database on small, high- will be compromised. Having a flow model, which can bypass ratio engine performance and operability. In accurately predict the Reynolds number lapse of a LPT is particular, the program documents the impact of altitude key to the execution of successful designs. This is of (Reynolds Number) on the aero-performance of the low- particular importance today because of the emphasis on pressure turbine (fan turbine). A second objective was to reducing design time and reducing LPT blade count assess the ability of a state-of-the-art CFD code to predict without sacrificing LPTefficiency. In addition the recent the effect of Reynolds number on the efficiency of the low- interests in Uninhabited Aerial Vehicles (UAV) for high pressure turbine. CFD simulation performed prior and altitude surveillance has added even more emphasis on after the engine tests will be presented and discussed.
the need for models that can accurately predict the LPT Reynolds number lapse in efficiency.
Key findings are the ability of a state-of-the art CFD code to accurately predict the impact of Reynolds Number on The work in this paper outlines a test program in which the efficiency and flow capacity of the low-pressure aero-performance data is acquired for an LPT turbine turbine. In addition the CFD simulations showed the operating in an engine environment over a range of turbulent intensity exiting the low-pressure turbine to be Reynolds number typical of a UAV application. The high (9%). The level is consistent with measurements engine used in this test is a Pratt & Whitney Canada taken within an engine.
(PWC) PW 545 jet engine. The PW 545 engine is a high- bypass engine with a thrust rating of 3000 Ibs. The LPT INTRODUCTION turbine in the PW 545 engine has three stages. This It is well known that the aero-performance (adiabatic engine was highly instrumented in order to determine the efficiency) of the low-pressure turbine (LPT) of a turbofan Reynolds number lapse in efficiency of the LPT. CFD engine decreases with altitude or Reynolds number. This simulations were preformed using the CFD code reduction in aero performance is commonly referred to as APNASA, Adamczyk et.al. (1990) prior to the tests. These Reynolds number lapse.
This is a preprint or reprint of a paper intended for presentation at a conference. Because changes may be made before formal publication, this is made available with the understanding that it will not be cited or reproduced without the permission of the author. Copyright _ 2002 by ASME measured by the boundary layer rakes. The forward edge simulations established the ability of a state-of-the-art of the duct was the metric break location for the entire CFD code to predict the Reynolds number lapse inlet assembly.
of the LPT efficiency. Results from simulations executed after the tests will also be presented. These post-test simulations further support the ability of the CFD code to _-%,_ "_t accurately estimate the Reynolds number lapse of the ,_= '\ LPT efficiency.
t3oundary ",, Engine Test Set-up Layer Rakes'_, The PW 545 High Altitude Test was performed in the Cross NASA Glenn Research Center Propulsion Systems Duct Laboratory, Cell 4. An overall installation sketch of the test Rake
engine installation is shown in Figure (1). I
/ / Inlet Plenum PW545 Exhaust .>'_ Duct Static Engine Collector ................ " Pressures Figure 2 - PW 545 Station 1.0 Instrumentation Six thermocouple rakes with 12 elements each were installed near station 1.0 to measure the incoming total temperature profile. The station 2.0 duct mounts directly to the engine compressor inlet case. Four boundary layer rakes were installed at this location to record the total pressure profile and airflow at this location. The Figure 1 - PW 545 in PSL4 measurement accuracy of the incoming engine mass flow was estimated to be better than 1%.
Turbine Measurements The test engine, a PW 545 turbofan, was hard-mounted For this test program the LPT was heavily instrumented to via engine stand to a single-axis thrust stand. The metric record the lapse in turbine performance with Reynolds bed is designed to accommodate the PWC engine stand, number. Over 150 pressure and temperature sensors were installed in the LPT module. The LPT and also support the inlet ducting. One measurement and one calibration load cell are included on the stand. An on- instrumentation plus the instrumentation in the fan and board hydraulic cylinder and hand pump is used to core compressor were used to infer the incoming flow calibrate the stand prior to all testing. The engine was conditions to the LPT (corrected flow) and its efficiency.
built, instrumented, and acceptance tested in a sea level The accuracy of the efficiency estimates derived from the test facility in Mississauga, Ontario, Canada and flight data was +/- 1 point.
tested on the PWC Boeing B720 Flying Test Bed (FTB).
Test and CFD Simulation Results The flow capacity of the test cell ranges from 250 pps to 400 pps at a typical inlet pressure of 4 psia. For these Prior to the engine tests CFD simulations were performed tests the inlet flow was reduced to 12 pps at an inlet at four Reynolds numbers. The rotation speed of the LPT was fixed at a corrected speed parameter of 280. This pressure of 1.13 psia. The low flow rates required for this test program, relative to what is typical, necessitated speed parameter is defined as the physical rotational additional inlet instrumentation. The station 1.0 speed (rpm) of the LPT divided by the square root of the instrumentation duct (Figure 2.) provided mounting for total temperature (measured in Degs. Rankine) of the four boundary layer rakes, one cross-duct rake, and wall gas stream entering the LPT. The inlet flow conditions static instrumentation ports. The cross-duct rake was specified in the simulations (total temperature, total used to verify the total pressure profile at the duct inlet as pressure, flow angles, and turbulence level) were provided by Pratt and Whitney Canada.
2 Copyright @) 2002 by ASME Normalized Efficiency vs Reynolds The CFD simulations were performed using the APNASA Number code, Adamczyk et al. (1990). The APNASA code solves the average passage equation system as formulated by 1.02 ] Adamczyk (1985). This equation system governs the time average flow field within a typical passage of a blade row embedded in a multistage axial flow turbomachine.
oo94
APNASA .__ogB 4 • _'} 0 97 l The turbulence model used by APNASA in this study is an _ ---C-'-- NASA Data enhanced k-E model (CMOTT turbulence model)
i
developed by Shih et al. (1995) modified to be consistent J ..... -::_-. FTB Data 0.94 4 with the average passage equation system. The uses of 093 !
300 350 this turbulence model in CFD simulations of 5O 100 150 2OO 25O turbomachinery is reported upon by Shabbir et al. (1996), ReynoLds Number (xl000) Adamczyk et al. (1998). The CMOTT turbulence model Figure 3 does not explicitly account for transition. As reported by Adamczyk et al. (1998) the model appears to provide The corresponding experimental results derived from the reasonable estimates of the aero performance of axial engine tests at NASA Glenn are shown on the figure.
flow compressors in flow regimes where the flow is known This data ranges from a Reynolds number of 50,000 to to be in transition.
165,000. In addition efficiency estimates derived from a fight test bed (FTB) program have also been included.
The CFD simulations performed prior to the engine tests The Reynolds numbers for the fight test program range incorporated models for purge flows and leakage about from 110,000 to 235,000. All efficiency estimates (engine the rotor tip knife-edge seals (all rotors are tip shrouded).
results are not a direct measurement) have been The predicted normalized efficiency lapse with Reynolds normalized by their respective value at a Reynolds number is shown on Figure (3). The Reynolds number is number of 165,000. Thus all results have a value of one at defined in terms of the chord of the first LPT vane, and the Reynolds number of 165,000.
flow conditions at midspan at the exit of the vane. The simulations span a large range of Reynolds number Figure (3) shows that the Reynolds number efficiency ranging from 30,000 to 295,000. The efficiency estimates lapse as predicted by the CFD simulations is in very good derived from the CFD simulations used mass averaged agreement with that deduced from the engine data. From inlet and exit flow values of total temperature and total a Reynolds number of nearly 300,000 to 30,000 the CFD pressure. In addition the efficiency estimates explicitly simulation predicts nearly a 7 percent reduction in accounted for leakage and purge flows.
efficiency. The predicted lapse is well within the accuracy of the efficiency estimates derived from engine data.
The next figure shows the lapse in LPT inlet corrected flow with Reynolds number. Once again CFD simulation results as well as results derived from both engine tests (Tests at NASA Glenn, and fight test bed) are shown.
All inlet corrected flow estimates have been normalized with respect to their value at a Reynolds number of 165,000. All estimates show a well defined trend, and are self consistent with each other. Once again the CFD results are well within the accuracy of the flow rate estimates derived from the engine test data.
3 Copyright _ 2002 by ASME Figure (6) shows lapse in inlet corrected flow with Normalized Inlet Corrected Flow vs Reynolds number, both with and without purge and leakage flows. The results show that the dependence of Reynolds Number the lapse of inlet corrected flow on Reynolds number is a weak function of the leakage and purge flows.
Normalized Inlet Corrected Flow vs
1 ' O
m Reynolds Number, Fixed Corrected u. 0.99 i Speed Parameter _ _ NASA Data .,,., 0.98 ! --e--APNASA P 0.97 1.01 " _ FTB Data O O
0.96 ]
0.95 E O.99 "O 0 100000 200000 300000 400000 0.98 _, APNASA Reynolds Number .m O.97 O o 0.96 Figure 4 Purge and Leakage Flows, APNASA 0.95 b At the start of the CFD simulation exercise questions 0 100000 200000 300000 400000 arose as to the dependency of the simulation results on Reynolds Number the values specified for purge and leakage flow rates. It Figure 6 was hypothesized that the value for efficiency is highly dependent upon the values set for these flow rates Based on the results presented in Figures (5) and (6) all but that the Reynolds number efficiency lapse is a weak additional CFD simulation results to be presented will not function of these flow rates. This was the case include the effect of purge and leakage flows.
irrespective of whether the efficiency estimates were derived from CFD results or engine tests data. To A second set of CFD simulations were generated establish the validity of this hypothesis a series of CFD (executed after the engine tests was completed) at a simulations were performed without purge and leakage speed parameter of 250. These simulations were flows. These CFD simulation results are shown in Figure executed in order to compare with engine data for (5) along with the results from Figure (3). Once again the Reynolds numbers below 50,000. Four simulations were efficiency results are normalized as outlined above. With done ranging from a Reynolds number of 25,000 to the exception of the results at the lowest Reynolds 135,000. Results from this second set of simulations are number, the Reynolds number efficiency lapse for this shown in Figure (7). This time the efficiency has been engine is a weak function of the purge and leakage flow normalized with respect to the efficiency at a Reynolds rates.
number of 135,000. Efficiency estimates derived from engine data are also shown. These results have also Normalized Efficiency vs Reynolds Number, been normalized with respect to their value at a Reynolds Fixed Corrected Flow number of 135,000. The range of the engine results is from 30,000 to 135,000.
1.02 o 1 = .2 _.0.98 ILl -00.96 N "__Flow, APNASA _0.94 E _0.92 V + Leakage and Purge Flow, APNASA z 0.9 0 100000 200000 300000 400000 Reynolds Number Figure 5 4 Copyright/¢; 2002 by ASME Log of Entropy Rise vs Log of Reynolds Number Normalized Efficiency vs Reynolds Number at Corrected Speed of 250 Ln Reynolds Number 8 10 12 14 2 4 6 1.02 O Corr Spd=270, APNASA -0.5 O C D Corr Spd=250, APNASA "6 0.98 .e E LU "o 0.98 o ._. ata =_ 0.94 I_-1.5 ta O C z 0.92 ...I ;
ol
0.91 -,: ] 150000 50000 100000 Reynolds Number -2_.5 - Figure 7 Figure 8 This time the correlation between engine data and CFD results is not as clear as that on Figure (3).
There is even a lack of correlation between the estimates Key to the ability of APNASA to predict the efficiency derived from the two sets of engine test data.
lapse of the LPT with Reynolds number is capturing the However, if one neglects the estimate at a Reynolds turbulence level through the LPT. The turbulence level is number of 40,000 derived from the tests at NASA Glenn defined in terms of turbulent intensity as: there is good agreement between estimates derived from the CFD simulations and those derived from the tests at I= _/q (1) NASA Glenn. Given that the estimates derived from the engine tests data have an error of plus or minus one point where k is the axi-symmetric average of the turbulent one could also state that the CFD based estimates are in kinetic energy, and q is the axi-symmetric average of the reasonable agreement with those derived from the FTB absolute velocity. A plot of the turbulent intensity at mid- study.
span at various axial locations within the LPT is shown in Figure (9).
The CFD simulations provided a creditable estimate of the efficiency lapse of the LPT with Reynolds number. Based Turbulence Level Through LPT on these simulations an estimate of the dependency of entropy rise across the LPT on Reynolds number was attempted. The results are shown in figure (8) in which
<>
the log of the entropy rise across the LPT is plotted as a function of the log of the Reynolds number. A reasonable t.
<>
fit to the simulation results is a linear curve whose slope is ._= -0.274. This dependence of entropy rise across the LPT
<>
on Reynolds number lies between that for a laminar flow (-0.5) and that for a fully turbulent flow (-0.2).
U == 4
<>
"5 I- Inlet R1 exit R2 Exit R3 Exit Axial Location Figure 9 5 Copyright G: 2002 by ASME turbulence model play a key role in generating the results These values are derived from the CFD simulations for a presented in this paper.
Reynolds number of 295,000, which corresponds to Sea Level Take-Off conditions. The results show the turbulent References intensity increases through the first two stages of the LPT but decreases slightly across the last stage. The Adamczyk, J.J., 1985, "Model Equation for Simulating decrease is the result of a reduction in the aerodynamic Flows in Multistage Turbomachines," ASME paper 85-GT- loading of the last stage relative to the second stage.
226.
Figure (9) shows that the turbulence intensity is more than Adamczyk, J.J.,Celestina, M.I., Beach, T.A., and Barnett, doubled between the inlet and the exit of the LPT.
M., 1990,"Simulation of Three-Dimensional Viscous Flow Sharma (1998) reported measurements of turbulent within a Multistage Turbine", Trans. ASME, 112,370-376.
intensity downstream of a LPT at Sea Level Take-Off conditions ranging from 12% to 16%. The turbulent Adamczyk, J.J., Hathaway, M.D., Shabbir, A., and intensity derived from the CFD simulation is not out of line Wellborn, S.R.,1998,"Numericat Simulationof Multistage with the measurements reported by Sharma (1998). The Turbomachinery Flows," Presented at the Vehicle high level of turbulence intensity aft of the first stage Technology Symposium onDesign Principles and raises issues as to the nature of the transition process in Methods for Aircraft Gas Turbine Engines, hosted by the inner stages of a LPT.
AGARD in Toulouse, France May 11-15, 1998.
Halstead, D.E., Wisler, D.C., Okiishi, T.H., Walker, G.J., Conclusions Hodson, H.P., and Shin, H., 1997,"Boundary Layer Development in Axial Compressors and Turbines Part 1 of A comprehensive engine research program was 4: Composite Picture,"Journal of Turbomachinery, Vol.
conducted to establish the Reynolds number efficiency 119, p.114.
lapse of an LPT under engine operating conditions. In support of this engine research program a series of CFD Hodson, H.P., 1990, "Modeling Unsteady Transition and simulations were preformed to establish the ability of a Its Effects on Profile Loss," ASME Journal of CFD code (APNASA) to predict the Reynolds number Turbomachinery, Vol. 112, No. 4.
efficiency lapse as well as the lapse in LPT inlet corrected flow with Reynolds number. Both the CFD simulations LaGraff, J.E., and Aships, D.E., 1998, "Minnowbrook II and the engine tests points spanned a wide range of 1997 Workshop on Boundary Layer Transition in Reynolds numbers, which makes the current study Turbomachines," NASA/CP-1998-206958 important.
Shabbir, A., Zhu, J. and Celestina, M.L., The CFD simulation results presented in this paper 1996,"Assessment of Three Turbulence Models capture the lapse in aerodynamic efficiency with Reynolds in a Compressor Rotor",ASME Paper No. 96-GT-198 number quite well. It also appears the CFD simulations accurately capture the lapse in LPT inlet correct flow with Sharma, O.P., 1998,"Impact of Reynolds Number on LP Reynolds number. CFD simulations show that the effect of Turbine Performance", NASA/CP-1998-206958 leakage and purge flow on the efficiency lapse and inlet corrected flow lapse is small.
Shih, T.H., Liou, W.W., Shabbir, A., Zhu, J. and Yang, Z., 1995, "A New k-e Eddy Viscosity Model for High Reynolds The CFD simulations were executed using a turbulence Number Turbulent Flows", Computers Fluids, 24, 3, 227- model that does not explicitly account for flow transition.
238.
The model does however account for the production of turbulence due to the straining of wakes as they convect through a blade row. It is the straining of wakes that leads to an increase in turbulence intensity through the LPT.
The model also accounts for the damping of turbulent kinetic energy near solid surfaces, the extent of which increases as the Reynolds number is reduced. The high turbulence level of the free stream penetrates the outer region of the blade boundary layers to a depth established by the wall-damping model. The resulting state of the boundary layer thus determines its response to an imposed pressure gradient. These key elements of the 6 Copyright © 2002 by ASME