IMPROVED BLADE PROFILES FOR HIGH LIFT LOW PRESSURE TURBINE APPLICATIONS

Transcription

1 IMPROVED BLADE PROFILES FOR HIGH LIFT LOW PRESSURE TURBINE APPLICATIONS P. González *, I.Ulizar *, H.P.Hodson ** * ITP, Industria de Turbo Propulsores, SA. Parque Empresarial San Fernando Avda. Castilla N.2, Edificio Japón, Madrid, SPAIN paloma.gonzalez@itp.es, inaki.ulizar@itp.es ** Wittle Laboratory University of Cambridge, Madingley Road Cambridge, CB3 ODY, UK. hph@eng.cam.ac.uk 1. ABSTRACT Nowadays, there is a significant effort aimed toward improving LP turbine efficiency. This is because of the large effect that the efficiency of the LP turbine has on the SFC in comparison to the other modules in the engine. Low pressure turbines already operate at efficiencies well above 90% which makes the challenge of reducing the losses even more difficult. The loss generation processes basically depend on the suction surface boundary layers, the pressure surface boundary layers and the three dimensional regions of the flow. To date, the pressure surface has received very little attention. The dependence of the profile losses on the behaviour of the pressure surface flows has been investigated for the case of a high lift design that is representative of a modern civil engine LP turbine. Two profiles with different pressure surfaces were designed and tested over a range of conditions. The first profile is a thin-solid design. This profile has a large pressure side separation bubble extending from near the leading edge to mid-chord. The second profile is a hollow design. It has the same suction side as the thin-solid design, but there is no pressure side separation bubble. The study is part of a wider on-going research programme covering the effects of the different design parameters on losses. This paper describes the experiments conducted and the results obtained in a low-speed linear cascade facility. Steady state two-dimensional measurements are presented in the form of isentropic surface velocity distributions and wake traverses downstream the cascade. It is shown that the thick profile generates only around 90% of the losses of a thin-solid profile. Nomenclature: α 1 α 2 C ax I LP Inlet angle. Outlet angle. Axial chord. Incidence. Low Pressure.

2 Re Re d s SFC solidity V V 2 Reynolds number based on true chord at exit conditions. Design Reynolds number based on true chord at exit conditions. Surface length. Specific Fuel Consumption. Pitch to axial chord ratio. Local velocity. Exit velocity. 2. INTRODUCTION The modern large civil aero-engine LP turbines consist of several stages. This makes not only the efficiency but also the weight and manufacturing cost important parameters in the design process. SFC is highly influenced by the LP turbine efficiency, the weight of the LP turbine represents over 20% of the engine weight and the cost could be up to 15% of the whole engine total cost. In order to reduce weight and cost without penalising the efficiency, the number of aerofoils has been reduced in recent years as a result of increases in the lift coefficient, leading to the so called high lift profiles. The development of these profiles is supported by computational studies and experimental evidence [2], [8]. High lift profiles have been introduced into the latest LP turbines for civil applications such as the BR715 and Trent 500 engines [6],[7]. Profile losses are greatly dependant on the development of the blade surface boundary layer [1], [4]. Due to the large aspect ratios existing in LP turbines, the aerofoil loss is by far the largest percentage of the total loss, accounting for up to 80% of the profile loss according to [2]. Furthermore, reducing the 2D losses by 10% to 90% of their former value, can raise the efficiency of the LP turbine by approximately 0.5%. Therefore, it is important to be able to predict such changes as accurately as possible in order to control the loss generated. Two very different profile design options are available for use in engines today, either thin solid or hollow aerofoils. Hollow aerofoils are lighter, more efficient, mechanically more robust at large aspect ratios but they are more expensive because of the increased manufacturing complexity. The current LP turbine design philosophy is based on thin-solid profiles but LP turbines using thick hollow aerofoils have accumulated around 100 million hours of successful operation over the last thirty years. Turbine aerofoils are typically optimised for their design point, but the profiles do not always operate at their design conditions. Incidence, Reynolds number and Mach number vary across the operating range. This study is an attempt to define the differences between thin-solid and thick hollow aerofoils. The objective is to discover by which mechanisms and by how much the thickening of the profile influences the aerodynamic behaviour of the aerofoil. This is assessed in terms of changes in the losses, the boundary layer behaviour and tolerance to changes in incidence. This paper describes the experiments conducted in a low speed linear cascade in order to improve the understanding of the differences between using thin solid and thick hollow aerofoils in the LP turbine. 3. EXPERIMENTAL SETUP. The experiments were conducted in a low-speed cascade wind tunnel in the Whittle Laboratory, Cambridge University. Figure 1 shows the test section and some profile details.

3 Static pressure tapping at 25% and 50% axial chord downstream trailing edge 352),/( '(7$,/6 α α & D[ VROLGLW\ ž ž PP FLOW α 1 α 2 Instrumented aerofoil Inlet static pressure tapping Figure 1. Test section of the low-speed cascade. The cascade consists of six aerofoils with constant section. Two highly loaded profiles were tested. Each has the same lift coefficient, around 1. Both profiles have physically identical suction surfaces but different pressure surfaces. Profile F is a thin-solid high lift profile following the current LP turbine design philosophy. Profile G is the redesign of Profile F. It is a thick high lift profile representative of a hollow blade. Profile G was designed by thickening Profile F to the point that the pressure side bubble was just suppressed. The pressure surface was not modified close to the leading edge and trailing edge so as not to modify the overall behaviour of the profile. For the purposes of testing, Profile G was created by adding metal inserts to Profile F to fill in the profile on the pressure side. Special attention was paid to the junction near to the leading edge so that the boundary layer was not tripped in this area. Three different incidences were tested ( 0, +10 and 20 ) to study the off-design behaviour. This range is representative of some operating conditions in the turbine. Instrumentation: The stagnation temperature at inlet to the cascade was measured using a thermocouple that was placed in the upstream plenum. The inlet stagnation pressure was measured upstream of the leading edge of the blades. A Pitot probe was placed at mid-pitch 33% C ax upstream of each blade passage. Static pressure tappings were located in the same positions but in the opposite side-wall. The average values of inlet static pressure and inlet stagnation pressure were determined using the values provided by the above instrumentation. Static pressure tappings were also placed at mid-pitch behind each blade passage at 25% C ax and 50% C ax downstream of the trailing edge plane. One of the central two blades is instrumented with static pressure tappings at mid-span. A total of forty four tappings were used to measure the static pressure distribution over the surface. The locations of the measuring points are shown in figure 2. The tappings were

4 placed closer together on the suction side near to the leading edge in order to detect if a separation bubble is formed at positive incidence. Similarly, the tappings were placed closer together in the region of the separation bubble that was expected to form downstream of the throat on the suction surface in order to locate the separation and reattachment points. 1 TAPPING PROFILE G PROFILE F MISES V/V2 %Cax 0 Figure 2. Static pressure tapping location & isentropic surface velocity distributions: Profile F and Profile G at Re d. Downstream of the cascade, a 4-hole Neptune probe was used to measure the exit flow field. The probe was operated in a fixed orientation with its axis parallel to predicted flow direction. The local mean flow angle, static pressure and stagnation pressure were determined from the calibration of the probe. Integration of these local values was then carried out and a constant area mixing calculation was used to provide the mixed-out values of the cascade loss, exit flow angle and exit velocity. The traverse plane was located 0.25 C ax downstream the trailing edge plane of the cascade. Some data were also acquired 0.5 C ax downstream although they are not presented in this paper because there is no significant difference between the mixed out values at the two locations. 4. RESULTS & DISCUSSION. Profile F and Profile G were tested over a range of chord-based exit Reynolds numbers from to and at three different incidences (0, +10 and 20 ) under steady-state flow conditions. Unsteady measurements were conducted to study the suction surface behaviour but they are not included in the paper. Inlet turbulence level in the experiments is given by the characteristics of the tunnel, around 0.5%.

5 Isentropic surface velocity distributions: The static pressure data are presented in terms of the normalised velocity coefficient (V/V 2 ). In the majority of the figures, the data are plotted against the normalised surface length (%s) since it is the development of the boundary layers that is of particular interest. Figure 2 presents the isentropic surface velocity distributions for Profile G and Profile F plotted against the percentage of axial chord at the design representative Reynolds number of approximately (Re d ). These profiles were designed following the high lift design philosophy developed in former studies [2], [8]. The experimental results verify that the profiles fulfil the intentions of the design. There is a smooth acceleration over the leading portion of the suction side. The maximum Mach number is located close to 70% C ax. The strong deceleration on the suction surface leads to a separation of the laminar flow. Transition occurs in the separated flow region and as a result, the flow reattaches before the trailing edge. Figure 2 also shows that both profiles have essentially the same suction surface velocity distribution. The only noticeable difference in the suction side behaviour is that the velocity is slightly higher on the front part of the suction surface of Profile G as a result of the blockage caused by the thickening of the profile on the pressure side. On the pressure surface, the behaviour of the profiles is very different as was intended. In addition to the measurements, figure 2 also presents a prediction, obtained using the Mises code [8] for Profile G. This shows that there is a good agreement between computational and experimental data. The experiments show that for profile F there is a very large separation bubble which reattaches at about 70% C ax. This large separation bubble does not exist when the pressure side is filled in. V/V 2 1 Re=1.0E5 Re=2.2E5 Re=3.0E5 PROFILE G 0 PROFILE F %s Figure 3. Isentropic surface velocity distributions at three Reynolds numbers.

6 Figure 3 presents the isentropic velocity distributions of both profiles at three chord-based exit Reynolds numbers. Tests were conducted at about eight Reynolds numbers, but only three are shown for the sake of clarity. It has already been shown that the suction side distributions are very similar for both profiles. Therefore, the suction side distributions for only one profile have been plotted in this figure. The Reynolds number affects the evolution of the boundary layer on the suction surface and in particular the characteristics of the suction side bubble. Increasing the Reynolds number reduces the length to transition and causes earlier reattachment of the separated shear layer. While the reattachment point varies its location, the separation point essentially remains constant as expected from theoretical considerations. On the pressure surface, it seems that there is no noticeable effect of the Reynolds number on the behaviour of the separation bubble. In all cases, separation has occurred before the first measurement point and reattachment appears to take place as the free-stream flow begins to reaccelerate toward the trailing edge. 1 V/V % s degrees 0 degrees -20 degrees Profile G Profile F PRESSURE SURFACE SEPARATED REGIONS Profile G: +10 No bubble +0 No bubble % S Profile F: % S % S % S Figure 4. Isentropic surface velocity distributions at three incidences for Profile F and G. Figure 4 shows the isentropic velocity distributions of profiles F and G at different inlet flow angles. The incidence influences both the pressure and the suction sides. Three different incidences were tested (0, +10 and 20 ). Only the results obtained at the design Reynolds number are presented. The results at other Reynolds numbers are very similar to those presented in figure 4. As is usually the case, changing the incidence affects the suction side velocity distribution over the front part of the aerofoil between the leading edge and the location of maximum velocity. It also affects the pressure side by reducing the deceleration of the flow and so the size of the pressure bubbles when it is positive and increasing or

7 sometimes generating pressure side bubbles if it is negative. There is always a pressure surface bubble in Profile F. At an incidence of 20, this separation bubble is extremely long. Its extent is around 75 percent of the surface length, which is more than 80 percent of its axial chord. At +10 of incidence, the size of this bubble is reduced on Profile F but it is not suppressed. No bubble appears for Profile G when operating at the design incidence or at positive incidence. Operation at negative incidence provokes the separation of the flow on the pressure surface but the separation bubble is much shorter than on Profile F. Profile F Profile G Trend lines are curve fits to the experimental data 0, Reynolds number 3,3 Figure 5. Profile F and G design incidence losses distributions: (a) Variation of stagnation pressure loss coefficient with Reynolds number. (b) Stagnation pressure loss coefficient profile downstream of the cascade at Re d. 0 %Pitch 2 +10º 0º -20º Trend lines are curve fits to the experimental data Reynolds number %Pitch Figure 6. Profile G: Incidence influence in losses: (a) Variation of stagnation pressure loss coefficient with Reynolds number. (b) Stagnation pressure loss coefficient profile downstream of the cascade at Re d. Losses: Figures 5,6 and 7 summarise the loss data obtained for the two cascades. In each figure, the variation of the profile loss coefficient with the chord based exit Reynolds number is shown together with the pitchwise variation of loss downstream of the cascade at Re d. Figure 5(a) shows how the stagnation pressure loss coefficient of Profile F and Profile G varies with Reynolds number at the design incidence. Both trend lines are almost parallel. As the Reynolds number is reduced, the stagnation pressure loss coefficients of both profiles increase, as is usually the case. At the lowest Reynolds number, the velocity distributions of figure 3 indicate that the suction side boundary layer is still attached at the trailing edge. As the Reynolds number is increased, the reattachment of the suction side separation bubble occurs further from the trailing edge and the losses are substantially lower. Figure 3 also shows that the laminar length of this separation bubble does not change very much between a Reynolds number of 2.2x10 5 and 3x10 5. Over the same range of Reynolds numbers, the stagnation pressure loss coefficients are almost constant.

8 The only physical difference between Profile F and Profile G is the shape of the pressure surface as shown in figure 2. Therefore, if there is any difference in the losses of the two profiles, it must be due to the changes made to the pressure surface. It has already been noted that the suction side velocity distributions are subtly different as a result of the different geometries and pressure side blockage. However, figure 5(b) shows that this is not only reason for the difference in the losses of the two profiles. This plot shows how the stagnation pressure loss varies with pitchwise distance at the design Reynolds number. The wakes from the two central blades of the cascade are presented. The pressure side of each wake is to the right. The plot clearly shows that there is a so-called "loss tail" on the pressure side of the wake extending into the freestream in the case of Profile F. An assessment of the losses in this region reveals that they are of a similar order to the differences in the loss coefficients of the two profiles. Figure 5(a) shows that the stagnation pressure loss coefficients of Profile G are approximately 90 percent of those of Profile F at a wide range of Reynolds number around Re d as it is shown in figure 8. This difference is reduced to 5% at the lowest Reynolds number. The very weak dependence on Reynolds number arises because, as figure 3 shows, changing the Reynolds number does not significantly alter the characteristics of the pressure side separation bubble. +10º 0º -20º Trend lines are curve fits to the experimental data Reynolds number %Pitch 2 Figure 7. Profile F: Incidence influence in losses: (a) Variation of stagnation pressure loss coefficient with Reynolds number. (b) Stagnation pressure loss coefficient profile downstream of the cascade at design Re d. Figures 6 and 7 present the off-design behaviour of Profile G and Profile F. The variation of the stagnation pressure loss coefficient with the chord-based exit Reynolds number is presented in Figure 6(a) and figure 7(a). The wake of profiles are shown in the lower plot of each figure. Figure 6 presents the results for Profile G at the three incidences. Figure 7 presents the results for Profile F. Both figures show that, for a given profile, there is no significant change in the stagnation pressure losses between zero and positive incidence. In fact, the losses of Profile G increase very slightly whereas those of Profile F remain constant. This is because the increased loss that arises from changes on the suction surface is offset by the reduction in the length of the pressure side bubble as shown in figure 4. In the case of Profile G, as the Reynolds number is reduced, the losses increase at positive incidence more than at zero incidence. As a result, the differences between the two profiles are reduced especially at positive incidence. When both profiles are operated at negative incidence, there is a substantial increase in the profile losses of about 30%. This is a direct consequence of the pressure side separation bubble which now exists on Profile G and which has grown much larger in extent on Profile F. The loss tail associated with this additional loss is very clearly visible in the wake profiles. This relatively large deterioration in performance at -20 of incidence means that the inclusion of hollow aerofoils in the LP turbine will not lead to an unexpected over-speed problem ( in case of shaft failure a higher efficiency leads to a larger

9 terminal speed ). Hodson and Dominy [5] have also reported the presence of a significant pressure side loss tail behind a cascade of LP turbine blades at -20 of incidence. (lossesf-lossesg)/lossesf % ,0E Reynolds number 3,3E+05 Figure 8. Difference in stagnation pressure loss coeficient between Profile F and Profile G vs. Reynolds number. Future work could include the testing of these profiles when they are subjected to the simulated wakes of an upstream bladerow. The unsteadiness which arises is known to affect the suction surface boundary layer development, thus modifying the losses [1],[2],[4]&[7]. However, experiments of this type and cold flow rig data indicate that the unsteadiness has no significant effect on the static pressure distribution along the pressure surface. Therefore, it is expected that the results from these steady-state experiments are directly applicable to the engine environment. 4. CONCLUSIONS Two sets of cascades, a thin-solid one ( Profile F ) and a thick one ( Profile G ) were manufactured to study the pressure side influence in losses. Both of them have physically identical suction surfaces but different pressure surfaces; the pressure surface were modified without modifying the suction surface aerodynamic behaviour. As a result, the pressure side separation bubble which exists on the thin-solid profile at zero and positive incidence has been suppressed. Suppressing this pressure surface bubble, without modifying anything else, reduces the profile losses by approximately 10%. Due to the large aspect ratios existing in LP turbines, the profile loss is the largest portion of the total loss. If profile loss is reduced by 10%, the target of increasing efficiency is achieved. The result is approximately a 0.5% improvement in efficiency. The off-design behaviour of the profile is not penalised by the thickening of the profile. At negative incidence, the relative increase in the losses is similar for both profiles. At positive incidence, the difference in the losses of the two profiles is slightly reduced especially at lower Reynolds numbers were losses of Profile G increase more.

10 5. ACKNOWLEDGEMENTS The authors wish to thank all of the staff at the Whittle Laboratory and particularly T.Chandler for their help and also ITP for its support of the project and the permission to publish this paper. 6. REFERENCES [1] Banieghbal, M.R., Curtis, E.M., Denton, J.D., Hodson, H.P., Huntsman, J., Schulte, V.. (1995). Wake passing in LP turbines. Paper No.23, AGARD Conference. Loss Mechanisms and unsteady flows in turbomachinery, Derby, May. [2] Curtis, E.M., Hodson, H.P., Banieghbal, M.R., Howell, R.J., and Harvey, N.W.. (1997), "Development of Blade Profiles for Low Pressure Turbine Applications", ASME Jnl. of Turbomachinery, Vol 119, Jul. [3] Denton, J.D.. (1999), State of the art and future of turbine technology. Proceedings of the International Gas Turbine Congress. Kobe. Pg [4] Engber, M., Fottner, L.. (1995), The effect of incoming wakes on boundary layer transition of a highly turbine cascade. Paper No.21, AGARD Conference. Loss Mechanisms and unsteady flows in turbomachinery, Derby, May. [5] Hodson, H.P., Dominy, R.G.. (1987), "The Off-Design Performance of a Low Pressure Turbine and Cascade", ASME Jnl. of Turbomachinery, Vol. 109, Apr. [6] Harvey, N.W., Schulte, V., Howell, R.J., and Hodson, H.P.. (1999), "The Role of Research in the Aerodynamic Design of an Advanced Low Pressure Turbine ", 3 rd European Conf. on Turbomachinery, IMechE, London, Mar. [7] Howell, R.J., Ramesh, O.N., Hodson, H.P., Harvey, N.W., Schulte, V.. (2000), "High Lift and Aft Loaded Profiles for Low Pressure Turbines", ASME Paper No 2000-GT-0261, ASME Turbo Expo 2000, Munich, May. [8] Schulte, V., and Hodson, H.P.. (1998), Unsteady wake-induced boundary layer transition in high lift LP turbines, ASME Jnl of Turbomachinery, Vol 120, Jan [9] Giles, M., and Drela, M.. (1995) Two Dimensional Transonic Aerodynamic Design method, AIAA Journal, Vol.25, No9, Pg