You are on page 1of 7

Xinqian Zheng Yangjun Zhang

e-mail: zhengxq@tsinghua.edu.cn State Key Laboratory of Automotive Safety and Energy, Tsinghua University, Beijing 100084, China

Separation Control of Axial Compressor Cascade by Fluidic-Based Excitations


Flow separation control was investigated on a compressor cascade using three types of uidic-based excitations: steady suction, steady blowing, and synthetic jet. By solving unsteady Reynoldsaveraged NavierStokes equations, the effect of excitation parameters (amplitude, angle, and location) on performance was addressed. The results show that the separated ow can be controlled by the uidic-based actuators effectively and the time-averaged performance of the ow eld can be improved remarkably. Generally, the improvement can be enhanced when the amplitude of excitation is increased. The optimal direction varies with each type of excitations and is related to physical mechanisms underlying the separation control. For two types of steady excitations, the most effective jet location is at a distance upstream of the time-averaged separation point and the synthetic jet is just at the separation point. DOI: 10.1115/1.4002407

Weidong Xing Junyue Zhang


National Key Laboratory of Diesel Engine Turbocharging Technology, Datong, Shanxi 037036, China

Introduction

In the modern aero-engine design, the tendency toward weight reduction and pressure ratio enhancement leads to stronger adverse pressure gradient in axial compressors, making unsteady ow separation appears easily. The separated ow constrains both efciency and pressure ratio. Therefore, it is increasingly urgent to address the control of ow separation in axial compressor. Almost all the targets of separation control focus on accelerating the low momentum uid in boundary layer on adverse pressure gradient, notwithstanding the methods are different. There has been extensive work in research and development on separation prevention 1,2. These methods include surface-based actuators, e.g., vortex generators, riblets, slotted aps, and uidic-based methods, e.g., steady suction, steady blowing, and synthetic jet. Most of the ow control research was focused on the single airfoil 3,4. Several research groups have begun to investigate the application of ow control to compressor blades by using uidicbased methods. Merchant et al. 5,6 designed and tested an aspirated compressor stage, which achieved a total pressure ratio of 3.4 at a tip speed of 457.2 m/s with a suction ow rate of 3.5% of the inlet mass ow. Schuler et al. 7 designed an aspirated fan stage for pressure ratio of 1.6 at a low tangential tip Ma of 0.7. The mass ow was removed through slots on the suction surface of the rotor/stator and the rotor shroud and stator hub. Steady blowing was carried out on a highly loaded compressor cascade 8. Both the prole and the corner boundary layer could be controlled, and the performance was improved. Culley et al. 9 controlled the stator vane suction surface separation in a low-speed multistage compressor using steady and unsteady injection on the suction surface. These results suggested that air injection from the suction surface was relatively ineffective in mitigating endwall loss. Furthermore, a number of papers investigating an approach for tip critical instabilities are focused on injection 10,11. The referred research contributed to the understanding of ow separation control in axial compressors 12. However, to the authors knowledge, little study has been performed on separation control comparatively by using various types of uidic-based methods in axial compressors. In the present study, three types of uidic-based methods, steady suction SS, steady blowing SB,
Contributed by the International Gas Turbine Institute IGTI of ASME for publication in the JOURNAL OF TURBOMACHINERY. Manuscript received October 3, 2007; nal manuscript revised February 12, 2010; published online April 21, 2011. Editor: David Wisler.

and synthetic jet SJ, were investigated regarding their ability to control the separation inside an axial compressor cascade by solving the unsteady NavierStokes equations. In view of complex ows in axial compressors, two-dimensional case was considered only in this paper. Specically, ow conditions were xed at Ma= 0.5 and i = 10 deg.

Numerical Methods

2.1 Discretization Methods. The code used in this simulation was based on a 2D unsteady compressible nite difference scheme to solve the Reynolds-averaged NavierStokes equations in conservative formulation. A third-order TVD RungeKutta timemarching scheme was used for time integration because time accuracy was critical for this simulation. It was generally recognized that high-order spatial schemes were recommended especially for time-dependent simulations. A fth-order accurate, generalized compact scheme 13 was used for spatial difference of inviscid uxes and a sixth-order compact nite difference scheme 14 for viscous uxes. For turbulence model, the standard algebraic model of Baldwin and Lomax was adopted 15. 2.2 Grid and Initial Value. In the present paper, an H-type grid was adopted, which was generated by solving relevant elliptic differential equation. The grid inuences were analyzed to make sure that the computation results were independent of spatial discretization. The inuences of different factors were analyzed, such as the forward and backward extensions of the computational domain, the number of grid points, and the local densication of grid points. After a thorough consideration of compromise between computation work and computation precision, the grid adopted in the present paper was as follows. The computational domain covers six chord lengths, one chord length before the cascade and four chord lengths behind the cascade; the number of grid points was 321 axial direction 65 circumferential direction. For improving computation precision, grid resolution was rened locally near the leading and trailing edges and the cascade walls, in particular, within the boundary layers, the rst level grid points above the wall were located at y 1+ 10. The problem how to select initial values had also been studied. Commercial software NUMECA was rst used to obtain a converged steady solution, which was then taken as the initial oweld for our unsteady simulation. This greatly accelerated convergence of the unsteady simulation. For different initial ow-elds, the unsteady simulations could invariably lead to periodic or quaOCTOBER 2011, Vol. 133 / 041016-1

Journal of Turbomachinery

Copyright 2011 by ASME

Downloaded 21 Jun 2011 to 166.111.50.173. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm

cations of blowing and suction. Periodic conditions were imposed on the boundaries of the extended domains before and behind the cascade. 2.4 Validation of the Program Reliability. Validation of the simulations reliability is of key importance in numerical simulation. Comparisons are shown in Fig. 1 for static pressure distribution on the prole surface between the measurements obtained in the planar cascade experiments and the relevant simulation results time-averaged values. Overall, good agreement is obtained between the experimental and the simulation results, indicating the reliability of the simulation setup. However, there are some differences, e.g., the simulation results are somewhat smaller than measurements at the points behind the position of 50% chord. The main reasons can be explained as follows: 1 Simplied simulation modeleven though planar cascades are nonrotating parts with weak 3D effects, however, they are actually 3D straightblade cascades with end-wall, whereas the simplied simulation model is a strictly 2D one. Hence, the determinative conditions of physics and mathematics have differences between numerical simulations and experiments. 2 B-L model was adopted as turbulence model. This model is widely used in engineering applications; however, practice shows that it is always some differences between the simulation and experiment results.

Fig. 1 Comparisons of static pressure distribution on prole surface between the measurements and the simulation results

siperiodic converged solutions, and the inuences on the timeaveraged performances were also small. The inuence of the initial values lies mainly in the converging speed. Data were recorded only when the computation was converged to a periodic or quasiperiodic solution. Because for this kind of massively separated and unsteady ow the initial transient stage is quite random and has a profound effect on the detailed uctuation history of the ow, it is impractical to reach an exact grid and initial value independence. Overall inaccuracy due to the grid and initial solution was smaller than 0.5%. 2.3 Boundary Conditions. The present paper adopted the approach of modifying the boundary conditions on the prole surface to simulate the uidic-based actuators. The following velocity at the excitation location was imposed: uwall = Vex cos2 f et cos
vwall = Vex cos2 f et sin

Simulation Results

The modeled compressor cascade geometry is as follows: blade chord C = 80.6 mm, solidity = 1.5, inlet camber angle 1K = 41.6 deg, exit camber angle 2K = 1.23 deg, and stagger angle = 19.88 deg. The imposed uidic-based excitations were assumed to be completely 2D and the jet slot width is about 5% of the chord length. Figure 2 shows the main parameters involved. The total loss coefcient is chosen as the indication of performance. Taking into account the mass and kinetic energy of the jet ow, the total pressure loss coefcient is dened as =
min pin + mex p ex min + mex pout min pin pin + mex p ex pex

in which is the excitation direction angle, i.e., = 0 deg means the excitation is tangential to the suction surface and pointing to the trailing edge, and = 90 deg means excitation is normal to the suction surface. Equation 1 denotes the method of synthetic jet when f e 0; f e = 0 means the methods of steady excitations, i.e., steady suction 0 and steady blowing 0, respectively. Steady boundary conditions were given at the inlet and outlet boundaries. Total temperature, total pressure, and incidence were given at the inlet, whereas only the static pressure was xed at the outlet while velocity and density could be derived by extrapolation. Nonslip, impermeability, and adiabatic boundary conditions were imposed on all the prole surface points except for the lo-

where p in and pout are the mass-averaged total pressure of inlet and outlet of computational domain, respectively. pin is the massaveraged static pressure of inlet section. In Eq. 2, the denominator means the total kinetic energy entering the channel, and the numerator represents the total pressure loss. According to Eq. 2, the total pressure loss coefcient without 0 = 0.237 and the separation point on uidic-based excitation is the suction side is about 26.2% of chord length from the leading edge. The time-averaged performance is evaluated by using nor / 0. malized total pressure loss coefcient

Fig. 2 Cascade geometry and excitations parameters

041016-2 / Vol. 133, OCTOBER 2011

Transactions of the ASME

Downloaded 21 Jun 2011 to 166.111.50.173. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm

Fig. 3 The loss coefcient for different relative steady suction amplitude = 90 deg, l = 15.6%

Fig. 5 Loss coefcient for different steady suction location = 8%, = 90 deg A

For the three types of excitation methods, the jetted mass ow is as follows: mex =0 0 3.1

steady blowing synthetic jet or unexcitation steady suction

the suction mass ow rate reaches its maximum at the suction angle of 90 deg when other suction parameters are xed. The results can be proved herewith that a certain amount of the suction ow rate is the prerequisite of obtaining a positive suction effect. 3.1.3 Effect of Location. Figure 5 gives the relation between different suction locations and total pressure loss coefcients. Loss coefcient can be reduced by about 20% when 7% l 30%. However, the improvement of performance increases slightly when suction location is beyond 60% chord length. Particularly, when steady suction is imposed at the location of 15.6% chord length instead of the separation point l = 26.2%, the optimal effect could be obtained with a reduction by 28.3%. At this location, separation does not yet occur and the ow is still attached. However, the low momentum has been accumulated at this location. It is at this location that the ow could keep attached if low momentum uid is sucked or high streamwise momentum uid is injected, thereby delay or restrain the downstream separation. 3.2 Steady Blowing

Steady Suction

3.1.1 Effect of Amplitude. For a given suction angle and location, Fig. 3 shows the relation between suction amplitude and loss coefcient. Results give that improvement of performance is directly related to the suction amplitude. Even a small increase of the suction amplitude will result in an immense improvement in = 0.73%, the performance. As shown in Fig. 3, when A = 10% m loss coefcient can be reduced by 31.2%. Therefore, for steady suction, the separated ow could be under effective control without high suction amplitude. 3.1.2 Effect of Angle. Figure 4 gives the variation of loss coefcient as suction angle varies at two locations of l = 26.2% time-averaged separation point and 29.9%. The results indicate that, for a given suction amplitude A = 8%, the optimal efciency can be obtained at = 90 deg, and other angles are symmetric about the center of 90 deg. It can be explained by the fact that

3.2.1 Effect of Amplitude. Three locations were selected on the suction surface, one near to the leading edge l = 1.75%, one at a distance upstream of the time-averaged separation point l = 15.6%, and another near to the separation point l = 26.2%. The blowing angle is xed, and the blowing ow rate is controlled to 0.25%. The results at the three locations are given by Fig. 6. m The loss coefcient of the oweld generally decreases with an increase of the blowing amplitude. 3.2.2 Effect of Angle. Figure 7 shows the relation between blowing angle and cascade performance. The effectiveness of steady blowing control is in a wide range. For a given blowing amplitude, the blowing ow rate will increase with an augmented angle, and blockage is likely to appear, which leads to decrease the static pressure ratio. Blockage is present when 30 deg and the results is not given accordingly. The effectiveness of tangential blowing 30 deg is similar to that of the normal suction for the reason that when blowing tangentially, i.e., blowing streamwise, high momentum uid can fully join with the boundary layer ow, making low momentum uid in boundary layer accelerate on the one hand, and avoid serious disturbance to main ow on the other hand. At the same time, for a given slot width, tangential blowing brings high moOCTOBER 2011, Vol. 133 / 041016-3

Fig. 4 The loss coefcient for different steady suction direc = 8% tion A

Journal of Turbomachinery

Downloaded 21 Jun 2011 to 166.111.50.173. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm

Fig. 6 Loss coefcient for different relative steady blowing amplitude at three locations = 2 deg

Fig. 8 Loss coefcient for different steady blowing location = 90%, = 2 deg A

mentum and low mass ow rate injection, the blockage brought with it will not decrease the static pressure ratio while diminishing the separated region. 3.2.3 Effect of Location. Figure 8 shows the effect of the location of excitation on the performance. Again, loss coefcient can be reduced by about 20% when 7% l 30%. Particularly, when steady suction is imposed at the location of 15.6% chord length instead of the separation point l = 26.2%, the optimal effect can be obtained with a reduction by 29.1%. In addition, seen from Figs. 5 and 8, it is surprised that the similar results are obtained with the excitations by steady suction and steady blowing. 3.3 Synthetic Jet. Steady suction and blowing are belong to steady methods. Synthetic jet, the characteristic of zero net mass ux and nonzero momentum ux, is a typical unsteady method. The main difference between steady and unsteady methods is the importance of excitation frequency. As this paper focuses on the effect of amplitude, angle, and location, the role of excitation frequency can be referenced in Ref. 16. When the excitation frequency was approximately equal to the characteristic frequency of vortex shedding, i.e., f = 1, the optimal effect of unsteady excitation on separation ows can be obtained. Therefore, relative excitation frequency is chosen at f = 1.

3.3.1 Effect of Amplitude. Figure 9 shows the effect of excitation amplitude on the loss coefcient. It is demonstrated that the periodic excitation has a negligible effect on the ow dynamics when the excitation amplitude is relative small. Therefore, positive results obtained must satisfy the condition that the relative excitation amplitude is above a certain threshold value A 10%. Basically, the aerodynamic performance increases monotonically with excitation amplitude up to about A = 35%. If the excitation amplitude goes beyond the limit A = 35%, the effect of periodic excitation is weakened slightly. 3.3.2 Effect of Angle. Figure 10 shows the effect of excitation angle on the loss coefcient. The same best positive excitation effects are obtained in the range of 40 deg 140 deg. Beyond this range, the excitation effects decrease sharply. However, up to the extreme case of = 0 deg or = 180 deg, i.e., the excitation direction tangent to the suction surface, positive effects could be still obtained. There is another typical feature, the curves in Fig. 10 almost take the normal to the suction surface of = 90 deg as their symmetry axis, their left parts and right parts are generally symmetric. Therefore, the best positive excitation effects are obtained when the excitation direction is normal to the suction surface and the positive effects remain invariability in a rather wide range of . The nature of synthetic jet is periodic suction and blowing, and

= 90%, Fig. 7 Loss coefcient for steady blowing direction A l = 26.2%

Fig. 9 Loss coefcient for different synthetic jet amplitude f = 1, = 90 deg, l = 26.2%

041016-4 / Vol. 133, OCTOBER 2011

Transactions of the ASME

Downloaded 21 Jun 2011 to 166.111.50.173. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm

Fig. 10 Loss coefcient for different synthetic jet direction f = 1, l = 26.2%, A = 40%

a combination of steady suction and pulsed jet in temporal characteristic. Not only can it suck out the low momentum uid in boundary layer, but also it can facilitate the interaction of unsteady vortices. Therefore, synthetic jet is not sensitive to the jet angle. 3.3.3 Effect of Location. Figure 11 shows the loss coefcient for different excitation locations. The best results were obtained when synthetic jets are excited just at the time-averaged separation point. If the excitation location moves backward or forward further from the time-averaged separation point, the positive effect of excitation is reduced slightly. Generally, the positive results are obtained in a wide range of excitation location, which is very important for engineering application because the separation point is varied with the ow conditions. It is speculated that the boundary layer close to the separation point is sensitive to disturbance. Just around this specic local region, the boundary layer is receptive to unsteady excitation. The basic steady ow has a high velocity component upward at the local region near separation point, which will amplify the effect of any nonlinear streaming and thus make local oscillations an effective actuator in delaying separation on the blade. This amplication mechanism, which cannot be found in attached ows, will enhance the primary entrainment and rolling-up coalescence inherent in boundary layer, resulting in reduced separation region.

Fig. 12 Instantaneous contour plots for Ma, and total pressure plus streamlines. Baseline ow.

Flow Field Structure

The ow eld without excitation, i.e., the baseline ow, is shown in Fig. 12. It is obvious that a strong velocity gradient divides the oweld into main and separated ow. There is a sequence of vortices generating, developing, and shedding in the separated ow. The presence of a large scale separated vortex deteriorates the diffusion capacity, and loss increases. Figures 13 and 14 are the ow structures imposed by steady suction and steady blowing, respectively. Compared with Fig. 12, the strong velocity gradient weakens, and the ow velocity increases. The separated region is reduced remarkably, the scale of separated vortex diminishes, and the streamlines become smoother. The oweld structure by synthetic jet can be referenced in Ref. 16.

Concluding Remarks

= 1, A Fig. 11 Loss coefcient versus synthetic jet location f = 30%, = 90 deg

For this paper, investigations on separation control in an axial compressor cascade have been carried out using three types of uidic-based excitations: steady suction, steady blowing, and synthetic jet. The focus is on the inuence of time-averaged aerodynamic performances when the amplitude, angle, and location of excitation are varied. The results can be summarized as follows: 1 The effectiveness of above three types excitation varies OCTOBER 2011, Vol. 133 / 041016-5

Journal of Turbomachinery

Downloaded 21 Jun 2011 to 166.111.50.173. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm

Fig. 13 Instantaneous contour plots for Ma, and vorticity plus = 8%, = 90 deg, streamlines. Steady suction A l = 15.6%.

Fig. 14 Instantaneous contour plots for Ma, and vorticity plus = 75%, = 2 deg, streamlines. Steady blowing for A l = 15.6%.

Acknowledgment
signicantly with the excitation parameters e.g., jet angle, location, amplitude, or injected ow rate. As a whole, the positive effect is more remarkable when the excitation amplitude is increased. 2 For a given suction amplitude and location, normal suction = 90 deg can maximize the suction ow rate and obtain the best effect. The steady blowing energizes the low momentum region directly, so a tangential angle is effective. Synthetic jet is not sensitive to the jet angle and the positive effects remain invariability in a rather wide range about 40 deg 140 deg. 3 The optimal location of steady excitation i.e., steady suction and steady blowing and unsteady excitation synthetic jet are different. The optimal location of steady excitation is at a distance l = 15.6% upstream of the time-averaged separation point l = 26.2%. The optimal location of unsteady excitation is just at the time-averaged separation point and positive results can be obtained in a wide range of excitation locations. 041016-6 / Vol. 133, OCTOBER 2011 This research was supported by the National Natural Science Foundation of China Project No. 50806040. The authors are deeply grateful to Zhou Xiaobo for her efforts to make the simulation possible. X.Z. would also like to thank Joern Huenteler from RWTH Aachen University for his great effort to correct the English grammar.

Nomenclature
A relative excitation amplitude A = Vex / c C chord of the cascade blade c free stream velocity f shed characteristic frequency of vortex shedding f e excitation frequency f relative excitation frequency f = f e / f shed l distance between L.E. and excitation location l normalized distance between L.E. and excitation location l=l/C time-averaged excitation ow rate m m = mex / min Transactions of the ASME

Downloaded 21 Jun 2011 to 166.111.50.173. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm

Ma p p t uwall vwall Vex x,y

Mach number static pressure total pressure time, blade pitch velocity in wall-tangent direction velocity in wall-normal direction excitation amplitude Cartesian coordinate in axial direction, Cartesian coordinate in circumferential direction excitation angle with respect to the direction tangent to the blade suction surface total loss coefcient

Subscripts ex excitations in inlet out outlet

References
1 Greenblatt, D., and Wygnanski, I. J., 2000, The Control of Flow Separation by Periodic Excitation, Prog. Aerosp. Sci., 367, pp. 487545. 2 Seifert, A., Bachar, T., Koss, D., Shepshelovich, M., and Wygnanski, I., 1993, Oscillatory Blowing: A Tool to Delay Boundary-Layer Separation, AIAA J., 3111, pp. 20522060. 3 Liu, Y., Sankar, L. N., Englar, R. J., Ahuja, K. K., and Gaeta, R., 2004, Computational Evaluation of the Steady and Pulsed Jet Effects on the Performance of a Circulation Control Wing Section, AIAA Paper No. 2004-56. 4 Wu, J. Z., Lu, X. Y., Denny, A. G., Fan, M., and Wu, J. M., 1998, Post-Stall Flow Control on an Airfoil by Local Unsteady Forcing, J. Fluid Mech., 371,

pp. 2158. 5 Merchant, A. A., Drela, M., Kerrebrock, J. L., Adamczyk, J. J., and Celestina, M., 2000, Aerodynamic Design and Analysis of A High Pressure Ratio Aspirated Compressor Stage, ASME Paper No. 2000-GT-619. 6 Merchant, A., Kerrebrock, J. L., Adamczyk, J. J., and Brqunscheidel, E., 2004, Experimental Investigation of A High Pressur Ratio Aspirated Fan Stage, ASME Paper No. GT2004-53679. 7 Schuler, B. J., Kerrebrock, J. L., Merchant, A. A., and Drela, M., 2000, Design, Analysis, Fabrication and Test of an Aspirated Fan Stage, ASME Paper No. 2000-GT-618. 8 Sturm, W., Scheugenpug, H., and Fottner, L., 1992, Performance Improvements of Compressor Cascades by Controlling the Prole and Sidewall Boundary Layers, ASME J. Turbomach., 114, pp. 477486. 9 Culley, D. E., Bright, M. M., Prahst, P. S., and Strazisar, A. J., 2004, Active Flow Separation Control of a Stator Vane Using Embedded Injection in a Multistage Compressor Experiment, ASME J. Turbomach., 126, pp. 2434. 10 Suder, K. L., Hathaway, M. D., Thorp, S. A., Strazisar, A. J., and Bright, M. B., 2001, Compressor Stability Enhancement Using Discrete Tip Injection, ASME J. Turbomach., 123, pp. 1423. 11 Bae, J. W., Breuer, K. S., and Tan, C. S., 2005, Active Control of Tip Clearance Flow in Axial Compressors, ASME J. Turbomach., 127, pp. 352362. 12 Zheng, X. Q., Zhou, X. B., and Zhou, S., 2005, Investigation on a Type of Flow Control to Weaken Unsteady Separated Flows by Unsteady Excitation in Axial Flow Compressors, ASME J. Turbomach., 127, pp. 489496. 13 Shen, M. Y., Niu, X. L., and Zhang, Z. B., 2001, The Three-Point Fifth-Order Accurate Generalized Compact Scheme and Its Applications, Acta Mech. Sin., 172, pp. 142150. 14 Lele, S. K., 1992, Compact Finite Difference Schemes With Spectral-Like Resolution, J. Comput. Phys., 103, pp. 1642. 15 Baldwin, B. S., and Lomax, H., 1978, Thin Layer Approximation and Algebraic Model for Separated Turbulent Flows, AIAA Paper No. 78-257. 16 Zheng, X. Q., 2006, Investigation on the Conditions of Flow Pattern Transform in Unsteady Vortex Flow-Field of Axial Compressors, Ph.D. thesis.

Journal of Turbomachinery

OCTOBER 2011, Vol. 133 / 041016-7

Downloaded 21 Jun 2011 to 166.111.50.173. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm

You might also like