Download - Control of Flow Separation in S-Ducts via Flow Injection ... · Control of Flow Separation in S-ducts via Flow Injection and Suction Marco Debiasi1, Marco Robert Herberg2, Zeng Yan

Control of Flow Separation in S-ducts via Flow Injection and Suction

Marco Debiasi1, Marco Robert Herberg2, Zeng Yan 3, Shyam Sundar Dhanabalan 4, and Her Mann Tsai5

Temasek Laboratories, National University of Singapore, Singapore, 117508, Singapore

This work explores the use of injection and suction as a mean to control the separation of flow in S-duct inlets. The overall goal is to reduce the distortion of the S-duct outlet flow and to improve its pressure recovery by using the least expenditure of energy. We also aim to understand to what extent computational means can be practically useful for optimizing this flow-control method. Computations and experiments were used to study the effect of injection and suction of 2% on the main stream in an S-duct of M2129 geometry. The control flow can be recirculated with advantages in term of mass flow and energy management. The benefit of the method is assessed by contrasting controlled flows to corresponding baseline (no-control) flows over a range of subsonic inlet conditions. The experimental measurements substantially validate the results from computation and indicate that the method is highly effective in controlling the flow separation.

Nomenclature Cp = static pressure coefficient = ( ) 2

21

iii Upp ρ− D = diameter of S-duct l = length of S-duct centerline m& = mass flow rate M = Mach number p = static pressure p0 = total pressure PR = total to ambient pressure ratio at the S-duct inlet ReD = Reynolds number based on the diameter s = S-duct centerline coordinate T = static temperature T0 = total temperature U = flow velocity ρ = flow density ζ = azimuthal angle Subscripts i = conditions at the S-duct inlet o = conditions at the S-duct outlet ∞ = ambient conditions

I. Introduction HE inlets of aircraft turbines are designed to decelerate the freestream while achieving a uniform flow distribution with minimum stagnation pressure losses at the compressor face. In some aircraft a straight inlet

T

1 Research Scientist, Temasek Laboratories, National University of Singapore, Singapore, Member AIAA. 2 Associate Scientist, Temasek Laboratories, National University of Singapore, Singapore. 3 Research Scientist, Temasek Laboratories, National University of Singapore, Singapore. 4 Associate Scientist, Temasek Laboratories, National University of Singapore, Singapore. 5 Senior Research Scientist, Temasek Laboratories, National University of Singapore, Singapore, Member AIAA.

American Institute of Aeronautics and Astronautics

1

46th AIAA Aerospace Sciences Meeting and Exhibit7 - 10 January 2008, Reno, Nevada

AIAA 2008-74

Copyright © 2008 by Temasek Laboratories - National University of Singapore. Published by the American Institute of Aeronautics and Astronautics, Inc., with permission.

diffuser is unfeasible or undesirable and S-duct geometries are used instead. Such shape can have considerable impact on the flow quality. The effects of the curvature of the S-duct centerline lead to cross-stream centrifugal pressure gradients which cause the boundary layer fluid to move in the direction of the pressure gradient thus inducing cross flows perpendicular to the main flow. Such “secondary flows” create undesirable non-uniform total pressure distribution on the engine compressor face. The net effect of the secondary flow and of the diffuser adverse pressure gradient means an increased possibility of local boundary layer separation and total pressure loss at the exit. The problem is exacerbated when high degrees of centerline curvature are imposed by space and weight limitations. In general, efficient diffusion within the range where the engine can tolerate non-uniformity of total pressure and flow distortion becomes challenging over a broad spectrum of flight condition.

Passive and active devices for flow-control in S-shaped inlets are reported in the literature. The use of spoilers was examined by Guo and Seddon1, whereas Sullerey et al.2 investigated the effect of boundary layer fences lined along the surface of the duct. By far the most successful passive means appears to be based on the use of vortex generator arrays.3 Extensive studies were made using different low-profile vortex generators4, vortex arrangements5, and tapered fin types.6 These studies show that vortex generators are effective but the orientation and configurations arrangements are suitable only for certain flow conditions. Configurations with large vortex generators are the most effective in reducing the total-pressure distortion but do not produce the greatest total-pressure recovery. Tests also show that the flow separation in the S-duct cannot be completely eliminated, although its extent can be reduced. Moreover in regimes were the control is not necessary, the vortex generators induce unwanted losses.

This leads to the use of active flow control methods whose effect can be switched off.7 Many of these approaches are based on the success of the vortex generators, whereby small slots of jets are introduced such as to mimic the vortex generators. Commonly termed as vortex generator jets, they are extensively studied.8-10 The review by Johnston and Nishi9 on vortex generator jets for separation control shows that pulsed rather than continuous jets can provide a more efficient use of the jet fluid with possible improvement gain in control. An alternative to vortex jet generators is the use of a wall jet whose benefits on separation control are well known.11 Wall jets are normally made by injecting fluid along a wall at a velocity higher than in the ambient flow. This provides momentum to the boundary layer which is on the verge of separating due to the adverse pressure gradient encountered, for instance, on the upper surface of an airfoil. The shear between the main free stream and the jet creates strong turbulence diffusion and mixing, thus providing the lateral transport of energy that allows the boundary layer to remain attached even at large angles of attack. The turbulent wall jet is an effective method for control of flow separation but like any other active flow control method requires direct energy expenditure.

A recent method that holds great promise in reducing the energy required in such process is the so called co-flow jet.12-15 The rationale is that a significant energy reduction can be achieved if the wall jet is supplemented with a suction port. Thus the control flow could be recirculated with advantages in term of mass flow and energy management and potential benefits to the efficiency of the overall airframe-propulsion system. The suction port also helps the flow to remain attached in spite of the severe adverse pressure gradient.

The objective of this work is to utilize computations and experimental measurements to assess the effectiveness of using injection and suction for control separation in S-duct inlets. We also aim to understand to what extent computational means can be practically useful for optimizing such active flow-control methods. Injection and suction are introduced via slots in an S-duct of M2129 geometry.16 The approach used in the current work differs from that of Kwong and Dowling17 who investigated the use of steady and unsteady injection in diffusers including the effect of feedback flow control. Their findings suggest that using both steady and unsteady injection together can offer good diffuser performance in terms of pressure recovery and reduced pressure oscillations. Our method also differs from that proposed by Ball18 which prevents separation by using a porous wall to remove the low momentum fluid, similar to the methods for laminar flow control applications. In our approach we adopt a suction slot so to eliminate the suction losses through a small perforation and to capitalize on the residual momentum of the sucked flow. The effect of the method is assessed by contrasting controlled flows to corresponding baseline (no-control) flows over a range of Mach numbers. The overall aim is to reduce the S-duct outlet flow distortion and to improve its pressure recovery by using the least expenditure of energy.

II. Flow Geometry and Conditions The geometry of the M2129 duct used in the computations and in the experiments was obtained as a continuous

variation of the cross-sectional shape normal to the duct center line, Fig. 1. The method used is an extension of that presented by Yaras and Grosvernor.16 The interpolation functions are parameterized to allow the description of complex cross-section shapes between the duct inlet and its round outlet.19 The M2129 duct has an outlet-to-inlet


2

area ratio of 1.4 to which corresponds an outlet-to-inlet diameter ratio Do/Di of 1.183. Cylindrical extensions are added to the inlet and to the outlet sections to avoid the effect of boundaries in the computations and to allow the placement of a rotating Pitot rake in the experiments.

Table 1 lists the nomenclature and the flow conditions considered for this study. We covered total-to-ambient pressure ratios PR at the inlet representative of moderate subsonic speed. The table provides the inlet and outlet flow Mach number, velocity, and Reynolds number based on the corresponding diameter. The mass flow rate is also given in the last column of the table. All quantities were calculated by using inviscid, one-dimensional theory. Thus the flow losses associated to viscosity and to the separation and recirculation inside the duct are neglected. This analysis nevertheless provides first cut values against which the computational and experimental results can be compared.

For control, a slot for tangential injection is placed on the lower wall of the inlet cylindrical extension just before the S-duct. The control flow is conveyed to the slot through a hemicycle converging nozzle, Fig. 2. A similar slot for suction is placed just past the S-duct outlet. In the current study the gap of the injection and suction slots is Di/42 and Do/12 respectively and both slots span an angle of 30° on the wall. The mass flow rate used for injection and suction in all control cases is approximately 2% of the main stream. Equaling the injection and suction flows allows maintaining the overall mass-flow-rate balance of the baseline flow.

III. Flow Computations

A. Grid Generation The structured volume grid of the S-duct with cylindrical extensions was generated with Gambit by

interpolating/extending the 2D grids corresponding to the inlet and the outlet sections, Fig. 2. For accurate representation of the viscous boundary layer, the inlet and outlet grids have been clustered close to the walls. Grid cells in the sloping sections of the S-duct have moderate angular distortion which did not cause any convergence difficulty. Unstructured grids were used for the injection and suction nozzles. The nodes of the injection and suction slots match the nodes of the densely spaced structured grid of the main stream. An enlargement of the unstructured grid on the injection nozzle is provided in Fig. 2.

B. Solver and Turbulence Model The Reynolds-averaged Navier-Stokes (RANS) equations were solved using the commercial code FLUENT. The

solver is based on an unstructured finite volume scheme. Inviscid fluxes were calculated using a second-order upwind scheme, whereas the viscous fluxes were evaluated using a second-order central-difference scheme. A second-order implicit scheme was used for iterating the unsteady equations in pseudo-time to steady-time solution.

The wall boundary layer was assumed to be turbulent, and the two-equation shear stress transport (SST) model of Menter20 is used here. The SST model employs a k-ω formulation in the inner region of wall boundary layers and switches to a transformed k-ε formulation in the outer region of boundary layers and in the free shear layer. The compressibility correction of Sarkar et al.21 is adapted. Other modifications include the addition of a cross-diffusion term in the ω equation and a blending function to ensure that the model equations behave appropriately in both the near-wall and far-field zones.20 Detailed information on the governing equations and the modifications of parameters for the shear-stress-transport model are given in Ref. 20.

The boundary conditions were imposed as follows: the total pressure and total temperature at the inlet were set to be p0,i = PR×p∞ and T0,i = T∞, respectively, with p∞ = 101325Pa, and T∞ = 300K. The walls were specified to be adiabatic with no-slip condition. The static pressure at the S-duct outlet was set to the ambient pressure. The SIMPLEC algorithm, based on the collocated grids, was used to treat the velocity-pressure coupling.

The convergence criterion for the residual of energy equation was set to be 10-6, and the convergence criteria for the residuals of momentum equations ( and k ω ) was set to be 10-3.

IV. Experimental Apparatus and Procedures In order to check and validate the results from computation, we designed and constructed a small, modular

facility for measuring the flow pressure distribution in the S-duct. The facility consists of a rapid-prototyped S-duct connected to a circular test section inside which a polar Pitot

rake rotates, Figs. 3 and 4. S-ducts of different geometries and terminating with a circular section of diameter Do=50.8mm (2in) can be attached to the flanged test section. Air at room temperature is supplied to the S-duct by a


3

stagnation chamber. A pressure transducer (Setra Model 208E) connected to the stagnation chamber allows control of the flow total pressure within ±1% of the nominal value.

The Pitot rake comprises 15 probes with external diameter 1.0mm and internal diameter 0.5mm. The probes are distributed on three arms at optimal positions to minimize their mutual interference and to obtain a dense and even distribution of surveyed points with a 360° sweep of the rake, Fig. 5. The arms and shaft of the probe rake are positioned 15mm downstream of the probe tips and produce a blockage of about 12% of the test section area. We deem this to have minor effect on the low-speed measurements of the current study. The Pitot probes are individually connected to Setra Model 264 pressure transducers mounted close to the rake in order to minimize the length of the tubing between each probe and the corresponding transducer. 1000 samples from each probe were acquired at each new angular position with a National Instrument 6014 board installed on a Dell Optilex GX270 personal computer. Incremental-step motion of the rake angular position is achieved by a microstepping motor drive controlled by the digital acquisition system. The flow Mach number and velocity were computed from the Pitot data by assuming that the static pressure and total temperature are uniform and equal to ambient at the plane of the probe tips.

V. Results and Discussion Figures 6-15 illustrate the distributions of the total-to-ambient pressure ratio p0/p∞ obtained from computations

and experiments for S-ducts without control (baseline) and with control by injection and suction. Figure 6 presents the results for the PR=1.02 baseline case. Computations show that due to the duct curvature, a cross-stream pressure gradient with sub-atmospheric pressure exists along the low wall of the duct, Fig. 6a. In this same region reverse flow and separation occur. An enlargement of the pressure distribution at the outlet of the cylindrical extension is given in Fig. 6b. This cross section exhibits an upper, crescent-shaped high-pressure zone accompanied by a low-pressure bubble close to the duct base. Such spatial distortion of the pressure is very undesirable and severely impacts the performance of turbine engines, increases noise and vibration, and shortens the life span of the engine components. The corresponding experimental measurements, Fig. 6c, show similar overall features. The maximum value of p0/p∞ from the experiments is 1.015, slightly lower than due and within the error with which we can control the total pressure of the flow. Also, the pressure gradient observed in the experiment is slightly smaller than that from computations.

Application of injection and suction to the PR=1.02 flow is presented in Fig. 7 showing that this technique is effective in attenuating the flow separation. The separation observed inside the suction diffuser is inconsequential to the purpose of this study. In the longitudinal plane a small but strong separation bubble is visible past the injection point. Overall, the low-pressure zone in the centerline plane has been significantly reduced, and the flow finally reattaches before the suction slot, Fig. 7a. As a result, the cross-sectional distribution of the total pressure becomes more uniform and quasi axisymmetric, Fig. 7b. The low-pressure bubble close to the lower wall has been actually replaced by a narrow, higher than PR bubble. Spreading the same control flow over a wall angle larger than 30° should avoid this high-pressure concentration while eliminating the residual low-pressure lobes still visible at its sides. The experiments substantially confirm the computations by indicating the recovery of an overall higher and more axisymmetric pressure distribution with high pressure projecting towards the bottom of the duct, Fig. 7c.

Similar results for the baseline and actuated cases at higher PR values are presented in Figs. 8-15. The computations of the baseline cases at higher PR indicate that sub-atmospheric pressure and flow separation occur along the lower wall of the S-duct, Figs. 8, 10, 12, and 14a. In the corresponding cross sections at the duct outlet a crescent-shaped high-pressure zone is accompanied by a low-pressure bubble below it, Figs. 8, 10, 12, and 14b. It should be noted that the center of this low-pressure bubble narrows and tends to lift up from the wall as PR increases. This is confirmed by the experimental measurements, Figs. 8, 10, 12, and 14c. These, however, record slightly lower values of the maximum pressure and a narrower pressure range than the computations. Injection and suction of 2% of the main stream is effective in attenuating the flow separation at higher PR. In all cases computations show that the low-pressure zone in the centerline plane has been significantly reduced and the flow reattaches before the suction slot, Figs. 9, 11, 13, and 15a, and the cross-sectional distribution of the total pressure becomes more uniform and axisymmetric, Figs. 9, 11, 13, and 15b. The low-pressure bubble is replaced by a high-pressure one which tends to narrow and weaken with increasing PR. The experimental results at PR = 1.04, Fig. 9c, reveal that injection and suction produce a quasi axisymmetric pressure distribution with higher values in the center that decrease toward the wall. The maximum value of p0/p∞ is 1.045, slightly higher than due and within the error with which we can control the total pressure of the flow. The measurements at PR = 1.06, 1.08, and 1.10, Figs. 11, 13, and 15c respectively, reveal a zone with small total pressure deficit toward the center of the duct. This is


4

surrounded by an almost closed annulus with total pressure close to the nominal value. In all cases the flow is more symmetric and has a better pressure recovery than the corresponding baseline. The results of injection and suction at PR = 1.04 - 1.06 further support the idea that spreading the same control flow over a larger wall angle should produce a more uniform and symmetric pressure distribution.

The current experimental data substantially validate the results from computation as most features and trends

captured by the latter are consistent with the physical measurements. Further refinement should allow the use of computations for exploring S-duct flows and for optimizing their shape and control techniques.

VI. Conclusion The purpose of this work is to explore the effect of flow injection and suction for reducing the flow separation

inside an S-duct diffuser and for producing a more uniform and higher pressure distribution at its outlet. To this aim we used computations and experimental measurements of the flow in S-ducts with the M2129 geometry tested at an inlet Mach number in the range 0.24-0.58. The experiments were used to validate the flow computations which would be a more flexible and convenient tool to explore and optimize different flow-control configurations. The grids generated for computation are suitable for detailed analysis of the baseline flows and of the effect of injection and suction. The computational results are substantially supported by the experimental measurements. The computations capture the fundamental characteristics of S-duct flows like the separation along the convex side of the S-duct bend and the crescent-shaped, high-pressure zone accompanied by a low-pressure bubble at outlet plane. Injection of 2% of the main flow before the first bend of the S-duct accompanied by suction of the same quantity past the second bend is an effective means to control the flow separation. Both the computations and the experiments show that the total-pressure distribution of the controlled flow emerging from the S-duct has more axisymmetric distribution and higher overall pressure than the uncontrolled cases. The results also suggest that a more uniform and symmetric distribution of the total pressure should be obtained by spreading the same control flow over a larger angle in the duct walls.

Future studies will consider different positioning of the injection/suction ports, e.g. closer to the S-duct slope, and the effect of increasing the azimuthal angle of the injection and of the suction slots. The most effective configurations will then be applied to S-ducts with more aggressive centerline curvature.

Acknowledgments We are grateful to Mr. Kim Seng Lim and Mr. Jonathan Tay for their assistance with preparing the experimental

setup.


5

References 1Guo, R. W. and Seddon, J., “The Swirl in an S-duct of Typical Intake Proportions,” Aeronaut. Q., Vol. 34, pp. 99–129. 2Sullerey, R. K., Mishra, S., and Pradeep, A. M., “Application of Boundary Layer Fences and Vortex Generators in

Improving the Performance of S-duct Diffusers,” Journal of Fluids Engineering, Vol. 124, pp. 136–142. 3Wendt, B. J. and Reichert, B. A., “Vortex Ingestion in a Diffusing S-duct Inlet,” Journal of Aircraft, Vol. 33, No. 1, pp.

149–154. 4Reichert, B. A. and Wendt, B. J., “Experimental Investigation of S-duct Flow Control using Arrays of Low Profile Vortex

Generator,” AIAA Paper 93-0018. 5Reichert, B. A. and Wendt, B. J., “Improving Diffusing S-duct Performance by Secondary Flow Control,” NASA Technical

Memorandum 106492. 6Reichert, B. A. and Wendt, B. J., “Improving Curved Subsonic Diffuser Performance with Vortex Generators,” AIAA

Journal, Vol. 34, No. 1, pp. 65–72. 7Sullerey, R. K. and Pradeep, A. M., “Effectiveness of Flow Control Devices on S-duct Diffuser Performance in the Presence

of Inflow Distortion,” Journal of Turbomachinery, Vol. 19, pp. 259–270. 8Johnston, J. P. and Nishi, M., “Vortex Generator Jets-Means for Flow Separation Control,” AIAA Journal, Vol. 28, No. 6,

pp. 989–994. 9Khan, Z. U. and Johnston, J. P., “On Vortex Generating Jets,” International Journal of Heat and Fluid Flow, Vol. 21, pp.

506–511. 10Sullerey, R. K. and Pradeep, A. M., “Secondary Flow Control Using Vortex Generator Jets,” Journal of Fluids

Engineering, Vol. 126, pp. 650–657. 11Gad-el-Hak, M. and Bushnell, D. M., “Separation Control: Review,” Journal of Fluids Engineering, Vol. 113, pp. 5-30. 12Zha, G., Paxton, C., Conley, C. A., Wells, A., and Carroll, B. F. “Effect of Injection Slot Size on High Performance Co-

Flow Jet Airfoil,” Journal of Aircraft, Vol. 43, No. 4, pp. 987-995. 13Zha, G. and Paxton, C. “A Novel Flow Control Method for Airfoil Performance Enhancement Using Co-Flow Jet,”

Applications of Circulation Control Technologies, AIAA Book Series, Progress in Aeronautics and Astronautics, Vol. 214, 2006, Chapter 10, pp. 293-314, Editors: R. D. Joslin and G. S. Jones.

14Zha, G.., Gao, W., Paxton, C., and Palewicz, A. “Numerical Investigations of Co-Flow Jet Airfoil with and without Injection,” AIAA Paper 2006-1061.

15Rachman, A., Tsai, H. M., and Zha, G. “Application of Co-Flow Jet Concept on Thin Airfoil at High Angle of Attack,” AIAA Paper 2006-2849.

16Yaras, M. I. and Grosvernor, A. D., “Evaluation of one- and two-equation low-Re turbulence models. Part II—Vortex-generator jet and diffusing S-duct flows,” International Journal for Numerical Methods in Fluids Vol. 42, pp. 1321–1343.

17Kwong, A. H. M. and Dowling, A. P., “Active Boundary-Layer Control in Diffusers,” AIAA Journal. Vol. 32, No. 12, pp. 2409–2414.

18Ball, W. H., “Tests of Wall Suction and Blowing in Highly Offset Diffusers,” Journal of Aircraft, Vol. 22, No. 3, pp. 161–167.

19Dhanabalan, S., Won, K., and Tsai, H. M., “S-shaped Intake Duct Parametrization,” AIAA Paper 2006-3317. 20Menter, F. R., “Ten years of industrial experience with the SST turbulence model,” Proceedings of the 4th International

Symposium on Turbulence, Heat, and Mass Transfer, (ICHMT), Antalya, Turkey, Oct. 2003, pp. 625–632. 21Sarkar, S., Erlebacher, G., Hussaini, M. Y. and Kreiss, H. O., “The Analysis and Modeling of Dilatational Terms in

Compressible Turbulence,” Technical Report 89-1789, ICASE, 1989.

Table 1: Values of inlet and outlet isentropic flow in M2129 S-duct with Do = 50.8 mm (2 in)

PR Mi Ui(m/s)

ReDix105

Mo Uo(m/s)

ReDox105

m& (kg/s)

1.02 0.24 83 2.27 0.17 59 1.92 0.1392 1.04 0.34 119 3.20 0.24 82 2.71 0.1966 1.06 0.43 147 3.92 0.29 100 3.31 0.2407 1.08 0.51 173 4.52 0.33 115 3.82 0.2777 1.10 0.58 197 5.05 0.37 128 4.27 0.3103


6

Figure 1. Geometrical representation of the M2129 diffusing S-duct [from Ref. 16].

Figure 2. Centerline and outlet cross-section grid of the baseline M2129 S-duct (top) and of the S-duct with injection and suction ports (bottom). Enlarged are the unstructured grid of the injection port and structured grid of the cross section at the outlet.

Flow

Flow


7

Flow

Figure 3. Schematics of the experimental apparatus showing the S-duct, the test section, and the rotating pitot rake.

Figure 4. Picture of the experimental apparatus.

Figure 5. Detail of the rotating pitot rake inside the test section as seen from upstream (left) and map of the test section points surveyed by the pitot probes.

-30 -20 -10 0 10 20 30-30

-20

30

-10

0

10

20

mm

mm


8

Figure 6. Distribution of total pressure ratio in baseline S-duct with PR=1.02: a) centerline plane from computation; outlet cross section from: b) computation; c) experiments.

p0/p∞b)

Separation

a)

c)

1.015

1.010

1.020

1.010

Figure 7. Distribution of total pressure ratio in S-duct with flow control at PR=1.02: a) centerline plane from computation; outlet cross section from: b) computation; c) experiments.

p0/p∞

b)

Reattachment

Separation bubble

c)

1.015

1.012

a)

1.020

1.010


9


a)

p0/p∞b) c)

1.030

1.015

1.040

1.010


p0/p∞

b)

a)

c)

1.045

1.040

1.040

1.010


10


a)

p0/p∞b) c)

1.050

1.025

1.060

1.020


p0/p∞

b)

a)

c)

1.045

1.065

1.060

1.020


11


p0/p∞b)

a)

c)

1.070

1.030

1.080

1.030


p0/p∞

b)

a)

1.060

1.080

1.070

c)

1.060

1.030


12


p0/p∞

a)

b) c)

1.085

1.015

1.030

1.100


p0/p∞

b)

a)

1.070

1.100

1.080c)

1.100

1.030


13