47th AIAA Aerospace Sciences Meeting Including The New Horizons Forum and Aerospace Exposition 5 - 8 January 2009,

Orlando, Florida

AIAA 2009-1106

47th AIAA Aerospace Sciences Meeting, 5-8, Jan 2009, Orlando, Florida

Post-Stall Flow Control Using Flexible Fin on Airfoil

Tianshu Liu1, J. Montefort,2 W. Liou,3 S. Pantula,4 Y. Yang5 Western Michigan University, Kalamazoo, MI 29008 Q. A. Shams6 NASA Langley Research Center, Hampton, VA 23681 Abstract This paper explores a new concept of post-stall flow control for airfoils by using a thin flexible fin attached on the upper surface of an airfoil to passively manipulate flow structures in fully separated flows for drag reduction and oscillation/flutter suppression. The relevant similarity parameters are given for aerodynamic and aeroelastic scaling of an airfoil with a flexible fin. Experiments are conducted on a NACA0012 airfoil section in a water tunnel. Force measurements are made by using a balance and velocity fields are measured by using particle image velocimetry. Drag reduction and oscillation suppression particularly for the natural lowfrequency oscillation in deep stall are achieved by using a flexible fin attached at a suitable location on the airfoil. The analysis of the flow fields and the connection between the velocity fields and the fin kinematics provides insight into the physical mechanisms of the low-frequency oscillation and the post-stall flow control. 1. Introduction Flow separation control is of immense importance to the performance of air vehicles and other technologically important systems involving fluids. Generally, it is desired to postpone separation such that the form drag is reduced, stall is delayed, lift is enhanced, and pressure recovery is improved. Therefore, a great effort has been made over years for flow separation control (or stall control) of airfoils and wings by using various techniques including synthetic jets (Glezer & Amitay 2002, Mittal et al. 2005), plasma actuators (Post & Corker 2004, Patel et al. 2007), piezoelectric actuators (Seifert, et al. 1998, Mathew, et al. 2006), deflected flap (Greenblatt & Wygnanski 2000), vortex generators (Gad-el-Hak 2000), passive and active blowing (Gad-el-Hak 2000), local suction (Atik et al. 2005), and oscillating camber (Munday & Jacob 2002). In general, separation control has been achieved by either injecting highmomentum fluid into separation regions or removing low-momentum fluid from the partially separation flows for angle of attacks (AoA) that are smaller than or near the stall AoA. It has been demonstrated that a partially separated flow can be almost re-attached by unsteady forcing, and the stall is delayed. Post-stall flow control after the flow over an airfoil or wing is completely separated is more challenging due to complicated separated flow structures. Post-stall flows contain not only
Associate Professor, Corresponding Author, Department of Mechanical and Aeronautical Engineering, G-220, Parkview Campus, Western Michigan University, Kalamazoo, MI 49008, tianshu.liu@wmich.edu, 269-276-3426, Senior Member AIAA 2. Assistant Scientist, Department of Mechanical and Aeronautical Engineering, Western Michigan University, Kalamazoo, MI 49008 3. Professor, Department of Mechanical and Aeronautical Engineering, Western Michigan University, Kalamazoo, MI 49008, Associate Fellow, AIAA 4. Graduate Research Assistant, Department of Mechanical and Aeronautical Engineering, Western Michigan University, Kalamazoo, MI 49008 5. Graduate Research Assistant, Department of Mechanical and Aeronautical Engineering, Western Michigan University, Kalamazoo, MI 49008 6. Research Scientist, NASA Langley Research Center, Hampton, VA 23681 Copyright 2009 by the American Institute of Aeronautics and Astronautics, Inc. All rights reserved.
1

1 AIAA Paper 2009-0736

Copyright 2009 by the American Institute of Aeronautics and Astronautics, Inc. All rights reserved.

47th AIAA Aerospace Sciences Meeting, 5-8, Jan 2009, Orlando, Florida

vortex shedding due to the Kelvin-Helmholtz instability and random turbulence in shear layers, but also a natural, global low-frequency oscillation that could cause severe flutter in the marked regime in a diagram of AoA and Reynolds number (see Fig. 1). Zaman et al. (1989) have studied in detail the low-frequency oscillation of a LRN(1)-1007 airfoil near stall and suggested that it might be due to switching between the stalled and un-stalled flows. Bragg et al. (1993, 1996) and Broeren & Bragg (1999, 2001) have conducted subsequent studies on unsteady stalling of a LRN(1)-1007 airfoil and provided experimental evidences on the combination of thin-airfoil stall and the trailing edge stall as the mechanism of the low-frequency oscillation. Large-eddy simulation for a LRN(1)-1007 airfoil has captured the some characteristics of the low-frequency oscillation observed in the corresponding experiments (Mukai et al. 2006). Figure 1 is a diagram of AoA and Reynolds number, indicating the possible regimes of flow patterns over thin airfoils, which is similar to that given by Wu et al. (1998). Data of the stall AoAs in Rec = 3 10 4 107 are collected from Loftin & Smith (1945) and Selig et al. (1995). Data of the stall AoAs in Rec = 2 10 3 2 10 4 are from Laitone (1997), Kunz & Kroo (2001), Sunada et al. (2000) and Okamoto et al. (1996). The region marked by dashed line indicates a possible regime in which the natural low-frequency oscillation may occur. Data from the experiments of Zaman et al. (1989) and Bragg et al. (1993, 1996) and Broeren & Bragg (1999, 2001) are marked in the diagram. The cases of the low-frequency oscillation investigated in this work are also marked in Fig. 1. It can be seen that the low-frequency oscillation occurs in a range of Reynolds numbers for micro-air vehicles and low-pressure turbine blades. The effects of acoustic excitation on stalled flows over an airfoil have been studied by Zaman (1992). Stall flow control using periodic blowing and suction near the leading edge of an airfoil has been numerically simulated by Wu et al. (1998). The excitation is typically based on the vortexshedding modes. At this stage, however, there is still a lack of a systematical investigation of control of the low-frequency oscillation and vortex-shedding all together particularly at high AoAs. This paper explores the post-stall flow control for airfoils by using a thin flexible fin attached on the upper surface of an airfoil. Figure 2 illustrates a thin flexible fin attached to an airfoil at post-stall angles of attack. The flexible fin used in this study is a thin Mylar film for passive control, which could be replaced by a polymer or composite sheet embedded with sensors and actuators for active control. To a certain degree, the idea of using a flexible fin is inspired by the fact that the major propulsive and control elements of birds, insect and aquatic animals like wings, tails and fins are flexible. Nevertheless, the proposed flexible fin for drag reduction is not a replica of fish fins that are mainly used for locomotion and maneuver. Morphologically, a flexible fin attached to an airfoil is basically rectangular unlike fish fins. In this paper, the relevant similarity parameters are first discussed for aerodynamic and aeroelastic scaling of an airfoil with a flexible fin. Experiments conducted on a NACA0012 airfoil section in a water tunnel are described. The results obtained from force measurements by a balance and velocity field measurements by PIV are presented. It is shown that drag reduction and oscillation suppression in stall can be achieved by using a flexible fin attached at a suitable location on the NACA0012 airfoil. The spectral analysis of the flow fields and the correlation analysis between the velocity fields and the fin kinematics are given. These results reveal the physical mechanisms of the low-frequency oscillation and the post-stall flow control. This concept could be useful in applications like gust alleviation for micro-air-vehicle and separation suppression of low-pressure turbine blades.
2 AIAA Paper 2009-0736

47th AIAA Aerospace Sciences Meeting, 5-8, Jan 2009, Orlando, Florida

2. Similarity Parameters The similarity parameters for a rigid airfoil with a thin flexible fin are considered. The similarity parameter for a rigid airfoil in an incompressible flow is Reynolds number Rec = U c / , where U is the freestream velocity and c is the airfoil chord. The nondimensional time parameter is U / c , where is a characteristic time scale of a flexible fin. When is replaced by 1 / , this parameter is inversely proportional to the reduced frequency c / U , where is a characteristic frequency of the flexible fin. For a flexible fin attached to the rigid airfoil, additional similarity parameters should be obtained. The flexible fin is considered as a thin cantilever plate deforming under fluiddynamical loading. The differential equation for the displacement w of a thin plate (e.g. fin) is generally expressed by (Meirovitch 1967) 2 w( P , t ) (1) L [ w( P , t )] + C [ w( P , t )] + M ( P ) = F ( P ,t ) , t t 2 where P denotes the coordinates x and y, L is a linear differential operator, C is a linear homogenous differential operator for damping, M is also a linear operator, and F ( P ,t ) is an external distributed force on the plate. In our case, F ( P ,t ) = p is the fluid-dynamical pressure difference across the plate. For a homogenous plate, the operator L = DE 4 is the bi-harmonic operator, where DE = Eh 2 / 12( 1 2 ) is the plate flexural rigidity, and M ( P ) is the mass distribution of the plate. The rigidity DE depends on the plate thickness (h), Youngs modulus (E), and Possion ratio (). The operator C is a linear combination of the operator L and the mass function M, C = a1 L + a 2 M , where a1 and a 2 are constant coefficients. An analysis for the rectangular fin deformation is given in Appendix A. Let us introduce the following non-dimensional variables F' = F / q , w' = w / l , x' = x / l , y' = y / l , t' = t / , (2) 2 where q = 0.5 U is the dynamical pressure and l is the fin length. Thus, substituting Eq. (2) into Eq. (1), we have the non-dimensional form of Eq. (1) 2 w' a = F' , (3) G1' 4 w' + 1 G1 ' 4 w' + ( a 2 )G 2 w' +G2 t' t' 2 t' where the two similarity parameters related to the rigidity and mass distribution are D Ml G1 = 3 E and G 2 = 2 . (4) l q q Other similarity parameters related to the damping are a1 / and a 2 . Therefore, the functional relation for the force coefficient of an airfoil with a flexible fin can be generally expressed by C F = f ( , Re , G1 , G2 , a1 / , a 2 ) . (5) In this study, the effect of AoA is mainly considered at a fixed Reynolds number. Since a Mylar fin with the fixed length and thickness is used, the non-dimensional parameters related to the fin are fixed. Clearly, a considerable amount of work is required in the future in order to understand the behavior of the flow control in the whole parametric space.

3 AIAA Paper 2009-0736

47th AIAA Aerospace Sciences Meeting, 5-8, Jan 2009, Orlando, Florida

3. Experimental Setup Experiments were conducted in the water tunnel (The Rolling Hills Research Corporation Model 1520) in the Fluid Mechanics Laboratory at Western Michigan University. The test section is nominally 15 in wide, 20 in high and 60 in long. The tempered glass, 3/8 in thick on the sidewalls and 1/2 in thick on the bottom, is mounted with silicon rubber, allowing good optical access from the top, bottom, both sides and rear for flow visualization and particle image velocimetry (PIV) measurements. The tunnel is operated as a continuous flow channel and the water level in the test section is typically adjusted to be roughly 50 mm below the top of the walls. The free water surface provides simple access to the model and easy setup of an external force balance. There is a 6:1 contraction section before the test section for turbulence reduction and avoidance of local separation and vorticity development. The test section flow velocity is variable from 0 up to 0.3 m/s. In the test section, the turbulence intensity is less than 0.1%, and the velocity non-uniformity is less than 2%, and the mean flow angularity is less than 1% in both the pitch and yaw angles. A plastic NACA0012 airfoil section model built by a rapid prototype machine was tested. The chord and span of the model were 10 in and 12 in, respectively. A clear Mylar (PETpolyester) film was used as a flexible fin. The Youngs modulus and Poisson ratio for Mylar were 2.8 GPa and 0.37, respectively. The Mylar film was attached to the upper surface of the airfoil by Scotch tape. Figure 3 shows a flexible Mylar fin attached on the NACA0012 model in water in a typical PIV image. To reduce the three-dimensionality of flow, the model was mounted on two 6.35 mm thick Plexiglas end plates (406 mm by 279 mm) that were supported by an aluminum bar. The bar was directly connected to a home-made external force balance that was above the water surface. For measurements of very small drag force, the main balance element was a 275 mm long aluminum beam with a 30 mm thick and 12.75 mm wide crosssection. Two sections of the beam ware machined and considerably thinned, and the center locations of the two sections were at 6.4 and 11.3 mm from the corresponding end. The thickness and length of the thinned sections is 2.75 mm and 35 mm, respectively. Two Omega strain gauges were installed at the middle of the thinned sections. When the balance element (beam) was mounted perpendicularly to the incoming flow direction, the difference between voltage outputs from the two strain gauges simply equaled to a product of the drag and the distance between the two gauges. Balance calibrations verified this simple and direct relationship. The measurement uncertainty of drag was less than one gram. A similar beam was used for lift measurement when this beam was parallel to the incoming flow direction. The flow around the model was illuminated by a 2-mm thick laser sheet generated by a Big Sky laser (CFR190) from the rear window of the test section. The laser frequency was set at 15 Hz, which was sufficient for time-resolved measurements in this study since the dominant frequencies in flows were lower than 0.5 Hz. The interval between two pulses was 2 ms. The measuring area was imaged by a PIV camera (TSI PIVCAM 10-30, Model 630046) with an 85mm lens. PIV images were processed using the TSI Insight5 PIV software (3232 pixels window size). 4. Flow Control Using Flexible Fin 4.1. Drag Reduction and Oscillation Suppression A 0.25c long rectangular Mylar fin with a thickness of 0.1 mm was first used in preliminary tests, where c is the airfoil chord of 10 in (254 mm). To examine how a flexible fin
4 AIAA Paper 2009-0736

47th AIAA Aerospace Sciences Meeting, 5-8, Jan 2009, Orlando, Florida

affects the drag of the airfoil, the fin was placed at different locations on the upper surface of the NACA0012 model. In addition, two tandem fins at different locations were tested. The incoming flow velocity was 0.25 m/s, and the Reynolds number based on the chord was Rec = 6.3 10 4 . Figure 4(a) shows the time-averaged drag coefficient C D as function of AoA for different arrangements of flexible Mylar fins. When AoA is larger than 12o, C D of the NACA0012 model with a flexible fin is smaller than that of the baseline model. For nonseparated flow at smaller AoAs, it has been observed that a flexible fin was not naturally attached on the surface in water due to its buoyancy, and thus it became an intrusive object to the attached boundary layer. In this case, as also shown in Fig. 4(a), C D of the NACA0012 model with a flexible fin is greater. This problem could be practically solved by somehow making the flexible fin attached on the surface for small AoAs before stall. This study focuses on the effects of a flexible fin on separated flow after the airfoil stalls. Instead of completely removing separation, a flexible fin passively alters flow structures to suppress separation and reduce the drag. As a typical case for detailed investigation, a significant drag reduction at high AoAs by a 0.25c fin located at 0.1c is shown in Fig. 4(b). The absolute measurement uncertainty of drag is 1 gram, and the relative uncertainty of C D is 1-3% in a range of AoAs from 12o to 20o. The CFD results of the drag coefficient for the same case are also given in Fig. 4(b), which are obtained by using a combined Navier-Stokes immersed boundary solver coupled with a finiteelement structural code for laminar flows (Pantula 2008). The CFD prediction is roughly consistent with the measured results, particularly indicating a drag reduction when AoA is larger than 12o, although the CFD drag values are considerably larger. As shown in Fig. 5, CFD calculations indicate that the pressure drag is considerably reduced by the 0.25c fin at high AoAs. The skin friction drag is increased for small AoAs since the fin is detached from the surface in calculations to simulate the situations in the water tunnel tests. Figure 6 shows the lift coefficient as a function of AoA for the baseline model and the model with the 0.25c fin, where the correlation given by the McCormicks formula for the effective wing aspect ratio of 4.42 is shown as a reference (Liu et al. 2007). The lift of the model with the fin is considerably decreased. For low AoAs, the smaller lift is caused by the detached fin from the surface. After stall at high AoAs, the decreased lift is probably caused by suppressing the lift-generating organized vortices by the fin. It will be pointed out that the suppression of the organized vortices is also responsible to the drag reduction after stall. Therefore, the lift decrease after stall is considered as a side effect of this flow control method. Actually, the drag of the NACA0012 model varies with time due to the highly unsteady nature of separation. The root-mean-squared (r.m.s) variation of the drag coefficient C D of the baseline model is about 3% of the mean value. The power spectra of C D are shown in Fig. 7 for the baseline model and model with the flexible fin at AoAs of 14o and 18o. The spectral peaks at 0.07 and 0.12 Hz are observed for the baseline model at AoA of 14o, and these low-frequency components have the Strouhal numbers based on the projected chord St LF = f c sin( ) / U e of 0.017 and 0.03, respectively. For AoA of 18o, the spectral peaks at 0.04, 0.07 and 0.08 Hz correspond to St LF = 0.012, 0.022 and 0.025, respectively. These low-frequency fluctuations of the drag are basically consistent with the natural low-frequency oscillation studied by Zaman et al. (1989). The similar low-frequency components are also found at AoAs of 16o and 20o. A significant backward-forward swing of the baseline model has been observed in a range of AoAs from 14o to 20o. Interestingly, as indicated in Fig. 7, the flexible fin attached at 0.1c
5 AIAA Paper 2009-0736

47th AIAA Aerospace Sciences Meeting, 5-8, Jan 2009, Orlando, Florida

considerably dampens the low-frequency drag fluctuations and the model swing. The physical mechanisms of the drag reduction and oscillation suppression will be explored based on timerevolved PIV measurements of flow fields. 4.2. Development of Flow Structures 4.2.1. Mean Flow Properties To understand the physical mechanisms of the drag reduction by a flexible fin, PIV measurements of flow fields were conducted and velocity fields were obtained at 15 Hz. Figure 8 shows the time-averaged streamlines and vorticity fields for the baseline model and the model with a 0.25c fin located at 0.1c at AoA of 18o. The change of the velocity fields by the flexible fin is appreciable. For the baseline model, a time-averaged large vortex or a re-circulating flow region is observed on the upper surface near the trailing edge of the airfoil, which can be loosely considered as a separation bubble in a long-time average sense. Indeed, a time sequence of velocity and vorticity fields indicates that strong large-scale organized vortices are developed from the Kelvin-Helmholtz instability in the shear layer. Although the evolution process of these vortices is highly unsteady, they occur at a higher probability in that region. Due to the presence of the flexible fin, the time-averaged large vortex is largely destroyed. Figure 9 shows the profiles of the x-component of the mean velocity U at x/c = 0.4, 0.51, 0.66, 0.79 and 0.91 on the upper surface for AoAs of 18o and 20o, where x is the coordinate along the incoming flow direction and c is the projected chord onto the x-coordinate. Clearly, the momentum loss is reduced due to the presence of a flexible fin in the separated flow region, which corresponds to the drag reduction found in force measurements by the external balance. Further evidence is provided by the momentum thickness development along the x-direction, as shown in Fig. 10. Here, the momentum thickness is defined as U ( y ) U ( y ) = (6) 1 dy , Ue Ue ys

where U = U min( U ) , U is the x-component of the mean velocity, U e is the external velocity, and y s is the y-coordinate at the surface. As indicated in Fig. 10, the development of the momentum thickness is significantly suppressed by a flexible fin. 4.2.2. Unsteady Flow Properties Figures 11 and 12 show the power spectra of the x-component of velocity U across the separation region along the y-direction on the baseline model and the model with a 0.25c fin located at 0.1c at five x locations for AoA of 18o. As illustrated in Fig. 2, the x-coordinate is the coordinate starting from the leading edge along the incoming flow and the y-coordinate is the normal coordinate through the trailing edge. For the baseline model, the dominant spectral component at f = 0.07 Hz is developed downstream. The Strouhal number based on the initial shear layer momentum thickness is St shear = f 0 / U e = 8.4 10 4 that is much smaller than the most unstable mode ( St shear = 0.032 ) in the shear layer, where 0 = 2 mm is the estimated initial momentum thickness. Therefore, such a low-frequency component cannot be the shear layer mode from the linear shear-layer instability, and it rather represents a natural, global lowfrequency oscillation of the separation region. In fact, the Strouhal number based on the frontprojected chord c sin( ) is St LF = f c sin( ) / U e = 0.022 , which is consistent with the natural
6 AIAA Paper 2009-0736

47th AIAA Aerospace Sciences Meeting, 5-8, Jan 2009, Orlando, Florida

low-frequency oscillation observed by Zaman et al. (1989). This low-frequency component of the velocity U is responsible to the drag fluctuation shown in Fig. 7. Figure 12 shows the effect of a flexible fin on the power spectra of U at the same locations, indicating that the dominant component and other components of U are significantly suppressed. This is also consistent with the drag measurements shown in Fig. 7. Furthermore, Figure 13 shows the development of the spectral components averaged across the shear layer ( y / c = 0.29 0.37 ) along the x-direction (incoming flow direction) for the baseline model and the model with a 0.25c fin located at 0.1c. For the baseline model, the dominant low-frequency component at f = 0.07 Hz ( St LF = 0.022 ) increases in a nearly linear fashion in x / c = 0.2 0.8 , and then it saturates and decays. As indicated in Fig. 13(b), the development of all the spectral components particularly the dominant component at f = 0.07 Hz ( St LF = 0.022 ) is suppressed by the flexible fin. Accordingly, as shown in Fig. 14, the Reynolds stress uv in the whole separated flow region is considerably decreased due to the presence of the flexible fin, where u = U U and v = V V are the velocity fluctuations in the x- and ycoordinates, respectively. 5. Vortex Shedding and Low-Frequency Oscillation The above results indicate that a flexible fin dramatically dampens the dominant spectral components particularly the low-frequency oscillation in the separated flow region over a NACA0012 airfoil. In order to understand the underlying mechanisms, it is necessary to examine the development of vortices and the generation of the low-frequency oscillation on the baseline NACA0012 model. Figure 15 shows the streamlines and vorticity fields, indicating the growth of a counterclockwise vortex shedding from the leading edge of the baseline model in an interval of 1.0 s. The vorticity sheet in the shear layer rolls up into discrete vortices due to the Kelvin-Helmholtz instability and shed from the leading edge. The vortex travels downstream as its size increases, while another vortex remains near the trailing edge in a probabilistic sense. Then, the two vortices tend to merge together to form a larger vortex. This dual-vortex structure is typically observed in our experiments, which is similar to that in the dynamic stall (Carr 1988). The vortex shedding is at about 1 Hz that corresponds to the bluff-body vortex shedding Strouhal number St shed = f c sin( ) / U e = 0.31 . PIV measurements were conducted to focus on the region near the leading edge. Figure 16 shows the power spectra of the velocity components U and V at 0.45c for the baseline model. Interestingly, the dominant spectral peak at 1 Hz that is the bluff-body vortex shedding mode St shed = 0.31 appears in the spectrum of the normal velocity V. In contrast, the dominant component in the spectrum of U is at 0.07 Hz that corresponds to the natural low-frequency oscillation at St LF = 0.022 . The low-frequency component in the spectra of U is detected in the whole separated flow region. In the spectra of V, the vortex shedding component is strong in 0.3-0.5c and the low-frequency component become more evident near the trailing edge. This indicates that the low-frequency oscillation is a global phenomenon, and in contrast the vortex shedding is more active in the shear layer near the leading edge. These two salient modes coexist in the separated flow region over the NACA0012 model. Figure 17 shows the Strouhal numbers based on the local momentum thickness for the low-frequency mode (f = 0.07 Hz) and the vortex shedding mode (f = 1.0 Hz) for the baseline NACA0012 model. Although the vortex shedding Strouhal number increases with x/c due to the increasing momentum thickness, the
7 AIAA Paper 2009-0736

47th AIAA Aerospace Sciences Meeting, 5-8, Jan 2009, Orlando, Florida

median value of St is about 0.032 that is the most unstable mode in the shear layer. This provides a connection between the vortex shedding and the shear layer instability. On the other hand, the Strouhal number St for the low-frequency mode is so small compared to the most unstable mode in the shear layer that any direct relationship between the low-frequency mode and the Kelvin-Helmholtz instability is excluded. Zaman et al. (1989) have suggested that the low-frequency oscillation may be attributed from switching between stalled and un-stalled flows. Phase-averaged flow field measurements by Broeren & Bragg (1999) have found that the generation and growth of a separation bubble from the leading edge and its merging to the trailing-edge separation could cause the lowfrequency oscillation. However, due to the highly unsteady flows with a considerable degree of randomness in our PIV measurements, the scenario of a steadily growing bubble near the leading edge described by Broeren & Bragg (1999) is not clearly observed. To gain insights into the origin of the low-frequency oscillation, the center of the large-scale vortex in the time-averaged velocity field in Fig. 8(a) is selected as a reference point (the mean zero-crossing point of U) to examine the temporal signal of U. In addition, when a U-velocity profile across the reference point is plotted, there is an instantaneous zero-crossing point across which the velocity U changes its sign. The zero-crossing point defines the instantaneous boundary of the reversed flow. Figure 18 shows the time traces of the velocity U and the zero-crossing point position at the reference point. Although there are a considerable amount of higher-frequency components, the low-frequency oscillation of 0.1 Hz is visible, and the power spectra clearly show the dominant peak at about 0.1 Hz for both the velocity U and the zero-crossing point position. It is also observed in Fig. 18 that the phase angle between the velocity U and the zero-crossing point position is about 180o, which is confirmed by a correlation analysis. According to the continuity equation, the valley of the zero-crossing point position should roughly correspond to the peak of U. Further, the magnitude squared coherence between the velocity U and the zero-crossing point position y cp is calculated, i.e.,
2

C ycpU =

PycpU ( f ) Pycp ycp ( f )PUU ( f )

(7)

where PycpU ( f ) is the cross power spectral density, and Pycp ycp ( f ) and PUU ( f ) are the power spectral densities. C ycpU is a function of frequency with values between 0 and 1 that indicates how well y cp corresponds to U at each frequency. Figure 19 shows the magnitude squared coherence C ycpU , indicating a high correlation between the velocity U and the zero-crossing point position y cp at 0.1 Hz and 0.4 Hz. The above results indicate that the low-frequency oscillation of U is related to the fluctuation of the zero-crossing point position or the boundary of the reversed flow. In a certain sense, this phenomenon corresponds to a cycle of the growth and bursting of a separation bubble described by Broeren & Bragg (1999). It is required to extract the flow fields that are responsible to the low-frequency oscillation from the instantaneous velocity fields also containing the active vortex shedding at a higher frequency. In order to smooth out the velocity components associated with the vortex shedding in the instantaneous flow fields, the velocity fields are averaged over a relatively short time interval of 0.73 s that is the vortex shedding period at five selected phases in a low-frequency cycle of U. These phases are marked by open circles on the time trace of U at the reference point
8 AIAA Paper 2009-0736

47th AIAA Aerospace Sciences Meeting, 5-8, Jan 2009, Orlando, Florida

in Fig. 18. The short-time-averaged velocity fields at the five phases are shown in Fig. 20, which reveal the flow structures associated with the low-frequency oscillation. As shown in Fig. 20(a), at the valley (a) in U at the reference point, a strong re-circulating flow region is observed in a sense of the short-time average, which is caused by the occurrence of a strong vortex or a group of interacting vortices developed in the shear layer. For simplicity, this re-circulating flow region is informally called a vortex due to the closed streamlines although the definition of a vortex is an arguable topic (Jeong & Hussain 1995). Figure 20(a) indicates that a smaller vortex forms near the leading edge. As shown in and 20(b), the newly-formed vortex grows as the original large vortex travels downstream. The dual-vortex structure is typical in the separated flow region and a saddle point between the two vortices is usually observed. As shown in Fig. 20(c), at the peak (c) of U at the reference point the velocity of the reversed flow becomes much smaller since the strong organized vortices are gone, and the topological structure of flow is considerably different. Figures 20(d) and 20(e) indicate that the dual-vortex structure occurs and develops again. The low-frequency oscillation results from the periodic switching between the strongly and weakly re-circulating flows in the separated flow region, which may be triggered by the global instability of the dual-vortex structure. A simplified analysis is given in Appendix B to shed insight into the instability of the dualvortex structure. Figure 21(a) shows the flow field of two point vortices on a flat plate at AoA of 18o, which is generated by using the Joukowski transformation as a crude model of the dualvortex structure observed in the experiments. The normalized strengths of the vortex #1 (near the leading edge) and vortex # 2 (near the trailing edge) are 1 / cU = 0.7 and 2 / cU = 1 , respectively. A dynamical system for the two point vortices on a flat plate is given in Appendix B. The equilibrium positions of the two vortices are calculated depending on the nondimensional strengths of the two vortices and AoA. Calculations indicate that the equilibrium positions of the vortices are unstable for a finite-amplitude disturbance. This implies that the dual-vortex structure on an airfoil cannot remain stationary. Figure 21(b) shows typical trajectories of the two vortices moving away from their initial positions above a plate. This problem is similar to the stability problem of a single vortex on an airfoil in trapping of a free vortex for lift enhancement (Saffman and Sheffield 1977, Huang and Chow 1982, Chow et al. 1985). A free vortex on an airfoil is generally unstable for a finite-amplitude disturbance without active flow control. After the two vortices move away from the airfoil, the dual-vortex structure is re-generated from interacting organized vortices developed in the shear layer. Although the dual-vortex structure is initially induced by the organized vortices developed in the shear layer due to the Kelvin-Helmholtz instability, the motion of the dual-vortex structure is governed by its own dynamical system once it forms. 6. Interaction between Flexible Fin and Unsteady Flow As pointed out before, a flexible fin attached to the upper surface of the NACA0012 airfoil passively interacts with flow, suppresses separation and reduces the drag at high AoAs. To understand the interaction between the flexible fin and its surrounding flow, it is necessary to measure the fin kinematics and fluid dynamics from PIV measurements near the fin. A typical PIV image near the fin is shown in Fig. 3. The fin is clearly visualized in PIV images as an intersection between the fin and laser sheet. The high-contrast edges of the fin in images are detected by using the Cannys edge detector. The coordinates of the fin edges are fitted using the characteristic beam functions and the fin magnitude is determined as a function of time. In Appendix A, the dynamics of a rectangular fin is discussed. For the nearly 2D deformation of
9 AIAA Paper 2009-0736

47th AIAA Aerospace Sciences Meeting, 5-8, Jan 2009, Orlando, Florida

the fin observed in our experiments, the first eigenfunction is sufficient to describe the fin deformation. Therefore, the normalized displacement of a thin fin is expressed as w( x3 ) / l = 1 ( t ) X 1 ( x3 ) , (7) where x 3 is the coordinate along the incoming flow from the root of the fin and 1 ( t ) are the time-dependent amplitude. The first characteristic beam function is X 1 ( x3 ) = cos 1 x cosh 1 x + k 1 (sin 1 x sinh 1 x ) , (8) where k1 = (sin 1 sinh 1 ) /(cos 1 + cosh 1 ) , x = x3 / l is the coordinate normalized by the fin length and 1 = 1.875 . By using Eq. (7) for least-squares fit to the detected fin edges, the time-dependent amplitude 1 ( t ) is determined. Figure 22 shows the time-dependent amplitude 1 ( t ) and power spectrum of the 0.25c flexible fin located at 0.1c for AoA of 18o. The dominant spectral peak is at 1.3 Hz and the secondary peak is at 0.12 Hz. For AoAs of 12o, 14o, 16o and 20o, the most dominant spectral peaks are 0.1 Hz, 0.1 Hz, 1.3 Hz and 0.25 Hz, respectively. In general, the fin amplitude has a dominant component around 1.3 Hz and some low-frequency components in a range of 0.07-0.25 Hz. This indicates that the flexible fin is more responsive to the vortex shedding associated with the shear layer instability. The velocity fields near the flexible fin are obtained in PIV measurements. A local coordinate system ( x" , y" ) located at the leading edge is used for its convenience, as shown in Fig. 3. The transformations between the coordinate systems ( x , y ) and ( x" , y" ) are x" = x and y" = y c sin . Figures 23 and 24 show the power spectra of the x-component U and ycomponent V of velocity across the fin along the y-direction on the NACA0012 model with a 0.25c fin located at 0.1c at several locations of x" / c for AoA of 18o. The power spectra of U have significant peaks at 0.1 Hz, 0.5 Hz and 1.3 Hz on the upper side of the fin in x" / c = 0.27 0.39 that is approximately within the fin. More spectral peaks occur downstream at x" / c = 0.45 . In contrast, the spectra of V have the significant peaks are around 1.3 Hz that corresponds to the vortex shedding mode. To provide a direct connection between the fin kinematics and the surrounding flow, the velocity is averaged over a zone around the fin ( 0.25 x" / c 0.4 and 0.05 y" / c 0.05 ). Figure 25 shows the time traces of the relative fin amplitude and the zone-averaged velocity components U and V, where the fin amplitude and velocities are normalized by the fin length and the freestream velocity, respectively. The waveform of the fin amplitude approximately corresponds to that of the zone-averaged velocities except that there is a phase shift between them. Further, the magnitude squared coherences between the fin amplitude and the zone-averaged velocity components C U and C V are calculated, where 1 ( t ) is the fin amplitude, and U and V are the zone-averaged velocity components. The definition of C ( C
U 1
1 1

is C

U 1

= P

2 U( 1

f)

1 1

( f )PUU ( f ) , where P U ( f )
1

is the cross power spectral density, and P

1

1 1

( f ) and PUU ( f ) are the power spectral densities

is similarly defined). Figure 26 shows the magnitude squared coherence between the fin

amplitude and the zone-averaged velocity components. For both U and V , the dominant peaks in the magnitude squared coherence are in a range of 1 Hz to 1.7 Hz, indicating a high correlation between the fin amplitude and velocities at the vortex shedding mode. The phase

10 AIAA Paper 2009-0736

47th AIAA Aerospace Sciences Meeting, 5-8, Jan 2009, Orlando, Florida

differences between the fin magnitude 1 ( t ) and the zone-averaged velocity components U and V around the fin are 14o and 149o, respectively. 7. Conclusions Force measurements indicate that the drag is reduced and the natural low-frequency oscillation is suppressed in deep stall for AoAs from 12o to 20o at Rec = 6.3 10 4 by using a thin flexible polymer fin attached at a suitable location on the NACA0012 airfoil. Due to the presence of a flexible fin, the velocity deficit and the momentum loss in the separated flow region are reduced, and the time-averaged large vortex causing stronger reversed flow is largely destroyed. As a side effect, however, the lift is reduced in stall since the organized liftinggenerating vortices are suppressed by the fin. Based on the spectral analysis of velocity fields, it is found that the dominant low-frequency oscillation at St LF = 0.022 and other spectral components including the vortex shedding mode are suppressed by the flexible fin. For further understanding, the mechanism of generating the natural low-frequency oscillation on the baseline NACA0012 model is explored. It is found that the vortex shedding associated with the KelvinHelmholtz instability coexists with the low-frequency oscillation. The vortex shedding mode is mainly detected in the shear layer near the leading edge, and in contrast the low-frequency oscillation is a global phenomenon in the entire separated flow region. The short-time averaged velocity fields show that the low-frequency oscillation is related to the development of the dualvortex structure that is induced by the occurrence of a strong vortex or a group of interacting organized vortices developed in the shear layer. A simplified analysis indicates that the dualvortex structure on an airfoil is intrinsically unstable for a finite-amplitude disturbance, and therefore the low-frequency oscillation may result from the global instability of the dual-vortex structure. Interaction between the kinematics of the flexible fin and the flow around it is studied. The flexible fin is mainly responsive to the vortex shedding associated with the shear layer instability although it is also affected by the low-frequency oscillation. The flexible fin dampens the development of the shear layer instability and the organized vortices, and as a result the lowfrequency oscillation is largely eliminated and the drag is reduced at high AoAs. Acknowledgements This work was supported by the Air Force Office of Scientific Research, USAF, under the grant number FA9550-06-1-0187. References: Atik, H., Kim, C. Y., Van Dommelen, L. L., & Walker, J. D. A., Boundary-Layer Separation Control on a Thin Airfoil Using Local Suction, J. Fluid Mech., Vol. 535, 2005, pp. 415-443 Bragg, M. B., Heinrich, D. C., & Khodadoust, A., Low-Frequency Flow Oscillation over Airfoils near Stall, AIAA J., Vol. 31, No. 7, 1993, pp. 1341-1343 Bragg, M. B., Heinrich, D. C., Balow, F. A. & Zaman, K. B. M. Q., Flow Oscillation over an Airfoil near Stall, AIAA J., Vol. 34, No. 1, 1996, pp. 199-201 Broeren, A. P. & Bragg, M. B., Flowfield Measurements over an Airfoil During Natural LowFrequency Oscillations near Stall, AIAA J., Vol. 37, No. 1, 1999, pp. 130-132 Broeren, A. P. & Bragg, M. B., Unsteady Stalling of Thin Airfoils, in Fixed and Flapping Wing Aerodynamics for Micro Air Vehicle Applications, Ed: T. J. Mueller, AIAA Press, Reston, Virginia, 2001, Chapter 10
11 AIAA Paper 2009-0736

47th AIAA Aerospace Sciences Meeting, 5-8, Jan 2009, Orlando, Florida

Carr, L. W., Progress in Analysis and Prediction of Dynamic Stall, J. of Aircraft, Vol. 25, No. 1, 1998, pp. 6-17 Chow, C. Y., Huang, M. K. & Yan, C. Z., Unsteady Flow about a Joukowski Airfoil in the Presence of Moving Vortices, AIAA Journal, Vol. 23, No. 5, 1985, pp. 657-658 Fish, F. E. & Lauder, G. V., Passive and Active Flow Control by Swimming Fishes and Mammals, Annual Review of Fluid Mechanics, Vol. 38, 2006, pp. 195-224 Gad-el-Hak, M., Flow Control: Passive, Active and Reactive Flow Management, Cambridge University Press, 2000 Glezer, A. and Amitay, M., Synthetic Jets, Annu. Rev. Fluid Mech., 2002 Greenblatt, D. & Wygnanski, I. J., The control of flow separation by periodic excitation, Progress in Aerospace Sciences, 36, 2000, pp. 487-545 Huang, M. K. & Chow, C. Y., Trapping of a Free Vortex by Joukowski Airfoils, AIAA Journal, Vol. 20, No. 3, 1982, pp. 292-298 Jeong, J. & Hussain, F., On the Identification of a Vortex, J. Fluid Mech., Vol. 285, 1995, pp. 69-94 Kunz, P. J. & Kroo, I., Analysis and Design of Airfoils for Use at Ultra-Low Reynolds Numbers, in Fixed and Flapping Wing Aerodynamics for Micro Air Vehicle Applications, Ed: T. J. Mueller, AIAA Press, Reston, Virginia, 2001, Chapter 3 Laitone, E. V., Wind Tunnel Tests of Wings at Reynolds Numbers Below 70,000, Experiments in Fluids, Vol. 23, 1997, pp. 405-409 Liu, T., Montefort, J., Liou, W., Pantula, S. R. & Shams, Q. A. Lift enhancement by static extended trailing edge, Journal of Aircraft, Vol. 44, No. 6, 2007, pp. 1939-1947 Loftin, L. K. & Smith, H. A., Aerodynamic Characteristics of 15 NACA Airfoil Sections at Seven Reynolds Numbers from 0.7106 to 9.0106, NACA TN 1945, 1949 Mathew, J., Song, Q., Sankar, V., Sheplak, M., & Cattafesta, L. N., Optimized Design of Piezoelectric Flap Actuators for Active Flow Control, AIAA J., Vol. 44, No. 12, 2006, pp. 2919-2928 Meirovitch, L. Analytical Methods in Vibrations, The Macmillan Company, New York, 1967 Mittal, R., Kotapati, R. B., Cattafesta, L. N., Numerical Study of Resonant Interactions and Flow Control in a Canonical Separated Flow, AIAA paper 2005-1261, 2005 Mukai, J., Enomoto, S. & Aoyama, T., Large-Eddy Simulation of Natural Low-Frequency Flow Oscillations on an Airfoil near Stall, AIAA Paper 2006-1417, 2006 Munday, D. & Jacob, J., Active Control of Separation on a Wing with Oscillating Camber, AIAA J. of Aircraft, Vol. 39, No. 1, 2002 Okamoto, M., Yasuda, K. & Azuma, A., Aerodynamic Characteristics of the Wings and Body of a Dragonfly, The Journal of Experimental Biology, Vol. 199, 1996, pp. 281-294 Pantula, S. R., Modeling Fluid Structure Interaction over a Flexible Fin Attached to a NACA0012 Airfoil, Department of Mechanical and Aeronautical Engineering, Western Michigan University, Kalamazoo, MI, 2008 Patel, M., Sowle, Z. H., Corke, T. C., & He, C., Autonomous Sensing and Control of Wing Stall Using a Smart Plasma Slat, J. of Aircraft, Vol. 44, No. 2, 2007, pp. 516-527 Post, M. & Corke, T. C., Separation Control on High Angle of Attack Airfoil Using Plasma Actuators, AIAA J., Vol. 42, No. 11, 2004, pp. 2177-2184 Reddy, J. N., Theory and Analysis of Elastic Plates, Taylor & Francis, Philadelphia, 1999, Chapter 7

12 AIAA Paper 2009-0736

47th AIAA Aerospace Sciences Meeting, 5-8, Jan 2009, Orlando, Florida

Saffman, P. G. & Sheffield, J. S., Flow over a Wing with an Attached Free Vortex, Studies in Applied Mathematics, Vol. 57, 1977, pp. 107-117 Seifert, A., Eliahu, S., Greenblatt, D. & Wygnanski, I., Use of Piezoelectric Actuators for Airfoil Separation Control, AIAA J., Vol. 36, No. 8, 1998, pp. 1535-1537 Selig, M., Guglielmo, J., Broeren, A. P. & Giguere, P., Summary of Low-Speed Airfoil Data, Volumes 1, 2 and 3, SoarTech Publications, Virginia Beach, Virginia, 1995 Sunada, S., Ozaki, K., Tanaka, M., Yasuda, T., Yasuda, K., & Kawachi, K., Airfoil Characteristics at a Low Reynolds number, Journal of Visualization and Image Processing, Vol. 7, 2000, pp. 207-215 Tikhonov, A. N. & Arsenin, V. Y., Solutions of Ill-Posed Problems, John Wiley & Sons, New York, 1977, Chapter II Warburton, G. B., The Vibration of Rectangular Plates, Proceedings of the Institution of Mechanical Engineers, Vol. 168, No. 12 , 1954, pp. 371-383 Wu, J-Z, Lu, X-Y, Denny, A. G., Fan, M. & Wu, J-M, Post-Stall Flow Control on an Airfoil by Local Unsteady Forcing, J. Fluid Mech., Vol. 371, 1998, pp. 21-58 Zaman, K. B. M. Q., McKinzie, D. J. & Rumsey, C. L., A Natural Low-Frequency Oscillation of the Flow over an Airfoil near Stalling Conditions, J. Fluid Mech., Vol. 202, 1989, pp. 403442 Zaman, K. B. M. Q., Effects of Acoustic Excitation on Stalled Flows over an Airfoil, AIAA J., Vol. 30, No. 6, 1992, pp. 1492-1499 Appendix A. Flexible Fin Dynamics The solution to Eq. (1) for a rectangular thin plate can be written as an expansion based on the eigenfunctions wr ( P ) (Meirovitch 1967) w( P ,t ) =

w ( P ) ( t ) ,
r r r =1

(A1)

where r ( t ) are the time-dependent amplitudes. Assuming that the eigenfunctions are orthogonal and substituting Eq. (A1) into Eq. (1), one has a system of the ordinary differential equations for r ( t ) &&r ( t ) + c rs & s ( t ) + r2 r ( t ) = N r ( t ) ,
s =1

( r = 1, 2 ,L )

(A2)

where the damping coefficients are c rs = and the generalized force is Nr ( t ) =

w ( P )C [ w ( P )] dD( P ) ,
r s D

( r , s = 1, 2 ,L )

(A3)

w ( P )F( P,t )dD( P ) .

r D

( r = 1, 2 ,L )

(A4)

Note that appropriate dimensional constants are absorbed into Eqs. (A3) and (A4) to make Eq. (A2) dimensionally consistent. When the operator C is a linear combination of the operator L and the mass function M, C = a1 L + a 2 M , where a1 and a 2 are constant coefficients, the damping term in Eq. (A2) is not coupled such that c rs = 2 r r rs . Thus, Eq. (A2) becomes
13 AIAA Paper 2009-0736

47th AIAA Aerospace Sciences Meeting, 5-8, Jan 2009, Orlando, Florida

&&r ( t ) + 2 r r & r ( t ) + r2 r ( t ) = N r ( t ) , (A5) ( r = 1, 2 ,L ) where r are the damping factors and r are the natural frequencies. The solution to Eq. (A5) is

r ( t ) =

rd

t 0

N r ( ) exp[ r r ( t )] sin rd ( t )d
(A6)

r sin t + exp( r r t )cos rd t + rd r ( 0 ) , ( 1 r2 )1 / 2

1 &r ( 0 ) exp( r r t ) sin rd t + rd where rd = r ( 1 r2 )1 / 2 can be regarded as a natural frequency associated with the damping system. The second and third terms are the starting transients. For an impulse force F ( P ,t ) = ( t ) ( P Pim ) , when r ( 0 ) = 0 , the main behavior of r ( t ) is described by exp( r r t ) sin rd t .
The approximation for wr ( P ) is given by wr ( x , y ) = X n ( x )Ym ( y ) , (A7) where X n ( x ) and Ym ( y ) are the characteristic beam functions (Warburton 1965, Reddy 1999). For the edge x = 0 clamped and the other edge x = a free, the characteristic beam function X n ( x ) is X n ( x ) = cos n x cosh n x + k n (sin n x sinh n x ) , ( n = 1,2 ,3 ,L ) (A8) where x = x / a is the normalized coordinate and k n = (sin n sinh n ) /(cos n + cosh n ) . The values of n are given by solving the nonlinear equation cos n cosh n + 1 = 0 . The first three values of n are 1 = 1.875 , 2 = 4.694 and 3 = 7.8547 . For the edges y = 0 and y = b free, the characteristic beam function Ym ( y ) is Ym ( y ) = 1 , ( m = 0 ) Ym ( y ) = 1 2 y , ( m = 1 ) Ym ( y ) = cos m ( y 1 / 2 ) + m cosh m ( y 1 / 2 ) , ( m = 2 ,4 ,6 ,L )

Ym ( y ) = sin m ( y 1 / 2 ) + m sinh m ( y 1 / 2 ) , ( m = 3,5 ,7 ,L ), (A9) where y = y / b is the normalized coordinate, m = sin( m / 2 ) / sinh( m / 2 ) , and m = sin( m / 2 ) / sinh( m / 2 ) . The parameters m and m are given by solving the equations tan( m / 2 ) + tanh( m / 2 ) = 0 and tan( m / 2 ) tanh( m / 2 ) = 0 . The typical values of m

and m are 2 = 4.73 , 4 = 10.99 , 3 = 7.853 , and 5 = 14.137 . The combinations of X n ( x ) and Ym ( y ) are w1 ( x , y ) = X 1 ( x )Y0 ( y ) , w2 ( x , y ) = X 2 ( x )Y0 ( y ) , w3 ( x , y ) = X 3 ( x )Y0 ( y ) , w4 ( x , y ) = X 1 ( x )Y1 ( y ) , w5 ( x , y ) = X 2 ( x )Y1 ( y ) , w6 ( x , y ) = X 3 ( x )Y1 ( y ) , w7 ( x , y ) = X 1 ( x )Y2 ( y ) , w8 ( x , y ) = X 2 ( x )Y2 ( y ) , w9 ( x , y ) = X 3 ( x )Y2 ( y )

(A10)

14 AIAA Paper 2009-0736

47th AIAA Aerospace Sciences Meeting, 5-8, Jan 2009, Orlando, Florida

For a plate with any combination of fixed and free edges, the non-dimensional natural frequencies can be theoretically determined (Warburton 1954). Nevertheless, from an experimental point of view, optical deformation measurements provide the time-dependent deformation data w( P ,t ) at a number of locations. Given the eigenfunctions wr ( Pi ) ( r = 1, 2 ,L M ) and measured w( Pi ,t ) for a number of points Pi ( i = 1, 2 ,L N ) on the plate, a system of equations for the coefficients r ( t ) ( r = 1, 2 ,L M ) at a given instant is

w( Pi ,t ) =

w ( P ) ( t ) .
r i r r =1

( r = 1, 2 ,L M , i = 1, 2 ,L N )

(A11)

The coefficients r ( t ) can be determined using least-squares method. Furthermore, to determine the damping factors r and the natural frequencies r , the impulse transient solution, r ( t ) exp( r r t ) sin rd t can be utilized in experiments when an impulse force is applied to the plate. The use of spectral analysis for r ( t ) gives rd = r ( 1 r2 )1 / 2 and measurement of the exponential decaying rate gives r r . In general, once r and r are known, the generalized force N r ( t ) can be estimated from the measured magnitudes r ( t ) by using Eq. (A5). An inverse problem is how to recover the fluid-mechanical force F ( P ,t ) from the generalized force N r ( t ) . According to Eq. (A4), N r ( t ) is a projected component of F ( P ,t ) on the eigenfunction wr ( P ) and therefore Fredholm integral equations should be solved for F ( P ,t ) . In general, this inverse problem can be solved by using the regularization method (Tikhonov and Arsenin 1977). Appendix B. Two Vortices on a Flat Plate In order to simulate the observed dual-vortex structure associated with the natural lowfrequency oscillation, the potential flow of two point vortices on a flat plate airfoil is considered as a preliminary model. In the -plane, according the Thomsons circle theorem, the complex potential for an incoming uniform flow over a circle at the origin and two vortices outside the circle is 1 2 a 2 e 2 i 1 +i w( ) = U e i ln + i 2 ln + 2 2 a 2 / 2 2 a / 1 , (B1) 0 + 1 + 2 +i ln 2 where is the angle of attack, a is the radius of a circle, 1 and 2 are the strengths of the vortices #1 and #2, respectively, 1 and 2 are the strengths of the vortices #1 and #2, respectively, and 0 is the strength of a vortex located at the origin to meet the Kutta condition. The complex velocity is

15 AIAA Paper 2009-0736

47th AIAA Aerospace Sciences Meeting, 5-8, Jan 2009, Orlando, Florida

. (B2) 0 + 1 + 2 1 1 1 a2 / + i 2 2 2 By using the Joukowski transformation z = + a 2 / , the circle is transformed to a flat plate The inverse Joukowski transformation is with a chord of c = 4 a in the z-plane.

a 2 e 2 i dw( ) 1 = U e i d 2

1 +i 2

1 1 2 a / 1 1

+i 2 2

= z / 2 ( z / 2 ) 2 ( c / 4 ) 2 , where the positive and negative signs are taken for Re( z ) 0 and Re( z ) < 0 , respectively. The complex velocity in the z-plane is dw[ ( z )] dw( ) 1 = . (B3) dz d dz / d At the trailing edge of the plate, the Kutta condition must be satisfied at z = c / 2 or = a such that 1 1 1 1 0 = ( 1 + 2 ) + 4 a U sin 1 1 / a 1 a / 2 1 / a 1 a / 1 1 2 2 . (B4)
Figure 21(a) shows a typical flow field over a flat plate with two vortices. A question is whether or not this dual-vortex structure is stable. By introducing the non-dimensional variables z* = z / c , t * = tU / c , * = / a and

* = / aU , a dynamical system for the two point vortices on a flat plate in the z-plane is
* dz1

* * = V2 ( z 1 , z 2 ; 1* , 2* , ) , dt * where the complex velocities of the two vortices are * 2 i 1 1 i * i 1 1 e V1 ( z* , z ) e = 1 2 2 2 * * * * 2 1 1 / 1 1 1/ 1 1

dt dz * 2

* * = V1 ( z1 , z 2 ; 1* , 2* , ) ,

(B5)

0* + 1* + 2* 1 1 1 * * * * 1/ * + i 2 1 2 1 2 1 2* 1 1 e 2i i * * V2 ( z 1 , z 2 ) = e 1 2 i 2 * * 2 2 1 / 2 * 11/ * 2 2 . * * * * 0 + 1 + 2 1 1 1 +i +i 1 * * * * * 2 2 2 2 1 2 1/ 1 The non-dimensional strength of the vortex added to meet the Kutta condition is

2* +i 2

16 AIAA Paper 2009-0736

47th AIAA Aerospace Sciences Meeting, 5-8, Jan 2009, Orlando, Florida

0* = ( 1* + 2* ) + 4 sin 1*

1 1 * * 11 1 1/ 1

1 2* 1 * * 1 11/ 2 2

,
2

and the non-dimensional inverse Joukowski transformation is * = 2 z * 4 z * 1 . equilibrium positions of the two vortices satisfy * * V1 ( z 1 , z 2 ; 1* , 2* , ) = 0

The

* * * (B6) V2 ( z * 1 , z 2 ; 1 , 2 , ) = 0 . The equilibrium positions are obtained by solving Eq. (B6), which depend on the three parameters 1* , 2* and . However, calculations indicate that the equilibrium positions are unstable for a finite-amplitude disturbance. This means that the dual-vortex structure on an airfoil cannot remain stationary.

17 AIAA Paper 2009-0736

47th AIAA Aerospace Sciences Meeting, 5-8, Jan 2009, Orlando, Florida

Figure 1. Diagram of stall regimes

Figure 2. Schematics of a thin flexible fin attached to an airfoil for post-stall flow control

Figure 3. A flexible Mylar fin attached on a NACA0012 airfoil model illustrated in a typical PIV image

18 AIAA Paper 2009-0736

47th AIAA Aerospace Sciences Meeting, 5-8, Jan 2009, Orlando, Florida

(a)

(b)

Figure 4. The drag coefficient as a function of AoA for (a) different fin arrangements and (b) a 0.25c fin located at 0.1c

Figure 5. Pressure and skin friction drag coefficients Figure 6. Measured lift coefficient as a function of obtained by CFD AoA

19 AIAA Paper 2009-0736

47th AIAA Aerospace Sciences Meeting, 5-8, Jan 2009, Orlando, Florida

(a) (b) Figure 7. Power spectra of the drag coefficient due to unsteady separation for the baseline NACA0012 model and model with a 0.25c fin located at 0.1c at (a) 14o and (b) 18o

Baseline, AoA 18o

(a)

Fin, AoA 18o

(b)

Figure 8. Mean velocity vectors and vorticity distributions for (a) the baseline NACA0012 model and (b) model with a 0.25c fin located at 0.1c at AoA of 18o

20 AIAA Paper 2009-0736

47th AIAA Aerospace Sciences Meeting, 5-8, Jan 2009, Orlando, Florida

(a)

(b)

Figure 9. Profiles of the mean velocity component U at different locations on the upper surface for (a) AoA 18o and (b) AoA 20o. The solid line: baseline NACA0012 model; the dash line: NACA0012 model with a 0.25c fin located at 0.1c

Figure 10. Effect of the flexible fin on the development of the momentum thickness in the separated region

21 AIAA Paper 2009-0736

47th AIAA Aerospace Sciences Meeting, 5-8, Jan 2009, Orlando, Florida

Figure 11. Power spectra of the velocity component U across the separated region for different x locations for the baseline NACA0012 model at AoA of 18o

22 AIAA Paper 2009-0736

47th AIAA Aerospace Sciences Meeting, 5-8, Jan 2009, Orlando, Florida

Figure 12. Power spectra of the velocity component U across the separated region for different x locations for the NACA0012 model with a 0.25c flexible fin located at 0.1c at AoA of 18o

23 AIAA Paper 2009-0736

47th AIAA Aerospace Sciences Meeting, 5-8, Jan 2009, Orlando, Florida

(a)

(b)

Figure 13. Development of spectral components of the velocity component U across the shear layer in the x direction at AoA of 18o for (a) the baseline NACA0012 model and (b) model with a 0.25c flexible fin located at 0.1c

(a) (b) Figure 14. Reynolds stress in the separation region at AoA of 18o for (a) the baseline NACA0012 model and (b) model with a 0.25c flexible fin located at 0.1c

24 AIAA Paper 2009-0736

47th AIAA Aerospace Sciences Meeting, 5-8, Jan 2009, Orlando, Florida

(a)

(d)

(b)

(e)

(c)

(f)

Figure 15. Snap-shot velocity fields of vortices shedding from the leading edge at (a) 0 s, (b) 0.2 s, (c) 0.4 s, (d) 0.6 s, (e) 0.8 s, and (f) 1.0 s

25 AIAA Paper 2009-0736

47th AIAA Aerospace Sciences Meeting, 5-8, Jan 2009, Orlando, Florida

(a)

(b)

Figure 16. Power spectra of the velocity components U and V on the baseline NACA0012 model

Figure 17. The Strouhal numbers based on the local momentum thickness for the low-frequency mode (f = 0.07 Hz) and the vortex shedding mode (f = 1.0 Hz) for the baseline NACA0012 model

26 AIAA Paper 2009-0736

47th AIAA Aerospace Sciences Meeting, 5-8, Jan 2009, Orlando, Florida

Figure 18. The time traces of the velocity U and the zero-crossing point position at the reference point, where the circles mark the phases for examination of the short-time-averaged velocity fields

Figure 19. The magnitude squared coherence between the velocity U and the zero-crossing point position at the reference point

27 AIAA Paper 2009-0736

47th AIAA Aerospace Sciences Meeting, 5-8, Jan 2009, Orlando, Florida

(a)

(d)

(b)

(e)

(c)

Figure 20. Velocity fields averaged over a short time of 0.73 s at the five phases indicated in Fig. 18 in a low-frequency cycle of the reference velocity U (about 12 s), (a) the first valley, (b) the first middle point between the peak and valley, (c) peak, (d) the second middle point, and (e) the second valley

28 AIAA Paper 2009-0736

47th AIAA Aerospace Sciences Meeting, 5-8, Jan 2009, Orlando, Florida

(a)

(b)

Figure 21. (a) Two vortices on a flat plate at AoA of 18o, and (b) typical trajectories of the two vortices

(a)

(b)

Figure 22. Fin deformation, (a) time trace, and (b) power spectrum of the fin amplitude

29 AIAA Paper 2009-0736

47th AIAA Aerospace Sciences Meeting, 5-8, Jan 2009, Orlando, Florida

Figure 23. Power spectra of the velocity component U across the fin for different locations for the NACA0012 model with a 0.25c fin located at 0.1c at AoA of 18o

30 AIAA Paper 2009-0736

47th AIAA Aerospace Sciences Meeting, 5-8, Jan 2009, Orlando, Florida

Figure 24. Power spectra of the velocity component V across the fin for different locations for the NACA0012 model with a 0.25c fin located at 0.1c at AoA of 18o

31 AIAA Paper 2009-0736

47th AIAA Aerospace Sciences Meeting, 5-8, Jan 2009, Orlando, Florida

Figure 25. Time traces of the relative fin amplitude and zone-averaged velocity components around the fin, where the fin amplitude and velocities are normalized by the fin length and the freestream velocity, respectively

Figure 26. Magnitude squared coherence between the fin amplitude and zone-averaged velocity components

32 AIAA Paper 2009-0736