Aerodynamic flow vectoring of a wake using asymmetric synthetic jet actuation

This paper reports an experimental investigation on the wake of a blunt-based, flat plate subjected to aerodynamic flow vectoring using asymmetric synthetic jet actuation. Wake vectoring was achieved using a synthetic jet placed at the model base 2.5?mm from the upper corner. The wake Reynolds number based on the plate thickness was 7,200. The synthetic jet actuation frequency was selected to be about 75?% the vortex shedding frequency of the natural wake. At this actuation frequency, the synthetic jet delivered a periodic flow with a momentum coefficient, C _??, of up to 62?%. Simultaneous measurements of the streamwise and transverse components of the velocity were performed using particle image velocimetry (PIV) in the near wake. The results suggested that for significant wake vectoring, vortex shedding must be suppressed first. Under the flow conditions cited above, C _?? values in the range of 10?C20?% were required. The wake vectoring angle seemed to asymptote to a constant value of about 30° at downstream distances, x/h, larger than 4 for C _?? values ranging between 24 and 64?%. The phase-averaged vorticity contours and the phase-averaged normal lift force showed that most of the wake vectoring is produced during the suction phase of the actuation, while the blowing phase was mostly responsible for vortex shedding suppression.     

An experimental study of a turbulent wing-body junction and wake flow

Extensive measurements were conducted in an incompressible turbulent flow around the wing-body junction formed by a 3∶2 semi-elliptic nose/NACA 0020 tail section and a flat plate. Mean and fluctuating velocity measurements were performed adjacent to the wing and up to 11.56 chord lengths downstream. The appendage far wake region was subjected to an adverse pressure gradient. The authors' results show that the characteristic horseshoe vortex flow structure is elliptically shaped, with ? (W)/?Y forming the primary component of the streamwise vorticity. The streamwise development of the flow distortions and vorticity distributions is highly dependent on the geometry-induced pressure gradients and resulting flow skewing directions. The primary goal of this research was to determine the effects of the approach boundary layer characteristics on the junction flow. To accomplish this goal, the authors' results were compared to several other junction flow data sets obtained using the same body shape. The trailing vortex leg flow structure was found to scale on T. A parameter known as the momentum deficit factor (MDF = (Re _T)² (θ/T)) was found to correlate the observed trends in mean flow distortion magnitudes and vorticity distribution. Changes in δ/T were seen to affect the distribution of u′, with lower ratios producing well defined local turbulence maxima. Increased thinning of the boundary layer near the appendage was also observed for small values of δ/T.     

On the Vortex Dynamic of Airflow Reattachment Forced by a Single Non-thermal Plasma Discharge Actuator

Commercial and military aircrafts or miniature aerial vehicles can suffer from massive flow separation when high angles of attack are required. Single dielectric barrier discharge (DBD) actuators have demonstrated their capability of controlling such a separated flow at low external velocity. However, the processes resulting in the improvement of the flight performances remain unclear. In the present study, the reattachment process along the suction side of a NACA 0015 placed at an angle of attack of 16° is experimentally investigated for an external velocity of 20 m/s (Re = 260,000). A single DBD actuator is mounted at the leading edge of the model. The velocity fields above the suction side of the airfoil are measured by a high-speed acquisition system (3 kHz). The results indicate that the baseline flow presents shed vortices that form at the leading edge and linearly grow along the free shear layer axis. This vortex shedding is organized and exhibits a specific frequency of about 90 Hz. The continuous actuation produces a partial flow reattachment up to 70% of the chord length. Temporal cross-correlation function indicates the presence of a vortex shedding at the trailing edge of the controlled flow. Finally, the temporal analysis demonstrates that the reattachment process requires 50 ms to reach a stabilized attached flow. The time-resolved analysis of the reattachment suggests that the actuation by plasma discharge acts as a catalyser by reinforcing one of the coherent flow structures already existing in the natural flow.     

INFLUENCE OF SEPARATION ON SOUND GENERATED BY VORTEX-STEP INTERACTION

An analysis of the sound produced when a line vortex interacts at low Mach number with forward or backward facing steps is made. The radiation is dominated by an aeroacoustic dipole whose strength is equal to the unsteady drag on the step. The drag is determined by the vorticity distribution, and a correct estimate of the sound must therefore include contributions from vorticity in the separated flow induced by the vortex. The separation is modelled by assuming that the shed vorticity rolls up into a concentrated core, fed by a connecting sheet from the edge of the step of negligible circulation. The motion everywhere is irrotational except at the impinging vortex and the separation core, and the trajectory of the core is governed by an emended Brown & Michael equation. For large steps it is found that estimates of the generated sound that neglect separation are typically an order of magnitude too large. The sound levels predicted for small steps with and without separation are of comparable magnitudes, although the respectivephasesare different.Turbulentflow over a step frequently involves separation and large surface pressure fluctuations at reattachment zones. The results of this paper suggest that numerical schemes for determining the noise generated by turbulent flow over a step must take proper account of “forcing” of the separation region by the impinging turbulence and of vorticity production via the no-slip condition.     

Experimental investigations of controlled transition of a laminar separation bubble in an axisymmetric diffuser

The instability of a pressure-induced laminar separation bubble is examined experimentally on an axisymmetric diffuser for a Reynolds number range 7,800 ≤ ≤ 11,400 for an inlet pipe diameter D ₁ (50 mm) and as mean input flow velocity 4.2 m/s ≤ u _m ≤ 6.1 m/s. A characterization of the base flow shows a wide-spread separation at the smooth diverging contour which gives rise to a massive amplification of instabilities. Controlled disturbances are introduced by means of a slot and a membrane actuator to trigger the transition, and the receptivity of the perturbations to the laminar boundary layer is evaluated. Different axisymmetric and azimuthal disturbances are applied in order to study their influence on the laminar–turbulent transition. The measurements show a clear dependence of the transition scenario and the reattachment length on the actuation mode.     

Aerodynamics and heat transfer in a separated flow in an axisymmetric diffuser with sudden expansion

Results of a numerical study of the influence of a positive pressure gradient in an axisymmetric diffuser with sudden expansion of a circular tube on aerodynamics and turbulent heat transfer in regions of flow separation, reattachment, and relaxation are reported. The air flow prior to separation is assumed to be fully turbulent and to have a constant Reynolds number Re _D1 = 2.75 · 10⁴. The tube expansion degree is 1.78, and the apex half-angle of the diffuser is varied from 0 to 5°. It is found that an increase in the pressure gradient leads to a decrease in the heat transfer intensity in the separation region, and the maximum heat release point moves away from the flow separation point. The calculated results are compared with experimental data. It is shown that the behavior of the separated flow behind the step becomes significantly different as the streamwise pressure gradient changes.     

Control of diffuser jet flow: turbulent kinetic energy and jet spreading enhancements assisted by a non-thermal plasma discharge

An axisymmetric air jet exhausting from a 22-degree-angle diffuser is investigated experimentally by particle image velocimetry (PIV) and stereo-PIV measurements. Two opposite dielectric barrier discharge (DBD) actuators are placed along the lips of the diffuser in order to force the mixing by a co-flow actuation. The electrohydrodynamic forces generated by both actuators modify and excite the turbulent shear layer at the diffuser jet exit. Primary air jet velocities from 10 to 40 m/s are studied (Reynolds numbers ranging from 3.2 to 12.8 × 10⁴), and baseline and forced flows are compared by analysing streamwise and cross-stream PIV fields. The mixing enhancement in the near field region is characterized by the potential core length, the centreline turbulent kinetic energy (TKE), the integrated value of the TKE over various slices along the jet, the turbulent Reynolds stresses and the vorticity fields. The time-averaged fields demonstrate that an effective increase in mixing is achieved by a forced flow reattachment along the wall of the diffuser at 10 m/s, whereas mixing enhancement is realized by excitation of the coherent structures for a primary velocity of 20 and 30 m/s. The actuation introduces two pairs of contra-rotating vortices above each actuator. These structures entrain the higher speed core fluid toward the ambient air. Unsteady actuations over Strouhal numbers ranging from 0.08 to 1 are also studied. The results suggest that the excitation at a Strouhal number around 0.3 is more effective to enhance the turbulence kinetic energy in the near-field region for primary jet velocity up to 30 m/s.     

Effect of velocity ratio on the streamwise vortex structures in the wake of a stack

The time-averaged velocity and streamwise vorticity fields within the wake of a stack were investigated in a low-speed wind tunnel using a seven-hole pressure probe. The experiments were conducted at a Reynolds number, based on the stack external diameter, of Re_D=2.3×10⁴. The stack, of aspect ratio AR=9, was mounted normal to a ground plane and was partially immersed in a flat-plate turbulent boundary layer, where the ratio of the boundary layer thickness to the stack height was δ/H≈0.5. The jet-to-cross-flow velocity ratio was varied from R=0 to 3, which covered the downwash, crosswind-dominated and jet-dominated flow regimes. In the downwash and crosswind-dominated flow regimes, two pairs of counter-rotating streamwise vortex structures were identified within the stack wake. The tip vortex pair located close to the free end of the stack, and the base vortex pair located close to the ground plane within the flat-plate boundary layer, were similar to those found in the wake of a finite circular cylinder, and were associated with the upwash and downwash flow fields within the stack wake, respectively. In the jet-dominated flow regime, a third pair of streamwise vortex structures was observed, referred to as the jet-wake vortex pair, which occurred within the jet-wake region above the free end of the stack. The jet-wake vortex pair had the same orientation as the base vortex pair and was associated with the jet rise. The peak vorticity and strength of the streamwise vortex structures were functions of the jet-to-cross-flow velocity ratio. For the tip vortex structures, their peak vorticity and strength reduced as the jet-to-cross-flow velocity ratio increased.     

Time-resolved and volumetric PIV measurements of a transitional separation bubble on an SD7003 airfoil

To comprehensively understand the effects of Kelvin–Helmholtz instabilities on a transitional separation bubble on the suction side of an airfoil regarding as to flapping of the bubble and its impact on the airfoil performance, the temporal and spatial structure of the vortices occurring at the downstream end of the separation bubble is investigated. Since the bubble variation leads to a change of the pressure distribution, the investigation of the instantaneous velocity field is essential to understand the details of the overall airfoil performance. This vortex formation in the reattachment region on the upper surface of an SD7003 airfoil is analyzed in detail at different angles of attack. At a Reynolds number Re _c < 100,000 the laminar boundary layer separates at angles of attack >4°. Due to transition processes, turbulent reattachment of the separated shear layer occurs enclosing a locally confined recirculation region. To identify the location of the separation bubble and to describe the dynamics of the reattachment, a time-resolved PIV measurement in a single light-sheet is performed. To elucidate the spatial structure of the flow patterns in the reattachment region in time and space, a stereo scanning PIV set-up is applied. The flow field is recorded in at least ten successive light-sheet planes with two high-speed cameras enclosing a viewing angle of 65° to detect all three velocity components within a light-sheet leading to a time-resolved volumetric measurement due to a high scanning speed. The measurements evidence the development of quasi-periodic vortex structures. The temporal dynamics of the vortex roll-up, initialized by the Kelvin–Helmholtz (KH) instability, is shown as well as the spatial development of the vortex roll-up process. Based on these measurements a model for the evolving vortex structure consisting of the formation of c-shape vortices and their transformation into screwdriver vortices is introduced.     

Scanning PIV investigation of the laminar separation bubble on a SD7003 airfoil

A laminar separation bubble occurs on the suction side of the SD7003 airfoil at an angle of attack α =  4–8° and a low Reynolds number less than 100,000, which brings about a significant adverse aerodynamic effect. The spatial and temporal structure of the laminar separation bubble was studied using the scanning PIV method at α =  4° and Re = 60,000 and 20,000. Of particular interest are the dynamic vortex behavior in transition process and the subsequent vortex evolution in the turbulent boundary layer. The flow was continuously sampled in a stack of parallel illuminated planes from two orthogonal views with a frequency of hundreds Hz, and PIV cross-correlation was performed to obtain the 2D velocity field in each plane. Results of both the single-sliced and the volumetric presentations of the laminar separation bubble reveal vortex shedding in transition near the reattachment region at Re = 60,000. In a relatively long distance vortices characterized by paired wall-normal vorticity packets retain their identities in the reattached turbulent boundary layer, though vortices interact through tearing, stretching and tilting. Compared with the restricted LSB at Re = 60,000, the flow at Re = 20,000 presents an earlier separation and a significantly increased reversed flow region followed by “huge” vortical structures.     

Visualization of a locally-forced separated flow over a backward-facing step

A laboratory water channel experiment was made of the separated flow over a backward-facing step. The flow was excited by a sinusoidally oscillating jet issuing from a separation line. The slit was connected to a cavity in which water was forced through a rigid pipe by a scotch-yoke system. The Reynolds number based on the step height (H) was fixed at Re _H =1200. The forcing frequency was varied in the range 0.305?St _H ?0.955 at the forcing amplitude A ₀=0.3. Time-averaged flow measurements were made by a LDV system, especially in the recirculating region behind the backward-facing step. To characterize the large-scale vortex evolution due to the local forcing, flow visualizations were performed by a dye tracer method with fluorescent ink. The vortex amalgamation process was captured at the effective forcing frequency (St _H =0.477) for laminar separation. This vortex merging process enhances flow mixing, which leads to the shortening of the reattachment length.     

DNS and LES of Passively Controlled Turbulent Backward-Facing Step Flow

A passive control approach (no external energy input) for an unsteady separated flow case was investigated numerically. A surface-mounted control fence was positioned upstream of a backward-facing step, and as an oncoming flow a thin and fully developed turbulent boundary layer with a thickness of δ/h = 0.8 was used. The objective of the passive control was to enhance the entrainment rate of the shear layer bounding the separation zone behind the step, thereby reducing the mean reattachment length,〈 X _r0 〉. Direct Numerical Simulations (DNS) and Large-Eddy Simulations (LES) at Re _h = 3000 (based on the step height, h, and the free stream velocity, U _∞) were carried out for the uncontrolled and the controlled flow case. The LES results were in good agreement with the DNS reference solutions. Adaptively controlled feedback simulations showed that a certain minimum distance between the step edge and the upstream position of the control fence is required to achieve a maximum reduction of the reattachment length.     

Accuracy of out-of-plane vorticity measurements derived from in-plane velocity field data

 A study of the errors in out-of-plane vorticity (ω _z ) calculated using a local χ² fitting of the measured velocity field and analytic differentiation has been carried out. The primary factors of spatial velocity sampling separation and random velocity measurement error have been investigated. In principle the ω _z error can be decomposed into a bias error contribution and a random error contribution. Theoretical expressions for the transmission of the random velocity error into the random vorticity error have been derived. The velocity and vorticity field of the Oseen vortex has been used as a typical vortex structure in this study. Data of different quality, ranging from exact velocity vectors of analytically defined flow fields (Oseen vortex flow) sampled at discrete locations to computer generated digital image frames analysed using cross-correlation DPIV, have been investigated in this study. This data has been used to provide support for the theoretical random error results, to isolate the different sources of error and to determine their effect on ω _z measurements. A method for estimating in-situ the velocity random error is presented. This estimate coupled with the theoretically derived random error transmission results for the χ² vorticity calculation method can be used a priori to estimate the magnitude of the random error in ω _z . This random error is independent of a particular flow field. The velocity sampling separation is found to have a profound effect on the precise determination of ω _z by introducing a bias error. This bias error results in an underestimation of the peak vorticity. Simple equations, which are based on a local model of the Oseen vortex around the peak vorticity region, allowing the prediction of the ω _z bias error for the χ² vorticity calculation method, are presented. An important conclusion of this study is that the random error transmission factor and the bias error cannot be minimised simultaneously. Both depend on the velocity sampling separation, but with opposing effects. The application of the random and bias vorticity error predictions are illustrated by application to experimental velocity data determined using cross-correlation DPIV (CCDPIV) analysis of digital images of a laminar vortex ring. Received: 31 October 1997/Accepted: 6 February 1998     

On the generation of yawing moment using active flow control

The generation of control moments without moving control surfaces is of great practical importance. Following a successful flight demonstration of creating roll motion without ailerons using differential, lift oriented, flow control the current study is a first step towards generating yawing motion via differential flow controlled drag.A wind tunnel study was conducted on a 21% thick Glauert type airfoil. The upper surface flow is partially separated from the two-thirds chord location and downstream on this airfoil at all incidence angles. An array of mass-less Piezo-fluidic actuators, located at x/c = 0.65, are capable of fully reattaching the flow in a gradual, controlled manner. The actuators are individually operated such that the boundary layer could be controlled in a 3D fashion.Several concepts for creating yaw motion without moving control surface are examined. The ultimate goal is to generate the same lift on both wings, while decreasing the drag on one wing and increasing the drag on the other, therefore creating a yawing moment. Decreased drag is created by effective part-span separation delay while increased drag can be created by enhanced generation of vortex shedding or by highly localized 3D actuation.Detailed measurements of 3D surface pressure distributions and wake data with three velocity and streamwise vorticity components are presented and discussed along with surface flow visualization images. The data provide evidence that yawing moments can be generated with AFC.     

A vorticity stretching diagnostic for turbulent and transitional flows

Vorticity stretching in wall-bounded turbulent and transitional flows has been investigated by means of a new diagnostic measure, denoted by Γ, designed to pick up regions with large amounts of vorticity stretching. It is based on the maximum vorticity stretching component in every spatial point, thus yielding a three-dimensional scalar field. The measure was applied in four different flows with increasing complexity: (a) the near-wall cycle in an asymptotic suction boundary layer (ASBL), (b) K-type transition in a plane channel flow, (c) fully turbulent channel flow at Re _τ = 180 and (d) a complex turbulent three-dimensional separated flow. Instantaneous data show that the coherent structures associated with intense vorticity stretching in all four cases have the shape of flat ‘pancake’ structures in the vicinity of high-speed streaks, here denoted ‘h-type’ events. The other event found is of ‘l-type’, present on top of an unstable low-speed streak. These events (l-type) are further thought to be associated with the exponential growth of streamwise vorticity in the turbulent near-wall cycle. It was found that the largest occurrence of vorticity stretching in the fully turbulent wall-bounded flows is present at a wall-normal distance of y ⁺?=?6.5, i.e. in the transition between the viscous sublayer and buffer layer. The associated structures have a streamwise length of ~200–300 wall units. In K-type transition, the Γ-measure accurately locates the regions of interest, in particular the formation of high-speed streaks near the wall (h-type) and the appearance of the hairpin vortex (l-type). In the turbulent separated flow, the structures containing large amounts of vorticity stretching increase in size and magnitude in the shear layer upstream of the separation bubble but vanish in the backflow region itself. Overall, the measure proved to be useful in showing growing instabilities before they develop into structures, highlighting the mechanisms creating high shear region on a wall and showing turbulence creation associated with instantaneous separations.     

Evaluation of unsteady pressure fields and forces in rotating airfoils from time-resolved PIV

The final stages of transitional phenomena in laminar separation bubbles play a key role in their reattachment process, and they condition the boundary layer properties and flow structure after reattachment. In this experimental study, the evolution of the perturbation velocity spectra found in this zone is first presented, showing the nonlinear growth of instabilities in their path to develop fully turbulent spectra. The study of the average flow field allows the scaling of the reattachment region, both in its extension and in the characterization of the integral boundary layer magnitudes. Experimental laws are proposed for the evolution of the momentum thickness and of the shape factor. In addition, a universal, wake-like mean velocity profile is found shortly after the reattachment station. The phase-locked characterization technique allows measurements conditioned to the presence of a fluid event. This technique is used to track the evolution of large-scale structures, whose dynamics is seen to dominate the fluid behavior in the reattachment zone. The simultaneous existence of two vortex blobs is found to characterize this flow region, with the longest lived one being convected toward the wall and stretched. This process results in the fast breakdown of the large-scale vorticity structure and the sudden formation of 3-D, small scales that promote the rapid flow evolution toward a fully developed turbulent state.     

Flow visualization of supersonic laminar flow over a backward-facing step via NPLS

An experimental study on a supersonic laminar flow over a backward-facing step of 5 mm height was undertaken in a low-noise indraft wind tunnel. To investigate the fine structures of Ma = 3.0 and 3.8 laminar flow over a backward-facing step, nanotracer planar laser scattering was adopted for flow visualization. Flow structures, including supersonic laminar boundary layer, separation, reattachment, redeveloping turbulent boundary layer, expansion wave fan and reattachment shock, were revealed in the transient flow fields. In the Ma = 3.0 BFS (backward-facing step) flow, by measuring four typical regions, it could be found that the emergence of weak shock waves was related to the K–H (Kelvin–Helmholtz) vortex which appeared in the free shear layer and that the convergence of these waves into a reattachment shock was distinct. Based on large numbers of measurements, the structure of time-averaging flow field could be gained. Reattachment occurred at the location downstream from the step, about 7–7.5 h distance. After reattachment, the recovery boundary layer developed into turbulence quickly and its thickness increased at an angle of 4.6°. At the location of X = 14h, the redeveloping boundary layer was about ten times thicker than its original thickness, but it still had not changed into fully developed turbulence. However, in the Ma = 3.8 flow, the emergence of weak shock waves could be seen seldom, due to the decrease of expansion. The reattachment point was thought to be near X = 15h according to the averaging result. The reattachment shock was not legible, which meant the expansion and compression effects were not intensive.     

Unsteady separated and reattaching turbulent flow over a two-dimensional square rib

The spatio-temporal characteristics of the separated and reattaching turbulent flow over a two-dimensional square rib were studied experimentally. Synchronized measurements of wall-pressure fluctuations and velocity fluctuations were made using a microphone array and a split-fiber film, respectively. Profiles of time-averaged streamwise velocity and wall-pressure fluctuations showed that the shear layer separated from the leading edge of the rib sweeps past the rib and directly reattaches on the bottom wall (x/H=9.75) downstream of the rib. A thin region of reverse flow was formed above the rib. The shedding large-scale vortical structures (fH/U₀=0.03) and the flapping separation bubble (fH/U₀=0.0075) could be discerned in the wall-pressure spectra. A multi-resolution analysis based on the maximum overlap discrete wavelet transform (MODWT) was performed to extract the intermittent events associated with the shedding large-scale vortical structures and the flapping separation bubble. The convective dynamics of the large-scale vortical structures were analyzed in terms of the autocorrelation of the continuous wavelet-transformed wall pressure, cross-correlation of the wall-pressure fluctuations, and the cross-correlation between the wall pressure at the time-averaged reattachment point and the streamwise velocity field. The convection speeds of the large-scale vortical structures before and after the reattachment point were U_c=0.35U₀ and 0.45U₀, respectively. The flapping motion of the separation bubble was analyzed in terms of the conditionally averaged reverse-flow intermittency near the wall region. The instantaneous reattachment point in response to the flapping motion was obtained; these findings established that the reattachment zone was a 1.2H-long region centered at x/H=9.75. The reverse-flow intermittency in one period of the flapping motion demonstrated that the thin reverse flow above the rib is influenced by the flapping motion of the separation bubble behind the rib.     

Mechanism of compressor airfoil boundary layer flow control using nanosecond plasma actuation

The flow control effects of nanosecond plasma actuation on the boundary layer flow of a typical compressor controlled diffusion airfoil are investigated using large eddy simulation method. Three types of plasma actuation are designed to control the boundary layer flow, and two mechanisms of compressor airfoil boundary layer flow control using nanosecond plasma actuation have been found. The plasma actuations located within the laminar boundary layer flow can induce a small vortex structure through influencing on the density and pressure of the flow field. As the small vortex structure moves downstream along the blade surface with the main flow, it can suppress the turbulent flow mixing and reduce the total pressure loss. The flow control effect of the small vortex structure is summarized as wall jet effect. Differently, the plasma actuation located within the turbulent boundary layer flow can act on the shear layer flow and induce a large vortex structure. While moving downstream, this large vortex structure can suppress the turbulent flow mixing too.     

Sensitivity of an asymmetric 3D diffuser to vortex-generator induced inlet condition perturbations

Modifications of the turbulent separated flow in an asymmetric three-dimensional diffuser due to inlet condition perturbations were investigated using conventional static pressure measurements and velocity data acquired using magnetic resonance velocimetry (MRV). Previous experiments and simulations revealed a strong sensitivity of the diffuser performance to weak secondary flows in the inlet. The present, more detailed experiments were conducted to obtain a better understanding of this sensitivity. Pressure data were acquired in an airflow apparatus at an inlet Reynolds number of 10,000. The diffuser pressure recovery was strongly affected by a pair of longitudinal vortices injected along one wall of the inlet channel using either dielectric barrier discharge plasma actuators or conventional half-delta wing vortex generators. MRV measurements were obtained in a water flow apparatus at matched Reynolds number for two different cases with passive vortex generators. The first case had a pair of counter-rotating longitudinal vortices embedded in the boundary layer near the center of the expanding wall of the diffuser such that the flow on the outsides of the vortices was directed toward the wall. The MRV data showed that the three-dimensional separation bubble initially grew much slower causing a rapid early reduction in the core flow velocity and a consequent reduction of total pressure losses due to turbulent mixing. This produced a 13% increase in the overall pressure recovery. For the second case, the vortices rotated in the opposite sense, and the image vortices pushed them into the corners. This led to a very rapid initial growth of the separation bubble and formation of strong swirl at the diffuser exit. These changes resulted in a 17% reduction in the overall pressure recovery for this case. The results emphasize the extreme sensitivity of 3D separated flows to weak perturbations.