The ultra-low Reynolds number airfoil wake

Lift force and the near wake of an NACA 0012 airfoil were measured over the angle (α) of attack of 0°–90° and the chord Reynolds number (Re _c ), 5.3 × 10³–5.1 × 10⁴, with a view to understand thoroughly the near wake of the airfoil at low- to ultra-low Re _c . While the lift force is measured using a load cell, the detailed flow structure is captured using laser-Doppler anemometry, particle image velocimetry, and laser-induced fluorescence flow visualization. It has been found that the stall of an airfoil, characterized by a drop in the lift force, occurs at Re _c  ≥ 1.05 × 10⁴ but is absent at Re _c  = 5.3 × 10³. The observation is connected to the presence of the separation bubble at high Re _c but absence of the bubble at ultra-low Re _c , as evidenced in our wake measurements. The near-wake characteristics are examined and discussed in detail, including the vortex formation length, wake width, spanwise vorticity, wake bubble size, wavelength of K–H vortices, Strouhal numbers, and their dependence on α and Re _c .     

The effect of a small isolated roughness element on the forces on a sphere in uniform flow

The effect of an isolated roughness element on the forces on a sphere was examined for a Reynolds number range of 5 × 10⁴ < Re < 5 × 10⁵ using a novel sting-mounted sphere apparatus. The roughness element was a circular cylinder, and its width and height was varied to be 1, 2, and 4% of the sphere diameter. At subcritical Re, a lateral force is produced in the direction of the roughness, while at supercritical Re, the force is in the opposite direction. This is caused by asymmetric boundary layer separation, as shown using particle image velocimetry. At supercritical Re, a roughness element that is only 1% the sphere diameter produces a lift to drag ratio of almost one. It was found that the isolated roughness element has the largest effect on the lateral forces when it is located between a streamwise angle of about 40° and 80°. In addition to the mean forces, the unsteady forces were also measured. It was found that at subcritical Re, vortex shedding is aligned to the plane of the roughness element. In addition, the probability distribution of the forces was nearly Gaussian for subcritical Re, but for supercritical Re, the skewness and kurtosis deviate from Gaussian, and the details are dependent on the roughness size. A simple model developed for the vortical structure formed behind the roughness element can be extended to explain aspects of nominally smooth sphere flow, in which external disturbances perturb the sphere boundary layer in an azimuthally local sense. These results also form the basis of comparison for an investigation into the effectiveness of a moving isolated roughness element for manipulating sphere flow.     

Unsteady force measurements in sphere flow from subcritical to supercritical Reynolds numbers

The flow over a smooth sphere is examined in the Reynolds number range of 5.0 × 10⁴ < Re < 5.0 × 10⁵ via measurements of the fluctuating forces and particle image velocimetry measurements in a planar cut of the velocity field. Comprehensive studies of the statistics and spectra of the forces are presented for a range of subcritical and supercritical Reynolds numbers. While the subcritical lateral force spectra are dominated by activity corresponding to the large-scale vortex shedding frequency at a Strouhal number of approximately 0.18, there is no such peak apparent in the supercritical spectra, although resolution effects may become important in this region. Nor does the large-scale vortex shedding appear to have a significant effect on the drag force fluctuations at either sub- or super-critical Reynolds numbers. A simple double spring model is shown to capture the main features of the lateral force spectra. The low-frequency force fluctuations observed in earlier computational studies are shown to have important implications for statistical convergence, and in particular, the apparent mean side force observed in earlier studies. At least one thousand dimensionless time units are required for reasonable estimates of the second and higher moments below the critical Reynolds number and even more for supercritical flow, stringent conditions for computational studies. Lastly, investigation of the relationship between the motion of the instantaneous wake shape, defined via the local position where the streamwise velocity is equal to half the freestream value, and the in-plane lateral force for subcritical flow reveals a significant negative correlation throughout the near wake, which is shown to be related to a structure inferred to arise from the large-scale vortex shedding convecting downstream at 61% of the freestream velocity. In addition to its utility in understanding basic sphere flow, the apparatus is also a testbed that will be used in future studies, examining the effect of both static and dynamic changes to the surface morphology.     

Reynolds Number Effects on the Coherent Dynamics of the Turbulent Horseshoe Vortex System

The adverse pressure gradient induced by a surface-mounted obstacle in a turbulent boundary layer causes the approaching flow to separate and form a dynamically rich horseshoe vortex system (HSV) in the junction of the obstacle with the wall. The Reynolds number of the flow (Re) is one of the important parameters that control the rich coherent dynamics of the vortex, which are known to give rise to low-frequency, bimodal fluctuations of the velocity field (Devenport and Simpson, J Fluid Mech 210:23–55, 1990; Paik et al., Phys Fluids 19:045107, 2007). We carry out detached eddy simulations (DES) of the flow past a circular cylinder mounted on a rectangular channel for Re = 2.0 × 10⁴ and 3.9 × 10⁴ (Dargahi, Exp Fluids 8:1–12, 1989) in order to systematically investigate the effect of the Reynolds number on the HSV dynamics. The computed results are compared with each other and with previous experimental and computational results for a related junction flow at a much higher Reynolds number (Re = 1.15 × 10⁵) (Devenport and Simpson, J Fluid Mech 210:23–55, 1990; Paik et al., Phys Fluids 19:045107, 2007). The computed results reveal significant variations with Re in terms of the mean-flow quantities, turbulence statistics, and the coherent dynamics of the turbulent HSV. For Re = 2.0 × 10⁴ the HSV system consists of a large number of necklace-type vortices that are shed periodically at higher frequencies than those observed in the Re = 3.9 × 10⁴ case. For this latter case the number of large-scale vortical structures that comprise the instantaneous HSV system is reduced significantly and the flow dynamics becomes quasi-periodic. For both cases, we show that the instantaneous flowfields are dominated by eruptions of wall-generated vorticity associated with the growth of hairpin vortices that wrap around and disorganize the primary HSV system. The intensity and frequency of these eruptions, however, appears to diminish rapidly with decreasing Re. In the high Re case the HSV system consists of a single, highly energetic, large-scale necklace vortex that is aperiodically disorganized by the growth of the hairpin mode. Regardless of the Re, we find pockets in the junction region within which the histograms of velocity fluctuations are bimodal as has also been observed in several previous experimental studies.     

Effect of self-issuing jets along the span on the near-wake of a square cylinder

The study herein focuses on the vortex shedding characteristics and near-wake vorticity patterns of a square cylinder having self-issuing jets through holes along its span. Three different values of spacing between the consecutive holes λ with respect to the cylinder diameter D, i.e., λ/D = 1.5, 3 and 4 are studied experimentally via Digital Particle Image Velocimetry for the Reynolds number range extending from 200 to 1,000. It has been observed that the three-dimensionality of the wake flow depends on the spacing between the holes and Re number. For sufficiently low Reynolds numbers, the jet flows issuing from the holes yield a non-uniform distribution of mean flow characteristics like the shedding frequency and the formation length of vortices along the span of the cylinder when the spacing between jets along centerline is close to wavelength of the naturally existing three-dimensional wake instability. Additionally, for Re number up to 500, the self-issuing jets emanating from the holes show an indirect interaction with shear layers originating from upper and lower separation lines of the cylinder. However, for higher Re numbers of 750 and 1,000, they directly interact with and modify the vortices forming from the cylinder.     

Control of vortex shedding in an axisymmetric bluff body wake

In the present experimental study the effect of a control disc mounted at the rear of an axisymmetric blunt-based body of revolution, first studied by Mair, is investigated in the Reynolds number range 3×10³≤Re_D≤5×10⁴ . As the distance of the control disc from the blunt base is increased, four vortex shedding regimes are identified: at small distances there is no effect, then a sharp increase of vortex shedding activity and total drag is observed, followed by an interval with reduced activity and drag and finally at large distances a regime where the flow around the main body and disc become essentially independent, i.e. where the drag forces of the two elements become additive. The near and far wake velocity fields corresponding to the different regimes are documented with time- and phase-averaged hot-wire and LDA measurements, with spectral analysis of the data and with flow visualizations of the near wake. The results are used to develop an improved understanding of the instability mechanism leading to high vortex shedding activity.     

Suppression of force fluctuations in flow past an aerofoil

Force fluctuations on a solid body are associated with unsteadiness in the wake, e.g. vortex shedding. Therefore, the control of force fluctuations can be realised by suppressing the flow unsteadiness. A NACA0024 aerofoil closed with a round trailing edge is chosen to represent the solid body throughout this investigation, with the Reynolds number fixed at Re = 1000 and angle of attack α ≤ 15^o, at which the uncontrolled flow is two-dimensional. A linear optimal control is calculated by analysing the distribution of sensitivity of unsteadiness to control around the entire surface of the body. The nonlinear effects of the calculated control, which can be actuated through surface-normal suction and blowing across the surface of the aerofoil, are tested through two-dimensional direct numerical simulations. It is observed that a surface-normal velocity control with a maximum magnitude less than 8% of the free stream velocity completely suppresses unsteadiness at α = 10° with an overall drag reduction of 14% and a 138% increase of lift.     

Effect of ventilation on the flowfield around a sphere

The flowfield around a sphere with and without ventilation was investigated in a wind tunnel over a range of Reynolds numbers in an incompressible flow. At supercritical Re, the pressure drag of a sphere can be nearly nullified by venting only 2% of the frontal area of the sphere to the base through a smooth internal duct. The drag reduction is achieved by increased pressures in the separated flow region close to the base. At high Re, the vent flow breaks through the near wake and brings about symmetry in the global flowfield. When the internal shear is increased by using a rough internal duct, the base pressure is unchanged, but the external flow is accelerated to velocities beyond that achieved by the potential flow around the basic sphere. The findings can be explained by a flow model in which the near wake is aerodynamically streamlined by a pair of counterrotating vortex rings at the base. A roughness element can be made to partially destroy the vortex system at the base and result in a steady asymmetric wake. A 1.2 mm diameter wire placed at 70° was found to overtrip the boundary layer and completely destroy the vortex system. Simultaneously, the turbulent separation is advanced and the drag increased.At subcritical Re, ventilation marginally increases static pressures all over the surface. Since the large pressure differential between the windward and leeward sides is not reduced, the internal flow has a rapid acceleration to a velocity close to that of the free stream. The reverse flow associated with the near wake forces the vent flow to rest within itself and the wake profile is unchanged. The main features of subcritical flow around the basic sphere are retained in spite of ventilation. The upstream effects of ventilation are greater for subcritical flow than for supercritical flow.The work reported was carried out under a study grant from the German Academic Exchange Service (DAAD) in Bonn. The authors wish to thank the Director of DAAD in Bonn for the same. Thanks are due to Dr. F. R. Grosche and colleagues at DLR in Göttingen who assisted in the design, fabrication and wind tunnel testing of the sphere model. Thanks are also due to Prof. D. G. Mabey, visiting Professor, Imperial College, London for useful discussions. The many useful discussions with the research advisors of the first author viz., Dr. P. R. Viswanath of National Aerospace Laboratories and Prof. A. Prabhu of Indian Institute of Science, Bangalore are acknowledged with thanks. The support given by the Head, Experimental Aerodynamics Division, National Aerospace Laboratories is thankfully acknowledged.     

Nonintrusive measurements of the boundary layer developing on a single and two circular cylinders

 The boundary layers developing on a single and two tandem circular cylinders were examined using multiple hot-film sensor arrays for Re=2.4–5.1×10⁴. Hot-wire and surface pressure measurements, and smoke-wire flow visualization were also made to better understand the flow pattern and the evolution of the vortex street. The results show that, by use of the sensor arrays in conjunction with a bank of constant-temperature anemometers, (i) the effects of the upstream cylinder on the boundary layer developing on the downstream cylinder, (ii) the frequency of the vortex shedding, and (iii) the locations of flow separation and reattachment can be determined non-intrusively and simultaneously. These measurement capabilities will provide a practical means for the characterization and manipulation of unsteady flow phenomena. Received: 27 October 1996 / Accepted: 13 February 1997     

Time resolved PIV analysis of flow over a NACA 0015 airfoil with Gurney flap

A NACA 0015 airfoil with and without a Gurney flap was studied in a wind tunnel with Re _c = 2.0 × 10⁵ in order to examine the evolving flow structure of the wake through time-resolved PIV and to correlate this structure with time-averaged measurements of the lift coefficient. The Gurney flap, a tab of small length (1–4% of the airfoil chord) that protrudes perpendicular to the chord at the trailing edge, yields a significant and relatively constant lift increment through the linear range of the C _L versus α curve. Two distinct vortex shedding modes were found to exist and interact in the wake downstream of flapped airfoils. The dominant mode resembles a Kàrmàn vortex street shedding behind an asymmetric bluff body. The second mode, which was caused by the intermittent shedding of fluid recirculating in the cavity upstream of the flap, becomes more coherent with increasing angle of attack. For a 4% Gurney flap at α = 8°, the first and second modes corresponded with Strouhal numbers based on flap height of 0.18 and 0.13. Comparison of flow around ‘filled’ and ‘open’ flap configurations suggested that the second shedding mode was responsible for a significant portion of the overall lift increment.     

Shallow wakes generated on smooth and rough surfaces

This study deals with the behavior of shallow turbulent wakes generated on smooth and rough surfaces. The wake generator used is a flat plate placed normal to the flow. Experiments were conducted at flow depths of 40 and 80 mm. The boundary layer thickness in the approaching flow occupies 60–75% of the flow depth. The Reynolds number based on the plate width and approaching freestream velocity varies from 13.0 × 10³ to 14.5 × 10³. Velocity measurements were carried out in the near-wake region (1–10 plate widths) using a laser-Doppler anemometer. The mean velocity distributions at various axial stations collapse onto a single curve by a proper choice of the length and velocity scales. It is important to note that a sense of self-similarity is attained even in the near-wake region. Attempts were made to clarify the relative effects of the transverse shear and bed friction in shallow open channel wakes. Received: 11 February 1999/Accepted: 30 August 2000     

PIV investigation of flowfield behind perforated Gurney-type flaps

The flow behind perforated Gurney-type flaps was investigated by using particle image velocimetry (PIV) at Re = 5.3 × 10⁴. The PIV measurements were supplemented by force balance and surface pressure data. The near wake was disrupted and narrowed, indicative of a reduced drag, with increasing flap perforation and had a drastically suppressed fluctuating intensity. Depending on the strength of the perforation-generated jet, the vortex shedding process behind the flap could be eliminated. The flap porosity also led to reduced positive camber effects and the decompression of the cavity flow (upstream of the flap), as well as decreased upper and lower surface pressures, compared to the solid flap. The reduction in the drag, however, outweighed the loss in lift and rendered an improved lift-to-drag ratio.     

Flow around a circular cylinder—structure of the near wake shear layer

The separated shear layer in the near wake of a circular cylinder was investigated using a single hot wire probe, with special attention given to the shear layer instability characteristics. Without end plates to force parallel vortex shedding, the critical Reynolds number for the onset of the instability was 740. The present data, together with all previously published data, show that the ratio of the instability frequency f_sl to the vortex shedding frequency f_v varies as Re^0.65, which is in agreement with the Re^0.67 dependence obtained by Prasad and Williamson [1997, J Fluid Mech 333:375–402]. However, the distribution of f_sl/f_v and the spectra of the longitudinal velocity fluctuation (u) suggest that, on either side of Re=5,000, the shear layer exhibits lower and upper subcritical regimes, in support of the observations by Norberg [1987, publication no. 87/2, Chalmers University of Technology, Sweden] and Prasad and Williamson [1997, J Fluid Mech 343:235–265]. The spectra of u provide strong evidence for the occurrence of vortex pairing in wake shear layers, suggesting that the near wake develops in a similar manner to a mixing layer.     

Supersonic leeside flow topology on delta wings revisited

The leeside vortex structures on delta wings with sharp leading edges were studied for supersonic flow at the Institute of Theoretical and Applied Mechanics of the Russian Academy of Sciences in Novosibirsk. The experiments were carried out with three wings with sweep angles of χ=68°, 73°, and 78° and parabolic profiles in the 0.6 × 0.6 m² test section of the blow-down wind tunnel T-313 of the institute. The test conditions were varied from Mach numbers M=2 to 4, unit Reynolds numbers from Re _l=26 × 10⁶ to 56 × 10⁶ m⁻¹, and angles of attack from α=0° to 22°. The results of the investigations revealed that for certain flow conditions shocks are formed above, below, and between the primary vortices. The experimental data were accurate enough to detect the onset of secondary and tertiary separation as well as other boundaries. The various flow regimes discussed in the literature were extended in several cases. The major findings are reported. Received: 6 September 1999/Accepted: 24 January 2000     

Effects of a trapped vortex cell on a thick wing airfoil

The effects of a trapped vortex cell (TVC) on the aerodynamic performance of a NACA0024 wing model were investigated experimentally at Re = 10⁶ and 6.67×10⁵6.67\times 10^{5}. The static pressure distributions around the model and the wake velocity profiles were measured to obtain lift and drag coefficients, for both the clean airfoil and the controlled configurations. Suction was applied in the cavity region to stabilize the trapped vortex. For comparison, a classical boundary layer suction configuration was also tested. The drag coefficient curve of the TVC-controlled airfoil showed sharp discontinuities and bifurcative behavior, generating two drag modes. A strong influence of the angle of attack, the suction rate and the Reynolds number on the drag coefficient was observed. With respect to the clean airfoil, the control led to a drag reduction only if the suction was high enough. Compared to the classical boundary layer suction configuration, the drag reduction was higher for the same amount of suction only in a specific range of incidence, i.e., α = −2° to α = 6° and only for the higher Reynolds number. For all the other conditions, the classical boundary layer suction configuration gave better drag performances. Moderate increments of lift were observed for the TVC-controlled airfoil at low incidence, while a 20% lift enhancement was observed in the stall region with respect to the baseline. However, the same lift increments were also observed for the classical boundary layer suction configuration. Pressure fluctuation measurements in the cavity region suggested a very complex interaction of several flow features. The two drag modes were characterized by typical unsteady phenomena observed in rectangular cavity flows, namely the shear layer mode and the wake mode.     

Wake flow pattern modified by small control cylinders at low Reynolds number

Passive wake control behind a circular cylinder in uniform flow is studied by numerical simulation for Re_D ranging from 80 to 300. Two small control cylinders, with diameter d/D=1/8, are placed at x/D=0.5 and y/D=±0.6. Unlike the 1990 results of Strykowski and Sreenivasan, in the present study, the vortex street behind the main cylinder still exists but the fluctuating lift and the form drag on the main cylinder reduces significantly and monotonously as the Reynolds number increases from 80 to 300. Obstruction of the control cylinders to the incoming flow deflects part of the fluid to pass through the gap between the main and control cylinders, forming two symmetric streams. These streams not only eliminate the flow separation along the rear surface of the main cylinder, they also merge toward the wake centerline to create an advancing momentum in the immediate near-wake region. These two effects significantly reduce the wake width behind the main cylinder and lead to monotonous decrease of the form drag as the Reynolds number increases. As the Reynolds number gets higher, a large amount of the downstream advancing momentum significantly delays the vortex formation farther downstream, leading to a more symmetric flow structure in the near-wake region of the main cylinder. As the Reynolds number increases from 80 to 300, both increasing symmetry of the flow structure in the near-wake and significant delay of the vortex formation are the main reasons for the fluctuating lift to decrease monotonously.     

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.     

Near-wake flow structure downwind of a wind turbine in a turbulent boundary layer

Wind turbines operate in the surface layer of the atmospheric boundary layer, where they are subjected to strong wind shear and relatively high turbulence levels. These incoming boundary layer flow characteristics are expected to affect the structure of wind turbine wakes. The near-wake region is characterized by a complex coupled vortex system (including helicoidal tip vortices), unsteadiness and strong turbulence heterogeneity. Limited information about the spatial distribution of turbulence in the near wake, the vortex behavior and their influence on the downwind development of the far wake hinders our capability to predict wind turbine power production and fatigue loads in wind farms. This calls for a better understanding of the spatial distribution of the 3D flow and coherent turbulence structures in the near wake. Systematic wind-tunnel experiments were designed and carried out to characterize the structure of the near-wake flow downwind of a model wind turbine placed in a neutral boundary layer flow. A horizontal-axis, three-blade wind turbine model, with a rotor diameter of 13 cm and the hub height at 10.5 cm, occupied the lowest one-third of the boundary layer. High-resolution particle image velocimetry (PIV) was used to measure velocities in multiple vertical stream-wise planes (x–z) and vertical span-wise planes (y–z). In particular, we identified localized regions of strong vorticity and swirling strength, which are the signature of helicoidal tip vortices. These vortices are most pronounced at the top-tip level and persist up to a distance of two to three rotor diameters downwind. The measurements also reveal strong flow rotation and a highly non-axisymmetric distribution of the mean flow and turbulence structure in the near wake. The results provide new insight into the physical mechanisms that govern the development of the near wake of a wind turbine immersed in a neutral boundary layer. They also serve as important data for the development and validation of numerical models.     

The effect of a splitter plate on the flow around a finite prism

The effect of a wake-mounted splitter plate on the flow around a surface-mounted finite-height square prism was investigated experimentally in a low-speed wind tunnel. Measurements of the mean drag force and vortex shedding frequency were made at Re=7.4×10⁴ for square prisms of aspect ratios AR=9, 7, 5 and 3. Measurements of the mean wake velocity field were made with a seven-hole pressure probe at Re=3.7×10⁴ for square prisms of AR=9 and 5. The relative thickness of the boundary layer on the ground plane was δ/D=1.5–1.6 (where D is the side length of the prism). The splitter plates were mounted vertically from the ground plane on the wake centreline, with a negligible gap between the leading edge of the plate and rear of the prism. The splitter plate heights were always the same as the heights of prisms, while the splitter plate lengths ranged from L/D=1 to 7. Compared to previously published results for an “infinite” square prism, a splitter plate is less effective at drag reduction, but more effective at vortex shedding suppression, when used with a finite-height square prism. Significant reduction in drag was realized only for short prisms (of AR≤5) when long splitter plates (of L/D≥5) were used. In contrast, a splitter plate of length L/D=3 was sufficient to suppress vortex shedding for all aspect ratios tested. Compared to previous results for finite-height circular cylinders, finite-height square prisms typically need longer splitter plates for vortex shedding suppression. The effect of the splitter plate on the mean wake was to narrow the wake width close to the ground plane, stretch and weaken the streamwise vortex structures, and increase the lateral entrainment of ambient fluid towards the wake centreline. The splitter plate has little effect on the mean downwash. Long splitter plates resulted in the formation of additional streamwise vortex structures in the upper part of the wake.     

Direct Numerical Simulation of Turbulent Flow over a Rectangular Trailing Edge

This paper describes a direct numerical simulation (DNS) study of turbulent flow over a rectangular trailing edge at a Reynolds number of 1000, based on the freestream quantities and the trailing edge thickness h; the incoming boundary layer displacement thickness δ^* is approximately equal to h. The time-dependent inflow boundary condition is provided by a separate turbulent boundary layer simulation which is in good agreement with existing computational and experimental data. The turbulent trailing edge flow simulation is carried out using a parallel multi-block code based on finite difference methods and using a multi-grid Poisson solver. The turbulent flow in the near-wake region of the trailing edge has been studied first for the effects of domain size and grid resolution. Then two simulations with a total of 256 × 512 × 64 (∼ 8.4×10⁶) and 512 × 1024 × 128 (∼ 6.7×10⁷) grid points in the computational domain are carried out to investigate the key flow features. Visualization of the instantaneous flow field is used to investigate the complex fluid dynamics taking place in the near-wake region; of particular importance is the interaction between the large-scale spanwise, or Kármán, vortices and the small-scale quasi-streamwise vortices contained within the inflow boundary layer. Comparisons of turbulence statistics including the mean flow quantities are presented, as well as the pressure distributions over the trailing edge. A spectral analysis applied to the force coefficient in the wall normal direction shows that the main shedding frequency is characterized by a Strouhal number based on h of approximately 0.118. Finally, the turbulence kinetic energy budget is analysed. Received 4 March 1999 and accepted 27 October 2000