PIV compared to classical measurement techniques using the example of turbulent flow in the wake of an air outlet

Transcription

1 PIV compared to classical measurement techniques using the example of turbulent flow in the wake of an air outlet Timo Gericke 1,*, Peter Scholz 1 1: Institut für Strömungsmechanik, Technische Universität Braunschweig, Braunschweig, Germany * Timo Gericke: t.gericke@tu-bs.de Abstract Turbulent flow of a flapped air outlet (AO) was investigated. A flapped AO is used for internal flow systems of aircraft and basically consists of a rectangular channel with a deflection of 45 and an adjustable flap. Flow in the wake of the AO was investigated using mono digital particle image velocimetry (2C2D-PIV), stereo digital particle image velocimetry (3C2D-PIV) and hot-wire anemometry (HWA). Compared to HWA, PIV has a relatively poor temporal resolution due to the low system acquisition rates. The spatial resolution is in much better agreement even if there are differences due to the volume averaging of PIV. On the other hand, flow in the wake of the AO changes significantly in all directions so that a complete set of HWA data will result in a tremendous amount of preparation and measurement time. Therefore, PIV and HWA were used to measure the flow properties of interest at different locations in the wake. The aim of this study was the analysis of the main flow structures and the mean turbulent kinetic energy existing in the wake for different flap angles and three different flow improvement devices. Measurements were made for flap angles of 21, 33 and 45 degree, whereas the flow control devices and geometry changes were applied only for a flap angle of 33 degree. A flap angle of 33 degree is the maximum angle in flight and was chosen for comparison of flow improvement devices due to the larger flow structures compared to 21 degree. All flap angles where tested with a mass flow through the rectangular channel of 0.6 kg/s. Results from both measuring techniques will be compared against each other. The main outcome of the comparison is that especially for the stereo PIV measurements significant errors can occur due to high frequency noise of the PIV measurement chain. In a second step passive flow control devices and geometry modifications were used to improve the flow. The main finding of the second set of experiments is that the flow is relatively insensitive against small changes of the geometry. In order to improve the flow characteristics massive changes of the geometry are necessary or flow control devices such as guide vanes need to be installed. It can be shown that a flap cover reduces the turbulent kinetic energy by about 20%. 1. Introduction Aircraft flow systems are needed for different purposes which are for example providing air for the cabin or for the cooling of heat sources. An AO plays an important role in aircraft flow systems, because the task of an AO is basically the returning of used air to the outside. Therefore, the design of efficient outlets through which the air is returned to the outside stream is essential in order to decrease the drag, increase the efficiency and improve the durability of such outlets and therefore of flow systems. One component which contributes to the drag is the turbulent kinetic energy in the wake of an AO. In order to achieve an accurate result for the turbulent kinetic energy over large areas PIV and HWA measurements were conducted and compared against each other to validate the calculation of turbulent kinetic energy from different PIV measurement techniques. Furthermore, in addition to the original AO, different passive flow control devices and geometry changes were tested in order to improve the flow characteristics. Particle image velocimetry (PIV) is a non-intrusive laser-based measurement technique. Compared to classical measurement techniques such as hot-wire anemometry (HWA), PIV is a relatively young measurement technique, but the amount of PIV experiments is increasing significantly from year to year (Westerweel et al. 2013). Although this technique offers various advantages over classical measurement techniques like the measurement of velocity over a field or a volume instead of a single point, the lack of temporal and spatial resolution compared with HWA is a drawback when resolving all scales of interest for turbulence studies. With the investigated geometry cover of all turbulence scales and their variations over the areas of interest with HWA would be result in a tremendous amount of preparation and measurement time. Moreover, regions exist where measurements are not possible due to geometry limitations or flow structures which cause a serious error in HWA measurements, like recirculation zones. Therefore, the use of both measurement techniques (HWA and PIV) promise to be a good combination to resolve all relevant turbulence scales

2 In the past a lot of experimental studies were carried out to evaluate the capability of PIV regarding turbulence scales (e.g. Knobloch and Fernholz, 2005 or Piirto et al., 2000). In most of these studies well defined flow fields such as grid turbulence, pipe flow, planar shear layers and boundary layers as well as synthetic images have been used for the evaluation. In the present paper an evaluation was carried out using a much more complex case combining single elements of pipe flow, secondary vortices, longitudinal vortices, boundary layers, a three dimensional wake and non planar shear layers. Lavoie et al. (2007) proposed a correction method for PIV measurements in order to achieve more accurate results of the relevant turbulence scales. Without correction of the PIV data the results can lead to a difference in the turbulence kinetic energy of about 20%. Here, PIV is underestimating the turbulent kinetic energy due to the coarser resolution. With the applied correction method the results were virtually identical. The drawback of the proposed correction method is that the velocity spectra must be known a priori (e.g. from CFD) in order to calculate the correction coefficients. This confinement reduces the applicability of this method to a manageable amount of cases where the velocity spectra is well known from theory or validated CFD calculations. Foucaut et al. (2004) showed that the measurement noise of a complete PIV chain is an important factor when comparing the power spectra of HWA and PIV. Interrogation window size plays a major role in order to increase the accuracy of the PIV power spectra. An increase in interrogation window size leads to a decrease of the noise. The optimal interrogation window size is the best compromise between spectral noise and spatial resolution. Li et al. (2009) tested the effect of different pre- and post processing parameters as well as vector calculation parameters like different filters, interrogation window offsets (window shift), interrogation window size and aspect ratios to the mean velocity and the fluctuation velocities. The result of this study showed that the interrogation window size has almost no effect on the fluctuation component and that the smallest interrogation window size lead to the best result. This result is in complete contrast to the findings of Foucaut et al. (2004). Lecordier et al. (2001) evaluated the capability of PIV algorithms (conventional PIV and iterative PIV) to measure turbulence properties using synthetic images from direct numerical simulation. It is clear from the obtained results that the calculation of velocity fluctuations is less sensitive to the interrogation window size for the iterative PIV. Using a window size of 16x16 the conventional PIV overestimates the velocity fluctuations due to the small amount of particles inside a window which increases the high frequency noise. For bigger window sizes conventional PIV underestimates the fluctuations due to smoothing of the velocity over the window. Lin et al. (2008) compared the fluctuations measured with stereo particle image velocimetry in a turbulent boundary layer with hot-wire anemometry results from Carlier and Stanislas (2005). Compared to HWA the fluctuations measured with PIV are lower. Gomes-Fernandes et al. (2012) compared the fluctuations measured with 2C2D-PIV of different fractal grids with HWA results from Mazieller and Vassilicos (2010) with the outcome that the fluctuations measured with PIV can be 20% higher compared to HWA. Stanislas et al. (1998) investigated a turbulent boundary layer at high Reynolds number with PIV and compared the results with hot-wire measurements from Carlier (1996). Both measurement techniques are in quite good agreement for the mean and the fluctuation component except near the wall due to the high velocity gradient inside the interrogation window. It stays unclear from the above experiments if PIV in general over- or underestimates the fluctuations measured with HWA. This strongly depends on the noise level which on the other hand is directly connected to the interrogation window size. Measurements in the wake of the AO were made for flap angles of 21, 33 and 45 degree, whereas the flow control devices and geometry changes were applied only for a flap angle of 33 degree. A flap angle of 33 degree is the maximum angle in flight and was chosen for comparison of flow improvement devices due to the larger flow structures compared to 21 degree. All flap angles where tested with a mass flow through the rectangular channel of 0.6 kg/s. The AO basically consist out of a rectangular channel with a 45 degree deflection and an adjustable flap. From previous studies (e.g. Melling and Whitelaw 1976 or Hoagland 1960) it is well known that in curved pipes, ducts or channels secondary flow occurs. This will most likely also happen in the bend of the AO. The difference to curved pipes, ducts and channels is that those systems are fixed, in terms of the geometry, and closed so that the pipe, duct or channel will continue for at least several diameters. In the case of the flapped air outlet the geometry is only partially fixed and the geometry downstream of the bend changes directly upstream of the bend due to the adjustable flap. Furthermore, the flow from the inside of the AO interacts with the free flow. This interaction leads to an additional deflection of the flow of about 45 degree in flow direction, so that both deflections form an s-shape

3 The purpose of this work is the investigation of the turbulent kinetic energy in the wake of the original AO as well as for different flow control devices and geometry changes. As experimental techniques 2C2D-PIV, 3C2D-PIV and HWA are used. Results of PIV and HWA data will be compared against each other in order to achieve complete fields with validated turbulent kinetic energy. 2. Experimental Setup Experiments were performed in a Göttingen type wind tunnel with a test section of 0.8 m x 0.8 m (Tu = 0.2% at 53 m/s and 0-25 khz) located at the Technische Universität Braunschweig. Freestream velocity was constantly set to Mach = The AO was implemented at a position where the boundary layer has a thickness of around 53 mm. The maximum blockage ratio defined as the ratio of the projected surface of the flap to the test section is 3.6. Experiments were performed at flap angles of 21, 33 and 45 degree, whereas the flow control devices and geometry changes were applied only for a flap angle of 33 degree. All configurations where tested with a mass flow through the rectangular channel of 0.6 kg/s. The tested AO, schematically shown in Fig. 1, is mostly made out of Plexiglas in order to gain optical access, and to reduce reflections that can occur during PIV measurements, to the lowest possible level. Only the flap (fiber composite) itself and the mechanism to move the flap and hold it in position (aluminum) is made out of other materials. The AO consist of six parts which are the 1) diffusor, 2) side walls, 3) channel long wall, 4) channel short wall, 5) flap and 6) outlet wall connection. The inlet of the diffusor has a circular cross sectional shape with a diameter of 138 mm and a rectangular cross sectional shape at the outlet. Therefore, the diffusor opens only to two sides. In Fig. 1 the diffusor is turned by 90 degree to make it visible. Behind the diffusor a rectangular channel follows with a width of 180 mm, a maximum height of 138 mm and a deflection of 45 degree. With the given parameters a maximum Reynolds number of Re ext = based on the flap height in the external (wind tunnel) flow is reached. Mass flow through the AO was provided by a radial-blow fan with a p max of 100 mbar and a maximum mass flow of 1 kg/s. Mass flow was measured with the thermal mass flow measuring system "t-mass 65I" at a distance of 6 m before the AO model. In order to achieve a defined inlet condition an 1100 mm long Plexiglas tube was installed in front of the diffusor. The installation of a longer tube was not possible due to installation constraints. Wind tunnel U y x Wind tunnel y/d = 0.5 y/d = 0 U AO d U Channel short wall z x Flap HWA, 3C2D-PIV x/d = 2 Outlet-wall connection Diffuser 45 Bend U AO Channel long wall Hall a) b) Fig. 1 a) Schematic top view and b) side view of the AO showing the main parts and the measurement positions. In total three different passive flow control devices and geometry changes are investigated. Guide vanes (Fig. 2b), in the following referred to as configuration I, are used to improve the flow inside the AO at the 45 degree deflection. It is known from further investigations that the flow separates in an unsteady manner at both channel radii. These separations lead to an unsymmetrical velocity distribution further downstream between the flap tip and the outlet wall connection with a lower velocity at the flap side for flap angles - 3 -

4 higher than 21 degree. Guide vanes are well known to improve the flow for those geometries. The four guide vanes are designed after Traupel (1977) using conformal mapping as circular arcs with a partition of 26 mm and a vane angle of 31 degree leading to a flow deflection angle of 33 degree. As a result of that deflection angle a nozzle like geometry at the radius of the short wall and a diffuser at the radius of the channel long wall evolve. This arrangement would improve the flow behavior in the upper radius given an adequate mass flow, but would definitely deteriorate the flow separation in the region of the lower radius. Therefore, the vane angle of the guide vanes is adjusted so that the vanes and the radii are parallel. The original diffuser was replaced by a dump-diffusor (Fig. 2c), in the following referred to as configuration II. The aim of this device was a reduction in length of the system and a fixed separation due to the geometry with the scope of a more steady channel flow. The third device was a cover for the flap (Fig. 2d), in the following referred to as configuration III, based on a spindle-like form which was cut on the sides and on the top to achieve the same width and height as the flap in order to not artificially widen the wake. It was expected that this device significantly reduces the size of the wake by reducing the dimensions of the vortices shed from the flap. Wind tunnel a) b) Wind tunnel U U U AO z z x Hall Wind tunnel c) d) x U AO Hall Wind tunnel U U z x U AO Hall Fig. 2 a) Schematic side view of the original AO, b) configuration I with the guide vanes, c) configuration II with the dump diffusor and d) configuration III with the flap covering. z x U AO Hall As experimental techniques PIV and HWA were used. PIV measurements (e.g. Raffel et al or Adrian and Westerweel 2010) were made with a setup for mono and stereo PIV mainly consisting of one or two LaVision Imager pro X 11 megapixel cameras and a Brilliant double-cavity pulsed Nd:YAG laser with a repetition rate of 10 Hz. Zeiss recording lenses were used with a focal length of 50 mm. Three lenses, one concave with a focal length of -50 mm, one convex lens with a focal length of 100 mm and one cylinder lens with a focal length of -25 mm was used to generate the light sheet with a size of about 300 mm x 250 mm. Time between images is chosen as a function of the in-plane pixel shift and was set to 6 µs. Light sheet thickness was set to 1 mm for mono and 2 mm for stereo PIV. The output power of both laser cavities per pulse was 100 mj. The frame rate was 1.3 Hz and in total 1000 double images were recorded. DEHS with a particle size of about 1-2 µm were used as tracer particles. Seeding was inserted to the wind tunnel before the first guide vane array after the measuring section and into the tube from the radial blow fan to the AO. Both streams were adjusted so that the seeding concentration is nearly similar. In total four different measurement planes were investigated for all tested configurations. 2C2D-PIV measurements were made in flow direction at the centerline (y/d = 0) and at the edge of the flap (y/d = 0.5) whereas 3C2D-PIV measurements were made in transverse direction at x = 360 mm (x/d = 2) and x = 450 mm (x/d = 2.5) (Fig. 1). Size of the FOV was 300 mm x 250 mm for the mono setup and 350 mm x 250 mm for the stereo setup. In the case of stereo PIV 2x2 binning was used to achieve a sufficient brightness of the particles in the - 4 -

5 image. Reflections from the wind tunnel wall and the AO were reduced by applying a specific car window foil (Foliatec) at distinct positions. Data analysis was done with the Davis software 7.2 from LaVision. All image pairs were processed by first subtracting the mean background image. Thereafter a multi pass with an interrogation window size of 64x64 and 32x32 was conducted for mono PIV and 32x32 and 16x16 for stereo PIV giving an interrogation window size of 1.5 mm x 1.5 mm and 3.5 mm x 3.5 mm. Overlap was set to 50% using 1 and 2 passes. Hot-wire measurements (Bruun 1995) were made using a single-sensor probe with a plated tungsten wire and a constant temperature anemometer bridge from Dantec. Data was aquired on a data acquisition computer for 26 seconds at a sampling frequency of 20 khz. The hot-wire was calibrated against a prandtl tube. Hot wire measurements were made 90 mm (x = 360 mm) behind the edge of the of the outlet-wall connection at y/d = 0 and 0.5. Every 10mm a set of data was recorded beginning from the wall. The measurements were stopped far away from the last shear layer in the free flow. 3. Results Comparison between HWA and PIV In the following section results of the HWA and PIV measurements are compared to each other at y/d = 0.5 and x = 360 mm (x/d = 2) for a flap angle of 33 degree and a mass flow of 0.6 kg/s. This configuration works as a reference case and the results are similar to all other configurations. In Fig. 3a the mean velocity profiles for the HWA and PIV measurements are plotted over z and in Fig. 3b the corresponding flow field is shown. Overall the agreement of both measurement techniques is good. Only in the convex region at a height of about 60 mm and in the free stream region minor deviations are visible. In the free stream region both PIV method are in good agreement but differ slightly from the HWA measurements. Here, two mechanism lead to this deviation. Namely the relatively steep calibration curve of the HWA system at high velocities, which makes the detection of small changes in velocity difficult and the accuracy of PIV which is 0.2-2% of full scale. The deviation of the 2C2D measurements in the convex region (z = 60 mm) from the other measurements is caused by the fact that in this region the particle distribution has a negative gradient towards the wind tunnel floor due to the merging of the free stream flow with the emanating AO flow in this region. An investigation of the results for different interrogation window sizes showed that for the 2C2D measurements an increase in window size to 64x64 makes no visible difference. For the 3C2D measurements an increase in window size (32x32 and 64x64) directly lead to an edged shape of the velocity profiles due to the increased averaging effect over the interrogation volume. a) b) Fig. 3 a)mean velocities of the HWA and PIV measurements for the original AO with an flap angle of 33 degree and a mass flow of 0.6 kg/s at y/d = 0.5 and x = 360 mm (x/d = 2) b) the corresponding flow field. Additionally to the comparison of the mean velocity the mean fluctuation velocities were also compared. In Fig. 4a the 2C2D-PIV results for an interrogation window size of 32x32 and 64x64 are plotted together with the HWA results. Differences between the PIV results for different window sizes are predominantly an outcome of the averaging effect over a volume of this measuring technique. It can be seen that with a - 5 -

6 window size of 32x32 PIV overestimates the fluctuation components slightly whereas for a window size of 64x64 the fluctuation component is in most parts underestimated by a higher diviation compared to 32x32. The minor difference in the free stream values between HWA and PIV is a clear sign for the existence of high frequency noise from the PIV measuring chain even if it is relatively small. For the 3C2D measurements the presence of high frequency noise is much more obvious (Fig. 4b), which is a well known effect of stereo PIV measurements. Here, the values in the free stream are much higher compared to the HWA measurements. Even for larger interrogation windows the noise is higher compared to 2C2D-PIV. Smaller interrogation windows lead to a lower noise level but have the drawback that due to the higher averaging effect the profiles have a more edge shape like profile and underestimate the fluctuation component clearly. As a result from this comparison only the 2C2D-PIV measurements with an interrogation window size of 32x32 were used to compute the mean turbulent kinetic energy. The use of the 3C2D-PIV measurements for the computation of the turbulent kinetic energy is questionable in the outer parts due to the high frequency noise. Here, further detailed investigations are necessary to improve these results to a level were the agreement between both techniques is more trustful. a) b) Fig. 4 a) Fluctuation velocity from 2C2D-PIV with an interrogation window size of 32x32 and 64x64 compared to the HWA results and b) fluctuation velocity from 3C2D-PIV with an interrogation window size of 16x16, 32x32 and 64x64 compared to the HWA results. Flow fields In the following subchapter the flow fields in the wake of the AO measured with 2C2D- and 3C2D-PIV will be discussed and compared against each other. For both PIV-techniques the original AO with a flap angle of 33 degree and the three flow control devices are combined into one figure. Furthermore, only the flow fields of the 2C2D measurements in the centerline at y/d = 0 are shown. The mass flow rate was set to 0.6 kg/s for all shown cases. Figure 5 show the results of the 2C2D-PIV measurements at y/d = 0 for the original AO with a flap angle of 33 degree and the three investigated flow control devices. The basic flow structures are a main shear layer, a non-uniform flow distribution between the tip of the flap and the channel long wall and recirculation zones from the flow separation at the edge of the outlet-wall connection and from the interaction of the first recirculation zone with the upper channel flow and the main shear layer. As can be seen from Fig. 5 the main shear layer results from the conjunction of the free flow with the channel flow. For the original AO and configuration II the position of the shear layer is nearly identical with an upward direction. In the case of configuration I the shear layer has also an upward direction but is slightly lifted as a result of the stretched recirculation zone at the outlet-wall connection. Enlargement of the recirculation zone in the case of configuration I is a direct result of the better flow deflection in the bend area and the therefore uniform flow distribution parallel to the wall between the flap tip and the channel long wall. Here, the flow leaves the AO at an angle of nearly 45 degree. The shear layer of configuration III has a downward direction which can be explained by the short horizontal approach section upstream of the flap tip instead of an inclined one and by the massive reduction of the flap vortices which for the case of the original AO and configuration I and II - 6 -

7 th 17 International Symposium on Applications of Laser Techniques to Fluid Mechanics lead to an upward motion in this area. Moreover, the stream leaving the AO has the highest velocity at the channel long wall which lead to a suction force due to the low pressure in this area with the result that the high velocity stream coming from the AO and the shear layer converge further downstream. This convergent behavior was only observed in this configuration, whereas in all other cases the high velocity part of the AO stream stays separated from the shear layer over the field of view. The recirculation zone on the outlet wall connection has its largest dimensions for configuration I (see explanation above) and is nearly identical for all other cases with the difference that the recirculation zone for the flap cover is more extended in flow direction than vertically due to the downward direction of the whole flow structure. The second recirculation zone behind the first one is present in the original configuration as well as in configuration I and II to different expansions. For configuration I only a small second recirculation zone is present due to the massive first one. Compared to that configuration the recirculation zone in the original configuration is much more present and smaller than the one existing in the wake of configuration II. Differences in the size of the second recirculation zones are a direct outcome from the velocity distribution of the AO stream and its magnitude. Without guide vanes the stream can be grouped into two regions a low velocity stream underneath the flap and a high velocity stream at the channel long wall. The high velocity stream separates at the edge of the outlet-wall connection and forms the first recirculation zone. Due to the downward motion at the end of the first recirculation zone fluid from the upper half of the wake is sucked downward forming a second recirculation zone which increases in size with an increasing stream velocity. a) b) c) d) Fig. 5 Flow fields in the wake of the AO measured with mono PIV at y/d = 0 for a) the original AO, b) configuration I, c) configuration II and d) configuration III. Flow structures at the edge of the flap tip (y/d = 0.5) show much higher velocities compared to the centerline where the velocity in the lower part has almost reached the free stream value (Fig. 3b) due to the contraction of the wake. Only for the configuration with the flap cover the velocity distribution in the wake is more or -7-

8 less uniform which is a sign for a wider wake compared to the other cases. From the flow structures for a flap angle of 21, 33 and 45 degree it is clear that with an increasing flap angle the efficiency of the AO decreases due to increasing losses. A flap angle of 21 degree shows no obvious recirculation zones in the wake which is a result of the smaller outlet area and the therefore higher outlet velocity which led to a much better flow deflection in this region. Compared to 33 degree a flap angle of 45 degree leads to an enlargement of the first recirculation zone of about 50% and no second recirculation zone is present. The absence of the second recirculation zone is due to the relatively low velocity at the end of the first recirculation zone. Results of the 3C2D-PIV measurements (Fig. 6) at x = 360 mm (x/d = 2) confirm the conclusions made based on the 2C2D-PIV measurements. Moreover, a mushroom like flow structure is present for the configurations with no flap covering. Due to the lower pressure at the flap sides the flow on the flap shows a diverging behavior and leaves the flap under an angle. This transverse flow component together with the free flow forms a vortex at both sides of the flap. With a flap cover the vortices shed from the flap are increased significantly leading to a wider wake. In the present result of the flap cover two counter rotating vortices at the upper edges of the geometry are visible. One vortex pair has its origin at the edges of the small horizontal part upstream of the flap tip. The origin of the second "counter rotating" vortex in this measurement needs to be investigated in further experiments. a) b) c) d) Fig. 6 Flow fields in the wake of the AO measured with stereo PIV at x = 360 mm (x/d = 2) for a) the original AO, b) configuration I, c) configuration II and d) configuration III

9 Turbulent kinetic energy In the following subchapter the turbulent kinetic energy of the 2C2D-PIV measurements will be discussed and compared against each other. The original AO with a flap angle of 33 degree and the three passive flow control devices are combined into one figure. Furthermore, only the results in the centerline at y/d = 0 are shown. The mass flow rate was set to 0.6 kg/s for all measurements. Turbulent kinetic energy in the wake for configuration I differs only slightly from the original AO (Fig. 7). The main difference is the decrease of turbulent kinetic energy in the first recirculation zone and the downstream shifted region with higher turbulent kinetic energy upstream of the recirculation zone. It seems that the guide vanes stabilize the first recirculation zone due to the uniform distribution of the outlet velocity between the flap tip and the edge of the outlet wall connection. A drawback of the uniform velocity distribution at the outlet is the lifted recirculation zone which led to a higher turbulent kinetic energy directly upstream of the recirculation zone. Here, the lifted recirculation zone leads to an upstream motion of high turbulent flow. Configuration II led to a higher velocity difference between the outer edges of the main shear layer because of the decrease in velocity in the upper part and an increase of velocity in the lower part of the stream leaving the outlet. a) b) c) d) Fig. 7 Turbulent kinetic energy in the wake of the AO measured with mono PIV at y/d = 0 for a) the original AO, b) configuration I, c) configuration II and d) configuration III. A higher difference in velocity between the edges of the shear layer led to a higher turbulent kinetic energy inside the shear layer due to an increase of the transverse fluctuation component, which can be seen in Fig. 7c for this configuration. The lowest turbulent kinetic energy is reached with configuration III as can be seen from Fig. 7d. Here, the massive reduction of the flap vortices, a decrease in the velocity difference of the shear layer boundaries and an overall smoother flow are the reasons for the reduction of the turbulent kinetic energy. Turbulent kinetic energy at the edge of the flap tip (y/d = 0.5) show much higher values compared to the centerline (y/d = 0). The first three configurations are very similar and only the results from the flap - 9 -

10 cover show less turbulent kinetic energy which is most likely a result of the wider wake of this configuration. As can be seen from Fig. 8 the mean turbulent kinetic energy over the field of view increases with an increasing flap angle with a higher turbulent kinetic energy at the outer sides of the AO. The higher turbulent kinetic energy is a direct result of the interaction between the flap vortex, the main shear layer and the conjunction of the free stream with the AO stream. Compared to the original AO only configuration III leads to a decrease in turbulent kinetic energy of about 20% in the wake. Configuration I lead to a slightly higher turbulent kinetic energy at the outer part of the flap whereas the kinetic energy at the centerline differs only slightly. For configuration II the turbulent kinetic energy is higher at both locations. Fig. 8 Comparison of the mean turbulent kinetic energy of all configurations calculated based on the 2C2D-PIV measurements at y/d = 0 (filled bars) and y/d = 0.5 (empty bars). 4. Conclusions The flow fields and the mean turbulent kinetic energy of an AO were investigated at three different flap angles. Additionally three different passive flow control devices were tested. All measurements were made with a mass flow through the AO of 0.6 kg/s. As experimental techniques HWA, mono and stereo PIV were used and compared against each other. It was shown that the computation of the fluctuation component and hence the turbulent kinetic energy from PIV data without comparison to classical measuring techniques can lead to significant errors especially for 3C2D-PIV measurements due to high frequency noise. For 2C2D- PIV measurements a study of the right interrogation window size is strongly recommended, were the right window size is always a compromise between the number of particles inside the window, the particle shift and noise from the PIV measurement chain. In future experiments measurements with no particle shift should be undertaken in order to determine the noise without any particle motions which then can be subtracted from the measurements with particle motion. This procedure covers only a part of the noise sources but should lead to more accurate results. Results for the mean velocity show a quite good agreement for all measurement techniques and are much less sensitive to the interrogation window size. Results of the mean turbulent kinetic energy calculated from 2C2D-PIV measurements show that a flap cover (configuration III) reduces the turbulent kinetic energy in the wake of about 20% whereas guide vanes and the dump diffuser lead to an increase of turbulent kinetic energy in the wake. The advantage of the flap cover is the massive decrease of the flap vortices resulting in a wider but smoother wake. Furthermore, an increase in flap angle leads to an increase of turbulent kinetic energy in the wake. Finally, the flow is relatively insensitive against small changes of the geometry. In order to improve the flow characteristics massive changes of the geometry are necessary or flow control devices such as flap covers or guide vanes need to be installed. 5. Acknowledgements We acknowledge the financial support from BMWi in LuFo IV project ELENA. We also acknowledge Martin Schmid from Airbus and Keith Weinmann from DLR for the helpful input and discussions throughout the project

11 6. References Adrian, R J and Westerweel, J (2010) Particle Image Velocimetry. Cambridge University Press, Cambridge Bruun, H H (1995) Hot-Wire Anemometry. Oxford, Oxford New York Tokyo Carlier, J (1996) Caracterisation par anemometrie a fil chauds dune couche limite turbulente a grand nombre de Reynolds. DEA report, Univ. of Lille Carlier, J and Stanislas, M (2005) Experimental study of eddy strucutres in a turbulent boundary layer using particle image velocimetry. J. Fluid Mech., Vol. 535: Fernandes, R G, Ganapathisubramani, B and Vassilicos, J C (2012) Particle image velocimetry study of fractal-generated turbulence. J. Fluid Mech., Vol. 711:1-31 Foucaut, J M, Carlier, J and Stanislas, M (2004) PIV optimization for the study of turbulent flow using spectral analysis. Meas. Sci. Technol., Vol. 15: Hoagland, L C (1960) Fully developed turbulent flow in straight rectangular ducts: Secondary flow its cause and effect on the primary flow. Dissertation, MIT Knobloch, K and Fernholz, H-H (2005) Hot-Wire and PIV Measurements in a High-Reynolds Number Turbulent Boundary Layer. Proceedings in Physics, Volume 101: Lavoie, P, Avallone, G, Gregorio, F, Romano, G P and Antonia, R A (2007) Spatial resolution of PIV for the measurement of turbulence. Exp. Fluids, Vol. 43:39-51 Lecordier, B, Demare, D, Vervisch, L M J, Reveillon, J and Trinite, M (2001) Estimation of the accuracy of PIV treatments for turbulent flow studies by direct numerical simulation of multi-phase flow. Meas. Sci. Technol., Vol. 12: 1-10 Li, C T, Chang, K C and Wang, M R (2009) PIV measurements of turbulent flow in planar mixing layer. Exp. Thermal and Fluid Science, Vol. 33: Lin, J, Foucaut, J M, Laval, J P, Perenne, N and Stanislas, M (2008) Assessment of Different SPIV Processing Methods for an Application to Near-Wall Turbulence. Topics Appl. Physics, Vol 112: Mazellier, N and Vassilicos, J C (2010) Turbulence without Richardson-Kolmogorov cascade. Phys. Fluids, Vol. 22: Melling, A and Whitelaw, J H (1976) Turbulent flow in a rectangular duct. J. Fluid Mech., Vol. 78: Piirto, M, Ihalainen, H, Eloranta, H and Saarenrinne, P (2001) 2D Spectral and Turbulence Length Scale Estimation with PIV. Journal of Visualization, Vol. 4, No. 1: Raffel, M, Willert, C, Wereley, S and Kompenhans, J (2007) Particle Image Velocimetry: A practical guide. 2 nd edn., Springer, Berlin Heidelberg New York Stanislas, M, Carlier, J, Foucaut, J M and Dupont, P (1998) Experimental study of a high Reynolds number turbulent boundary layer using DPIV. 9 th International Symposium on Applications of Laser Techniques to Fluid Mechanics, Lisbon

12 Traupel, W (1977) Thermische Turbomaschinen. 3 rd. edn., Springer, Berlin Heidelberg New York Westerweel, J, Elsinga, G E and Adrian, J R (2013) Particle Image Velocimetry for Complex and Turbulent Flows. Annu. Rev. Fluid Mech., Vol. 45: