Saarblitz, Wind farm in HDR,, creative commons by-nc-sa 2.0


Managed by

Site Description: 

All experiments were carried out at the small test section (1.4mx1.8m) of the National Technical University of Athens (NTUA) wind tunnel [1]. The wind tunnel is of the closed single-return type with a total circuit length of 68.81 m. The circuit has a contraction ratio of 6.45 to 1. The free-stream turbulence level in the 3.75 m long octagonal test section is 0.2% with a maximum test section velocity of about 60 m/s. Energy to the flow is given by a 2.67 m diameter eight-bladed fan driven by a 300 bhp DC motor.

Data Provider: 

National Technical University of Athens

Data accesibility: 

Creative Commons Licence
NTUA t18 Test Case by National Technical University of Athens is licensed under a Creative Commons Attribution 4.0 International License.


Pressure measurements
The wing model had 62 pressure taps located at the centre of the wing span. Chordwise they extended from the LE to 88.8% of the chord. All chord pressure channels were fed through a pressure scanner (model FCS421, Furness Controls Ltd) to either a Furness Manometer (FCO16, ±2000 Pa or a Scanivalve pressure transducer (Model J - 500PSI). The pressure measurements from the taps in the vicinity of the zigzag tape or the VGs are not included in the provided data.
The wake rake was 39.1cm wide and consisted of 45 total pressure tubes (of which 44 were used) and two static pressure tubes, located on a different plane from that of the rake. All the tubes were directly connected to the Scanivalve sensor and then through a 16 bit A/D card (National Instruments - USB6251) to the lab computer.
Stereo PIV measurements
Two different TSI Nd:YAG PIV lasers with dual cavities were used depending on availability. For the uncontrolled case a 30mJ laser produced a 2.5mm thick light sheet, whereas a 200mJ laser was used for the VG tests which produced a 1.8mm thick laser sheet at the measurement position. The flow was seeded with oil droplets of mean 1μm diameter created by a commercial generator (TSI model 9307). Two TSI Powerview Plus™ 4MP Cameras were used, located inside the test section 1.2c downstream of the TE. Depending on the size of the desired field of view, three different pairs of lenses, one Nikkon 50mm f/1.8 and one Tamron 90mm f/2.8 and one Sigma 150mm f/2.8, were used. For the uncontrolled case, the lenses were always set at the largest aperture to allow the maximum amount of light to reach the CCD sensor.

Measurement Campaign: 

Pressure measurements
For each set of measurements first the targeted free stream velocity was reached and then the model would be set at the desired angle of attack. The wing pressure measurements were then followed by the wake pressure measurements. All measurements were taken at 200Hz for 5 seconds. A constant misalignment of the model with the flow by 0.2o is allowed for in the results.
Stereo PIV measurements
For the uncontrolled case a pulse separation time of 20μsec was used. For the case with the VGs the pulse separation time was 12μsec.
Two different targets (one 100x100 mm and one 200x200 mm) were used for the calibration of the Stereo PIV experiments, depending on the size of the measurement plane. They were both dual plane double sided allowing the computation of the calibration coefficients without traversing the target in the out-of-plane direction. As a result the velocity component normal to the plane is first order accurate [2].
For some frames the camera field of view (FOV) was larger than the available 20cm x 20cm target (e.g. 10° case, plane at x/c = 1.06). In this way two separate perspective calibrations were performed one for each end of the image. Then the images were analysed twice, once for each side, using the corresponding calibration and mask. At the end, the two set of vectors were brought together based on the initial target positioning. Agreement in the overlapping regions was found to be excellent.
All image post processing was done using TSI Insight 4G software. The overlap between interrogation areas was set to 50% and a Gaussian peak estimator was used. Velocity derivatives were computed using the least squares method which is second order accurate and cancels out the effect of oversampling and produces smoother results (Raffel et al. 1998).
When a standard deviation filter was used to locate outliers, it was found that valid vectors would erroneously be considered as spurious. This was attributed to the fact that the velocity field in each measurement window did not follow a normal distribution (see e.g. Error: Reference source not found and Error: Reference source not found), since apart from the undisturbed flow (high above the airfoil) it included regions of separated flow (for the SC study case) or highly vortical flow (for the VG study case). It was hence decided to omit the standard deviation filter. Given the fact that 2000 snapshots were used to generate the average flow statistics, the random error introduced by a few outliers not excluded was insignificant. A double correlation filter was used to locally examine the validity of the processed vectors. Spurious vectors were replaced using a 3x3 local mean.
A schematic side view of the test set up is given in Figure 1, where the measurement planes for the case without VGs are also shown.

tunnel set up

Figure 1: Schematic planform view of the test set up. The wing, the fences, the zigzag tape and the cameras are shown along with the measurement planes. the measurement planes for the case without VGs are also shown. Planes normal to the flow (A, B and C) are indicated by vertical green lines, while planes normal to the wing span (α, β, γ, δ and ε) are shown with red horizontal lines. Note that planes A, B and C consisted of multiple adjacent measurement frames, which were "patched" together post processing. Planes α to ε were single measurement frames.


Wing Model
The 18% thick airfoil profile was designed at the NTUA [1] and coordinates of the actual wind tunnel model profile are given in the t18real.dat file. The profile belongs to the flat-top type experiencing TE separation leading to a gradual built-up of the lift and smooth post stall behaviour. Under separated flow conditions the flow becomes highly three-dimensional and a single or more Stall Cells (SCs) appear. The inherently unstable SC flow was stabilized by placing a zigzag tape at x/c=0.02 and for only 10% of the span [2, 3].
The wing model had a chord of 0.6m and spanned the test section vertically in order to minimize blockage. The solid blockage of the model was 6.9% of the tunnel section at 12° angle of attack and reached a maximum of 9.2 % at the highest measured angle, 16°, still below the usual upper bound of 10%, see [4].
Plexiglas fences were used to minimize the wind tunnel wall boundary layer effect and the wing aspect ratio was 2.0. The vortex generators will be constructed by a 0.2mm thick aluminum strip so that they will have adequate rigidity and impose minimum distortion to the boundary layer.
Counter rotating triangular vanes with common flow up were used in all measurements with VGs. The VG shape and positioning parameters are shown in Figure 3 and their values are given in Table I. shows the Stereo PIV set-up for the case with the VGs from inside the test section, upstream of the wing model.

Fences and zigzag tape

Figure 2: (a) The fences dimensions; (b) The zigzag tape dimensions.

x = 0.3c:

chordwise position of the VG array, where c is the wing chord

β = 20°:

VG angle to the free stream flow

δ = 6mm:

Boundary layer height at the location of the VG array

h = δ:

VG height

l = 3h:

VG length

D = 11.7h:

distance between two VG pairs

d = 3.7h:

spanwise distance between the LE of two VGs of the same pair

Table I: VG parameters

VG parameters 

Figure 3: Triangular vane vortex generator parameters. (Top left) Wing side view: global positioning parameter, (top right) VG side view: VG shape parameters, (bottom) Top view: relative positioning parameters.

Pressure Coefficients
The lift coefficient (Cl) was computed from the pressure distribution around the airfoil. Since the pressure taps only covered up to 88.8% of the chord, the values reported in the present study are not the full Cl of the profile.
For attached flow conditions the drag coefficient (Cd) was computed from the wake pressure distribution according to [3]. For separated flow conditions the pressure drag was used instead. In order to estimate the pressure drag, an approximation had to be used for the part of the chord that had no pressure taps (x > 88.8%c). It was assumed that the pressure on the suction side remained constant through the separated flow region up to the TE. Then a second order approximation for the pressure on the pressure side was used. No such approximation was used for the Cl computation.
For the case with VGs the drag varied significantly even under attached flow conditions due to the presence of the streamwise vortices shed by the actuators, as expected [4]. The drag was hence measured in four positions downstream of a VG pair and the average value is reported here. The central VG pair was selected, which was downstream of the ZZ tape. The four drag measurement positions are shown in Figure 4, and were the following: Position 0 was between the two VGs of the central VG pair, Position 3 was between two consecutive VG pairs and Positions 1 and 2 were in equal distance between Positions 0 and 3. Although the wake rake method has been used by other researchers in the past [5], in order to measure drag of an airfoil equipped with VGs, it should be noted that it might over predict Cd since rotational losses are also included in the measured drag [3].

Drag measurement positions

Figure 4: Drag measurement positions.

Wind tunnel corrections
Wind tunnel corrections were applied to the measured data according to [3] for the case of a wing spanning the tunnel height. In particular the corrections allowed for the model's solid blockage, the wake blockage and the tunnel walls. The horizontal buoyancy is considered insignificant for 2D airfoil models.
However, these corrections have been developed with the assumption of a two-dimensional flow. Even though the effect of the corrections is small, it has to be mentioned that their application in the present study is somewhat problematic since, as shown latter on, when a SC is formed the wake becomes highly three-dimensional.
It's worth noting that the wake rake position did not affect the measured Cl value at the centre of the wing, i.e. the measured values were always within ± 0.5% regardless of the rake position. It did however affect the static pressure measurements at the lowest positions due to increased blockage. The affected measurements were not used and an average pressure drop along the wind tunnel axis was used for all span positions.
Stereo PIV Sample size effect
The effect of sample size on the measured mean velocities was investigated on two different points, Point A, inside the SC, and Point B, outside of it. Figure 5 and Figure 6 show the results for the streamwise component (U) for the point inside and outside the SC, respectively. The behaviour of the other components was similar so they are not presented here. The mean streamwise velocity at each point based on the maximum sample size, 2000, is drawn as a straight solid line and the 95% confidence interval is given with a dashed line. Mean values computed from groups of 100, 250, 500, and 1000 samples are also plotted in the graphs. As expected the data variation is a lot higher for Point A, inside the SC. 2000 samples were taken for all measurement planes. For this number of samples the average velocity components are measured with a 95% confidence interval of at most ±0.23m/s or ±0.9% of the free stream. For the rms quantities the corresponding confidence interval is 6.0%.

Sample size effect

Figure 5: Sample size effect for the streamwise velocity component (U) at Point A, inside the SC. Data from plane C, normal to the flow at x/c = 1.06, 10° case. 

Sample size effect

Figure 6: Sample size effect for the streamwise velocity component (U) at Point B, outside the SC. Data from plane C, normal to the flow at x/c = 1.06, 10° case.


1. Manolesos, M., Experimental and computational study of three-dimensional separation and separation control using passive vortex generators, 2013, NTUA: PhD Thesis, Athens.

2. Ramasamy, M. and J.G. Leishman. Benchmarking PIV with LDV for Rotor Wake Vortex Flows. in 24 th AIAA Applied Aerodynamics Conference. 2006.

3. Barlow, J.B., W.H. Rae, and A. Pope, Low-speed wind tunnel testing1999, New York: John Wiley & Sons.

4. Timmer, W.A. and R.P.J.O.M. Van Rooij, Summary of the Delft University Wind Turbine Dedicated Airfoils. Journal of Solar Energy Engineering, 2003. 125(4): p. 488-496.

5. Fuglsang, P., et al., Wind tunnel tests of the FFA-W3-241, FFA-W3-301 and NACA 63-430 airfoils., in Risø-R 10411998.


No NDA is required.