Determination of the Angle of Attack on a Research Wind Turbine Rotor Blade Using Surface Pressure Measurements

Abstract. In this paper, a method to determine the angle of attack on a wind turbine rotor blade using a chordwise pressure distribution measurement was applied. The approach used a reduced number of pressure tap data located close to the blade leading edge. The results were compared with the measurements from three external probes mounted on the blade at different radial positions and with analytical calculations. Both experimental approaches used in this study are based on the 2-D flow assumption; the pressure tap method is an application of the thin airfoil theory, while the probe method applies geometrical and induction corrections to the measurement data. The experiments were conducted in the wind tunnel at the Hermann Föttinger Institut of the Technische Universität Berlin. The research turbine is a three-bladed upwind horizontal axis wind turbine model with a rotor diameter of 3 m. The measurements were carried out at rated conditions with a tip speed ratio of 4.35, and different yaw and pitch angles were tested in order to compare the approaches over a wide range of conditions. Results show that the pressure tap method is suitable and provides a similar angle of attack to the external probe measurements as well as the analytical calculations. This is a significant step for the experimental determination of the local angle of attack, as it eliminates the need for external probes, which affect the flow over the blade and require additional calibration.


correction was made based on wind tunnel measurement of static blade or airfoils representative of the studied blade section. It is remarkable that the case of the ECN exhibited better results without the upwash correction. This was assumed to be a result of the compensation effect of the downwash from the shed vorticity due to the variation of the bound circulation along the blade span (Schepers et al., 2002)." L46: it depends on the blade length and scaling; a 20mm pitot tube in comparison with a 60m bladeplease adjust this against wind tunnel testing Authors answer: This sentence was rewritten: being explicit that this is relevant in the case of small test turbine models, where this dimensions are comparable: "In general, according to the published literature, external probes can be used to determine the AoA. However, in the case of wind turbine models, such probes are intrusive and significantly disturb the flow over the blade section where are mounted."

L48: no references given
Authors answer: This general statement is now followed by several citations regarding each research that employs surface pressure data.
"Other complementary tools, used on research turbines are surface pressure sensors, located along the blade chord. These sensors are used to record the pressure distribution along the blade chord at a desired radial position and to calculate the aerodynamic loads. Different computational methods use this information has a source to estimate the AoA.
The inverse BEM method is probably the most common. From the surface pressure sensors, the normal and tangential forces are calculated. Assuming that they are uniform over an annulus containing the blade section. The wake-induced velocities are calculated according to momentum theory, yielding the effective velocity vector and subsequently the AoA (Whale et al., 1999). This method was implemented by ECN, NREL, DTU projects, obtaining similar results with their respective probes estimations. Schepers et al. (2012) presented the inverse free wake method applied to the MEXICO rotor, which follows the same BEM principle but using the normal and tangential forces into a free wake model. Several computational methods can be found in the latest phase of the project, summarized by Schepers et al. (2018), such as azimuth average, three point and lifting line average methods among others.
The surface pressure measurements also allow experimental estimations. Shipley et al. (1995) showed the stagnation point normalization method described as follows: the local dynamic pressure is estimated as the maximum value of the pressure side in each pressure distribution station. This value is used to estimated the freestream velocity and then the AoA based on the geometrical velocities defined by pitch, yaw and azimuth angles. Moreover, Brand (1994) presented the stagnation point method. The AoA is estimated as follows: The stagnation point is located as the previous method. Afterwards, the intersection of the chord line and a line normal to the surface at the stagnation point is used to estimate AoA. The position of the point of intersection can be determined 2D approaches either codes or wind tunnel measurement (Whale et al., 1999). The drawback of this method is that it relies only in the geometry of the blade section, assuming AoA and Reynolds number no influence. Furthermore, Bruining and van Rooij (1997) exposed an additional method that use two frontal pressure taps, one on the pressure side and one on the suction side, working as an built-in probe in the blade. The drawback of this is that requires calibrating the blade station where the taps are located. Schepers et al. (2002) reported the comparison between experimental probes, pressure taps and inverse BEM methods regarding the field measurement from ECN, NREL, DUT, DTU and Mie. The main conclusions found were: (1) The ambiguity of the 3D AoA definition implies that any check on accuracy can only be carried out with an arbitrary reference. (2) Before stall, the estimations of the AoA remain with differences below 1°. (3) Above stall conditions, the differences between methods can go up 4°. Table 1 shows field and wind tunnel experiments with the most common estimation methods mentioned above." L51: unclear sentence: With a pitot pressure sensor, you know the position geometrically.
Authors answer: This paragraph has been removed because the wording was confusing. Probes and pressure taps topics are now addressed in separate parts of the literature to improve the text structure. At the same time, the inflow showed some heterogeneity, i.e. was not fully uniform as is depicted in Fig. 2 (left). Figure 2 (right) shows four axial velocity distributions over at the radial positions 45; 65; 75 and 85%R. Therefore, due to these characteristics it was decided to analyze the measurement data over small azimuth angle stations."

L118:
The choice for using Clark-Y is not clear (high drag airfoil), see f. example DOI: 10.2514/6.2006-33 Conference: 44 th AIAA Aerospace Sciences Meeting and Exhibit.
Authors answer: Additional information has been added (see below). Regarding the high drag values in Fig 2 ( DOI: 10.2514DOI: 10. /6.2006) Cl/Cd plot, these are for very low Reynolds number (< 10 5 ). In the present experiment, under rated conditions, the Reynolds number range is 3 10 5 for the radial range 1.7 × 10 5 < < 3 × 10 5 .
"A slightly modified Clark-Y airfoil profile is used along the entire blade span and there is no cylindrical root section. The airfoil modification was necessary in order to account for a realistic trailing edge thickness with respect to manufacturing requirements. Aerodynamically, the design intended to avoid stall while keep offering optimal performance and the maximum internal space to include instrumentation (Pechlivanoglou et al., 2015).
In this way, the specific airfoil profile was chosen as it performs well at low Reynolds number (Re), i.e. at the conditions relevant to BeRT (Re range from 170k to 300k along the span). The blade twist was selected so that the local AoA stays constant over the span at rated conditions"

L113: Model Blockage and consequences for interpretation of results?
Authors answer: A model to analyze the blockage effect has been implemented. This is in terms of the equivalent freestream velocity. Additionally, this is coupled with the geometrical calculations. The hypothesis that the offset in AoA, Δ = 2.3°, between experimental approaches and analytical estimations ( Figure 13 of the original submission given below for reference) was a consequence of the blockage is now strengthened by this correction. More changes due to the inclusion of this correction follow in the next answers to the referees' comments. Figure 15 shows the effect of considering the blockage correction on the analytical calculation.
"The turbine rotor area ( ) produces a considerable blockage ratio in the wind tunnel, = / ≈ 0.4. The blockage effect was analyzed in terms of the equivalent freestream velocity ( ′) which produces the same torque. Glauert (1926) showed that for a propeller the ratio between the wind tunnel velocity ( ∞ ) and its corresponding equivalent freestream velocity is a function of the blockage ratio and the thrust coefficient ( ), Eq. 1. Using the BeRT rotor characteristics reported by Marten et al. (2019), a thrust coefficient of = 0.77 (expected at rated condition) was considered. Subsequently, applying Eq. 1, implemented on wind turbines, results in the velocity ratio of It is noted that this correction has also been applied successfully in wind tunnel experiments with even higher blockage ratio (45% Refan and Hangan, 2012)" L115: is the turbine yaw fixed or free?
Authors answer: Both the turbine yaw angle and the blade pitch angle are fixed during the measurements. The text has been rewritten: "The turbine yaw angle and the blade pitch angle were fixed during the measurements" L124: the statement of placement of pressure taps is not constant=0.45-why straight line placement?/why is it in this small scale experiment not following constant radius?
Authors answer: The reason is related to have good comparability with 2D airfoil studies. A drawing of the pressure taps together with the curvature (red line) is shown here. Although the curvature error is considered small (Δ <0.025 ), it was considered when the pressure was corrected by centrifugal effect as shown Eq 2.
: what is the max frequency (3 dB limit) of the detectable signal. L129 specs?
Authors answer: Spectra from both experimental tools (3-hole probes and pressure tap) have been included in Fig. 7.The same spectra shown in Figure 7 is presented here in [dB], the 3dB line is plotted in red with the max frequency at ≈ 6 . More information regarding the filtering decisions using the spectra can be found in the following answers.

L134: A miss why the use of flaps with consensus on title /intro & science objectives
Authors answer: The TE-flaps were set in their neutral position for all the experiments, and they are given in the description only for completeness of the blade information. The sentence below explicitly states this. Additionally, the same information has been added to the caption of Figure 5.
"The flaps were fixed without any deflection for all test cases presented in this study." L157: using a 3 hole probe-no side slip detection. What about the flow conditions when the turbine is in yaw? L237: the discussion of cross flow (sideslip) for the 2D probe is missing. Or may be your statement is to use a 2D probe in the 3D inflow as a representation of the normal(tangential) velocity components? Clarification and error calculation is needed. L310: the question is if yaw affetcs the pressure in the dynamic inflow field, observed here with a 2D-probe.

Authors answer:
The effectiveness of the 2D probe over the misaligned cases has been addressed and is added here for convenience: "As the turbine was set under yaw misalignments, it is important to verify the effectiveness of the 2D probe. The range of the AoA, in the probe stations, is 0°≤ ≤ 10°. Therefore, adding the corresponding twist angle, the range of the AoA relative to the probes ≤ 18°. Moreover, the probes are aligned with the chord, thus the yaw angle relative to the probe is the same, −30°≤ ≤ 0°. (2016) determined the mono-zone as ±30° ( , ). This zone represents where the calibration parameters of the probes remain invariant, i.e. , . These studies used probes with 7-and 5-holes, respectively. As a 3-hole probe sweep the same angle of these calibrations, its monozone should be the same. Moreover, Bruining and van Rooij (1997) employed 3-hole probes on field measurements with good agreement of the AOA, compared to inverse BEM and stagnation point methods. In addition, Klein et al. (2018) showed similar results from experimental and CFD simulations where the wind tunnel structure was considered. Therefore, based on these arguments, it was assumed that the 3-hole probes are able to estimate the AoA in the yaw misalignments here studied." (Disclaimer: This picture does not belong to any of the authors and is presented here only as a support of the statement. Author reference: Zilliac, G.: Modelling, calibration, and error analysis of seven-hole pressure probes, Experiments in Fluids, 14, 104-120, 1993.) L171: what is the explanation behind seeing the 1P in the signal for the interpretation?

Zilliac (1993) and Moscardi and Johnson
Authors answer: The exact cause of the 1P frequency in the signal is unconfirmed at the moment. It is conceivable that it is caused by either some rotor imbalance or by the tower effect or both.

L174: This is a surprising statement about resonator box system that doesn't damp frequencies.30 Hz filter? The cited reference (Berg) offers fig 21(assuming small tubes) with considerable amplitude and phase lag properties. This needs clarification
Authors answer: The authors realize that the 30Hz filter was unsuitable, as it was higher than the structural frequencies of the blade and rotor ( ≥ 13.5 , ≥ 18 ) and also for dynamic response purposes, as the reviewer rightly highlights.
"The structural design of BeRT results in eigenfrequencies of the blades ≥ 13.5 and the tower ≥ 18 . For this reason, the data were low pass filtered using a Butterworth filter with a cut off frequency of 12 to reduce the noise and structural vibrations." "The dynamic response of the pressure taps/tubes system was evaluated theoretically following the model proposed by Bergh and Tijdeman (1965). Figure 8 (left) shows a scheme of the model used to apply the analysis, based on the tube arrangement depicted in Fig. 5, while Figure 8 (right) shows the theoretical response of the system, based on Bergh and Tijdeman (1965). In order to minimize the attenuation and phase lag of the signal, an additional low pass filter was applied, with a cut off frequency of 6 . This was considered adequate as it shows the amplitude amplification and phase lag are less than 1% and 10°, respectively." To provide further insight, an additional figure is given below, where the effect of different filters is shown. Figure AA1 (left) shows the previous version, where only a low pass filter with a cut off frequency of 30Hz was used. Figure AA1 (right) shows the current results, a cut off frequency of 12 Hz in the case of the 3-hole probes and a lower cut off frequency (6Hz) for the pressure taps, in order to avoid large phase lag and damping. The two measurement tools present an improvement, reducing the vibrations and resulting in a smoother behavior. Figure AA1: AoA results from pressure taps and 3-hole probes approaches. Low pass filter with cut off frequency of 30Hz (left). Low pass filter with a cut off frequency of 12Hz and 6Hz over the 3-hole probes and pressure taps, respectively. Authors answer: This has been included.
"The difference between both curves Δ ≤ 0.05 until = 30% , except the peak at the suction side (Δ ( = 1% ) = 0.2). Afterwards, Δ varies between 0.05 − 0.10." L253: temperature increases in the flow during experiments effects on the pressure sensors (standard calibration at 25deg nom)? As I recall the HCL's have +-0.25%FS nonlinearity & hysteresis. So i would assume higher uncertainty on aoa. Table1 needs to state that uncertainty is [fraction/%] of FSR Authors answer: According to the manufacturer, the ±0.25% is the maximum, the nominal value is ±0.05% . During the experiments, the temperature range was 17.5-19.5°C. The nominal value was considered to calculate the errors.
"During the measurement campaign, between test cases, the tunnel was opened, meanwhile the changes on the pitch or yaw angle were made. This allowed to keep the temperature and relative humidity within range of 18 ± 5° and 40 ± 5%, respectively." Subsequently, the error calculation includes the nominal values, phase standard deviation and the conversion from pressure to AoA: "The measurement uncertainty, for of all quantities, was taken into account in order to quantify the error magnitude over the results. Both AoA estimation approaches have the same iteration in the error propagation, based in the following steps:   Table 2 shows the minimum and maximum values. An example of the uncertainty over the azimuth angle of each tool can be seen in App. D1" It was decided to keep the units in the table to provide the uncertainty of the pressure sensors and AoAs together. An additional figure was added in the appendix (Fig D1), showing the uncertainties over each measurement tool along the azimuth angle in order to visualize their level. Here is for convenience The figures from the section 4 of the Results were redone following the changes detailed above. The changes are regarding the new filtering L266: The results are expressed in Pascal, may be it is more clear to show it relative (normalisation), speaking of uncertainty and also from a point of measurement range.

Authors answer:
The pressure values were normalized by the dynamic pressure inflow. Consequently, the results are now non dimensional. The uncertainties described before are included in the plot and the measurement range and the level of the uncertainty on those ranges.
"In terms of the measurement range, the relative pressure 2.8 ≤ / ∞ ≤ 6.5 . Over this range, the uncertainty error represents the 4.5%. In the case of the pressure difference at 12: 5% , the range is 6 ≤ (12.5% )/ ∞ ≤ 10.3 , where the error takes a value of 4%." " Figure 13 shows the relative dynamic pressure at the radial position = 45\% for the aligned and misaligneds cases, normalized by the dynamic pressure ∞ . It can be seen the same trend between the geometrical case (dashed line) and the estimation from the pressure taps (PP, solid line) as well in the maximum ( = 0°) and minimum ( = 180°) azimuth positions."

Figure 12
Odd, with the 2-2½P variations(L316), except for the tower influence... Check! L334 Could this be the damping effects from the resonating tubes characteristics?, same p variation issue as above Authors answer: The addition of the new filters and the inflow heterogeneity explain the behavior.
"The explanation is due to the heterogeneity of the inflow. These variations, ∞ = ±0.2 −1 (see Fig. 2), can have the same influence that the tower over the AoA estimations. Using the geometrical estimation, , under this level of variation in the inflow, results in AoA difference of = ±0.4°, which support this statement." The main reason was that the filter decision in the first version, did not count the lower Eigen frequencies of the blade and tower. Also with the new filter, it seems that the vibration on the mounting of the 3-hole probes was suppressed.