You are on page 1of 11

Wind Tunnel Testing Of Airfoils Involves More Than Just Wall Corrections

Peter Fuglsang, Stefano Bove


LM Glasfiber
Rolles Møllevej 1
DK-6640 Lunderskov
e-mail: pfu@lmglasfiber.com
phone: +45 79840469
fax: +45 79840001

Summary
This paper reports an experimental investigation on how to obtain 2D airfoil data in the LM
Glasfiber wind tunnel. An experiment was designed with two airfoil models of different chord
having the same shape. The experiment was used to investigate the influence of model size on
wall corrections. Furthermore the influence from 3D flow for a typical 2D airfoil test setup was
assessed together with the unambiguity of results from different measurement methods. The
applied wall corrections were verified and results were homogenous for Reynolds numbers of 3
and 6 million obtained from different combinations of chord and flow speed. Comparisons of
pressure tap distributions at different span locations showed two dimensional flow for attached
flow, whereas the flow was clearly three dimensional when the flow was separated close to
maximum lift and in post stall. Comparison of lift obtained from airfoil pressure and wall
pressures respectively showed good agreement for attached flow, but some differences for
separated flow. Results for drag showed dependency on time and span averaging. The results
showed that two dimensional results were feasible but that knowledge of the measurement
method is essential when evaluating wind tunnel measurements. A benchmark against
measurements in other wind tunnels showed that the small remaining uncertainties on wall
corrections were insignificant compared to the variation of results between wind tunnels.

1 Introduction
Wind tunnel testing of airfoils for wind turbines has become increasingly popular within the wind
turbine industry. An increased need for experimental verification has appeared as a follow of the
trend to tailor airfoils for operation on specific wind turbine blades as opposed to previously
where airfoils were predominantly selected among a limited variety of existing families, e.g.,
from, [1], [2] and [3] .

The continuous optimization of new blades occurs to increase efficiency as well as to move the
barrier for how long blades can be for both existing and new turbine platforms. The tailoring of
airfoils opens up for further design optimization but also moves us away from our knowledge
base. Numerical predictions have not yet reached a level where the accuracy is sufficient to
eliminate the need for experimental verification.

Wind tunnel testing is also motivated from the need to have accurate blade design data that are
obtained at test conditions matching the operation conditions of the blade on the turbine. Due to
the large size of today’s wind turbine blades the number of suitable wind tunnels is limited with
only little availability of free time slots. This fact together with high ambitions to improve the
aerodynamics of wind turbine blades through extensive wind tunnel testing motivated LM
Glasfiber to build a new wind tunnel (LSWT) from scratch meeting up to date requirements to
testing conditions and test section flow quality. The wind tunnel was inaugurated in 2006 and
has a typical test setup with a 2D airfoil model in a closed test section where a model chord of
0.9 m together with a maximum flow velocity of 105 m/s, which yields a Reynolds number of
6
6x10 .

It is tempting to use a model with a large chord and a high flow speed to match the Reynolds
number of today’s large blades. However, wall corrections due to blockage and interference
become inaccurate. There are also other uncertainties from the three dimensional flow, which
appears when the flow on the upper surface starts to separate causing uncertainty on the lift in
post stall. Also the possible 3D variation of the flow in the span direction of the airfoil model
causes uncertainty on drag.

This paper reports an experimental investigation on how to obtain 2D airfoil data which was
done as a part of the run-in of the LSWT facility. A special experiment was designed with two
airfoil models of the NACA 64-618 shape having different chords of 600 mm and 900 mm. The
model span of 1.35 m spanned the test section which has a height of 2.70 m. The airfoil forces
were derived from different measurement principles typically being used in similar 2D test
setups. Two values were obtained for the airfoil lift coefficient from floor and ceiling wall
pressures and airfoil pressure tabs respectively. The airfoil drag coefficient was obtained from
wake rake measurements.

2 LSWT facility
The LSWT airline has an airline with a closed return loop (Figure 1). The overall outside
dimensions are 37 m x 14 m.

2.1 Airline layout


The airline layout is traditional with diffusers between corners, a single stage fan, a settling
chamber and a contraction. The diffusers as well as the corners are rectangular, except for the
in- and outlets of the fan. The diffusers expand only in the vertical direction between corners,
whereas the corners expand in the horizontal direction. The use of expanding corners is an
innovative feature that reduces the overall loss in the airline as well as the overall size of the
airline. Also, this avoids the need of a wide angle diffuser in front of the settling chamber. To
make this feasible, the corners have tailored vanes with optimized cascade airfoil shapes. The
fan has a maximum power consumption of 1 MW which is sufficient to maintain a flow velocity of
105 m/s in the test section with an inclined airfoil with a chord of 900 mm. The cooling system is
dimensioned to keep temperature constant at all fan powers. A honeycomb structure is
straightening the flow and three subsequent turbulence screens are dampening out fluctuations.
The contraction has a ratio of 10:1 [4].

Cooling
Settling system
Contraction chamber

Test section

Fan
Figure 1 Overview of the LSWT airline with main components

2
2.2 Test section
The main purpose of the facility is to do 2D aerodynamic testing of wind turbine airfoils, as
shown in Figure 2. The test section contains the airfoil model between the two turn tables and
the angle of attack (AOA) of the airfoil model relative to the horizontal inflow is changed by
turning the turn tables. A horizontal and vertical traverse is installed downstream of the turn
tables with a horizontal beam in the test section with a probe holder that moves in the horizontal
direction. This holds a wake rake, which is a structure with a row of total and static pressure
tubes to measure the deficit from the airfoil wake.
o o
The primary angle of attack range for the airfoil is from -20 to 30 to cover the normal operation
range of a rotor blade, which is within the negative stall angle and the angle of attack where
o o
leading edge stall occurs. The secondary range is defined as a full turn from -180 to 180 . This
is fully feasible with the design of the turn tables. However, it is not feasible for a large airfoil
chord due to blockage [5].

The flow quality of the test section has been documented in [6]. Here it was found that the
longitudinal component of the turbulence intensity was 0.1 for a flow speed of 100 m/s (high
pass filtered at 10 Hz). The variation of the flow in both time and space was found to be less
o
than 0.2% of the free flow velocity and the angularity was less than 0.2 .

Figure 2 Side view of test section with an airfoil model between two turn tables.

2.3 Measurement setup


A total of 276 signals are measured. These are mainly differential pressures in addition to a six
component load balance system on the turn tables and measurements of various test section
flow properties.

Figure 3 shows an overview of the measured signals:


• The airfoil model is equipped with pressure tabs on the upper and lower surfaces in the
centre region to measure the pressure distribution.
• The test section floor and ceilings are equipped with pressure tabs to measure the static
pressure variation in longitudinal and transverse directions.
• The wake rake, downstream of the airfoil, is equipped with total and static pressure tubes to
measure the vertical pressure distributions.
• The load balance system on the turn tables measure airfoil model forces.
• The inflow angle of attack is measured directly from the turn table encoder positions,
knowing the alignment of the airfoil model.

3
• Additional test section flow properties include temperature, free stream flow velocity /
dynamic pressure and atmospheric pressure.

-CP
Airfoil pressure ∆CP
Test section
conditions

-CP
Wake
pressure

Wall pressure Angle of attack

Load balance
-CP

Figure 3 Overview of measured signals in the test section.

2.4 Processing
The measured pressure distributions and the load balance signals can be used to derive the
airfoil force coefficients:
• The airfoil pressure distribution is used to calculate the lift and moment coefficients as well
as the pressure drag coefficient, which is typically used as a good approximation to total
drag at separated flow.
• The floor and ceiling longitudinal pressure distributions are used to calculate the lift
coefficient, which is only valid for attached flow at angles of attack below the maximum lift
coefficient.
• The wake rake total and static pressure distributions are used to calculate the total drag
coefficient, which is valid only when the airfoil flow is attached.
• The load balances on the turn tables are used to calculate airfoil lift, drag and moment
coefficients.

During the design of the measurement setup it had high priority to obtain redundant results. This
explains why each of the airfoil force coefficients can be obtained from at least two different
sources.

2.5 Wall corrections


The closed test section walls influence the flow in the test section causing differences in the
measurements compared to a free stream. Conventional wall corrections according to [7] where
implemented to take into account these errors so that the final corrected results correspond to
2D free air flow.

Solid blockage is due to displacement of the flow around the airfoil from the airfoil contour giving
a local increase in the flow speed around the airfoil. The correction is important for all angles of
attack, especially for thick airfoils. The classical method of Glauert was used.

Wake blockage behind the airfoil due to the airfoil wake leads to a local speed up of the flow
around the airfoil. This correction is important for separated flow but negligible for attached flow.

Streamline curvature from the test section floor and ceiling influences the circulation around the
airfoil causing the actual flow angle of attack to differ from the geometric angle of attack and
also causing changes to the airfoil force coefficients. The method of Allen and Vincenti for a
small model centrally located between floor and ceiling was used.

4
In addition it is necessary to address the influence from traverse system horizontal beam on the
airfoil model. The circulation of the flow over the airfoil model induces circulation around the
horizontal beam of the traverse system. In return the circulation around the beam induces back
a change in the lift force of the airfoil model and a change in the angle of attack [8].

3 Experimental setup
This section explains the experimental setup for the airfoil models as well as for the test section.

3.1 Airfoil models


Two models were manufactured for the experiment: A NACA 64-618 airfoil model with a chord
of 600 mm, called NACA 64-618_600 and a NACA 64-618 model with a chord of 900 mm,
NACA 64-618_900. Both models were made in aluminium and equipped with 64 pressure taps
in the centre region, Furthermore the NACA 64-618_600 model had two additional pressure tap
distributions away from the centre region, each having 32 pressure holes, Figure 4 shows the
pressure tap locations of both models in the centre region (top) and the additional pressure taps
for the NACA 64-618_600 model. Figure 5 shows a top view of the NACA 64-618_600 model.

Top view

Flow
A B C Model

Side wall

Outlet

Figure 4 Pressure tap distribution for NACA 64-618 Figure 5 Locations of


models in the centre region (top) and for the NACA pressure tap distributions
64-618_600 model in the off centre regions (bottom). for NACA 64-618_600.

3.2 Test section


The test section floor and ceiling were instrumented with 32 pressure taps each with a non
uniform spacing for optimum resolution. Because the pressures are only measured in a limited
region around the airfoil, a far field extrapolation is needed as described in [8]. The wake rake
was instrumented with arrays of static and total pressure tubes. It was located with the ends of
the tubes at approximately 800 mm behind the airfoil model trailing edge. A sensitivity study was
initially carried out to find the optimum distance between the airfoil model trailing edge and the
wake rake. This showed only little sensitivity of the wake rake measurements to the actual
location. In addition to the differential pressure measurements, the turn table angle was
measured together with test section flow properties. The free stream dynamic pressure was
taken directly from the LSWT control system, which measures a pressure differential in the
contraction.

3.3 Measurement campaigns


The following measurement campaigns were used:

5

6
The NACA 64-618_600 model clean surface flow at Re = 3x10 corresponding to a free
stream flow speeds of 75 m/s [9].

6 6
NACA 64-618_900 model clean surface flow at Re = 3x10 and 6x10 corresponding to free
stream flow speeds of 50 m/s and 100 m/s respectively [10]

All measurement series contained measurements from both airfoil pressure and wall pressures.
The balance system was not used during this investigation.

4 Results
The results in this section were used to assess the two dimensionality of the flow on the airfoil
model by comparing the pressure distributions on the airfoil model at different span locations.
Also the need for averaging of wake rake results for drag to obtain a representative value was
investigated. The influence of the measurement method on the resulting lift (cl) versus angle of
attack (AOA) was assessed by comparing results from the airfoil pressure distribution with
results from the wall pressure distributions. The wall corrections were verified by comparing
results for the same Reynolds number obtained for different combinations of model size and
flow speed and by comparing different Reynolds numbers. Finally, the results from LSWT were
benchmarked against results from other wind tunnels.

4.1 Two dimensionality of airfoil model flow


The span variation of the flow in the test section was evaluated by comparing pressure
distributions at different span locations together with the resulting cl versus AOA. Figure 6
shows the pressure distributions for AOA = 4 and 8 deg. It can be seen, that except for minor
deviations due to scatter and lack of resolution, the agreement between all three sections is
good. This verifies that the flow is two dimensional when the flow is attached. Figure 7 shows
the resulting cl versus AOA for the different sections. Except for a minor change of slope for
section A, which is nearest to the wall, the deviations start to grow at maximum cl and into the
post stall region. This indicates that the flow is no longer two dimensional in the separated
region. It appears from looking at the different levels of cl in post stall that separation advances
earlier on the centre part of the airfoil model than towards the side wall forming a single stall cell
in the centre region of the airfoil model.

Figure 6 Pressure distributions, A, B and C Figure 7 Airfoil pressure cl for the three
on NACA 64-618_600 for AOA = 4 and 8 pressure distributions on NACA 64-
deg. 618_600.

6
Figure 8 shows the span variation of cd for the NACA 63-618_900 model within the region of the
span where the wake rake was traversed. It can be seen that there is some scatter in cd versus
span position. This is due to variations in time but also due to small imperfections and grains in
the surface. The grey-hatched area indicates the approximate region with pressure taps.
Obviously the presence of the taps causes a measurable additional drag for the present model.
This is not necessarily the case for every model, but often model imperfections lead to a penalty
in cd. Such regions of additional drag are not representative to the clean model and need to be
taken out when processing for the final value of cd.

Top view 0.009

0.008
Flow

0.007

0.006
Model
0.005
Side wall

Wake rake
cd

traverse 0.004
region
0.003

0.002
Spanwise
0.001
coordinate
0
Outlet -400 -300 -200 -100 0 100 200 300 400
Spanwise coordinate [mm]

Figure 8 Variation of wake rake cd versus span at AOA = 4 deg for NACA 64-618_900
model

4.2 Influence of measurement method on resulting lift


Figure 9 shows the pressure distributions on ceiling (red) and floor (green) for NACA 64-
618_600 at AOA 8 and 12 deg respectively. The pressure difference over the airfoil is reflected
on the walls and integration of the wall pressure images including a far field approximation
yields cl [8]. The small scattering of the results is due to minor imperfections of the bushings
with taps in the walls, since it is difficult to keep these absolutely flush with the surface. Most of
this scatter was eliminated by using an empty section offset measurement. The remaining small
differences are negligible in the calculation of cl. The results for the ceiling at AOA = 12 shows a
kink downstream of the airfoil model. This is because the wake rake was parked near to the wall
for high angles of attack to avoid damage. The wake rake influenced the local wall pressure and
these points have to be taken out before calculating cl.

Figure 9 Wall pressure distributions for NACA 64-618_600 at AOA = 8 and 12 deg.

Figure 10 shows the resulting cl from the airfoil pressure distribution and the wall pressure
distributions for NACA 64-618_600. The curves are in excellent agreement for angles of attack
before maximum cl. This verifies the calculation principles for the wall pressures. In the
separated region, a difference appears with higher values for cl from wall pressure compared to

7
cl from airfoil pressure. The pressure image on the walls is an average of the airfoil model
surface pressure. Separation progresses most in the airfoil centre region according to Figure 7
and this causes the resulting cl to come out lower than cl from the wall pressure.

Figure 11 shows similar cl curves for NACA 64-618_900. Here, maximum cl and post stall cl are
in good agreement for both measurement principles but there is a small offset between the two
curves, which is due to the uncertainty on the offset calibration of the wall pressure
measurements..

These results reveal that the measurement method does not have influence on attached flow
results and on maximum cl around 10 deg but it has influence on the resulting cl in post stall.
Therefore, measurements in stall can only be properly assessed when knowing the
measurement method and even then they should be interpret with great care because of the
three dimensional behaviour of the flow over the airfoil model in this regime and the influence of
model aspect ratio.

Figure 10 Wall pressure cl compared with Figure 11 Wall pressure cl compared with
airfoil pressure cl for NACA 64-618_600. airfoil pressure cl for NACA 64-618_900.

4.3 Verification of wall corrections


6
Figure 12 shows polar results for NACA 64-648_600 and 64-618_900 for Re = 3×10 obtained
from airfoil pressure and Figure 13 shows the same polar results based on the wall pressures.
Due to the difference in chord, the free stream flow speed was approximately 75 m/s and 50 m/s
respectively. Wall corrections were applied to both measurements, and the different chord gave
differences in both blockage and stream line curvature. These comparisons can therefore be
used to verify that the applied wall corrections are valid to ensure that the results are universal.

For the NACA 64-618_900 model there was a minor difference in the rise of cd before stall,
which we attributed to the surface quality, which wad some small irregularities that promoted the
on-set of separation to a slightly lower AOA compared to the NACA 64-618_600 model. For the
airfoil pressure results in Figure 12 there is also a difference in the stall pattern for the two
models, where the NACA 64-618_900 model shows a higher value for maximum cl and a small
kink around AOA = 12. The agreement for maximum cl for the two models is better for the wall
pressure results in Figure 13.

The consistency of the cl curve slope and the value of cd for attached flow verify the wall
corrections, whereas the difference in maximum cl from airfoil pressure compared to wall

8
pressure for NACA 64-618_600 due to the stall cell introduces an uncertainty, showing that the
model chord in the present case had influence on the obtained value for maximum cl.

Figure 12 Airfoil pressure cl versus cd and cl versus AOA for NACA 64-618_600 and
NACA 64-618_900 at Re = 3x106.

Figure 13 Wall pressure cl versus cd and cl versus AOA for NACA 64-618_600 and NACA
64-618_900 at Re = 3x106.

4.4 Variation in Reynolds number


6 6
Figure 14 shows polar results for NACA 64-618_900 at Re = 3x10 and 6x10 . For the higher
Reynolds number the value of maximum cl is higher. There is hardly any difference in the
values of minimum cd. This is somewhat surprising since a higher Reynolds number normally
reduces minimum cd and also alters maximum cl. The irregularities on the model possibly
influenced the results.
6 6
The slope of cl for Re = 6x10 is a little steeper than cl for Re = 3x10 . The difference is
marginal but somewhat unexpected. At the high flow speed of 100 m/s which is needed to
6
obtain Re = 6x10 it is possible that the light compressibility introduce uncertainty on both wall
corrections as well as on the dynamic free stream pressure. Looking deeper into this would
require more investigations that were outside of the scope of this work.

9
Figure 14 Airfoil pressure cl versus cd and cl versus AOA for NACA 64-618_900 at Re =
6 6
3x10 and 6x10 .

Figure 15 Benchmark of cl versus cd and cl versus AOA for NACA 64-618 from LSWT,
6
LWK and LTPT at Re = 3x10 .

4.5 Benchmark of results


Figure 15 shows a benchmark of results from LSWT against results from Stuttgart Laminar
Wind Kanal (LWK) [11] and the Langley Low Turbulence Pressure Tunnel (LTPT) [1]. All the cl
curve results were obtained from wall pressure distributions, whereas a wake rake was used to
obtain cd. The results from LWK were obtained for the exact same NACA 64-618_600 model to
eliminate the uncertainty from the model shape accuracy.

The LSWT 600 mm model results agree very well with the results from LWK. The slope of cl as
well as for maximum cl and minimum cd are similar. The LWK results show a slightly higher
maximum cl as well as smaller cd for angles of attack close to maximum cl. Also the results for
the LSWT 900 mm are in good agreement with LWK, however, with a small difference in the cl
curve slope. The results from LTPT differ from the other measurements with a difference in both
cl curve slope and maximum cl. Also cd is slightly lower.

10
5 Conclusions
Two airfoil models of the NACA 64-618 shape having different chords of 600 mm and 900 mm
were tested in the LM Glasfiber LSWT wind tunnel at flow conditions corresponding to Reynolds
numbers at 3x106 and 6x106. The results verified the applied classical wall corrections but also
revealed that 2D airfoil testing is not trivial and more than just applying wall corrections. For
appropriate interpretation of measured results in post stall it is important to know how the
measurements were carried out and how the model aspect ratio affects the results. The
investigation verified the test setup of LSWT where high quality end results for 2D airfoil data
6
could be obtained and that results until Reynolds numbers of 6x10 are fully feasible.

An investigation of the span variation of the 600 mm chord airfoil model pressure showed that
the flow was two dimensional for angles of attack with attached flow, whereas measurements
around maximum lift and in post stall showed three dimensional flow with a single stall cell in the
centre region of the airfoil model. Results from the 900 mm chord model did not show the stall
cell making the presence of this depended on the model aspect ratio.

The span variation of drag was measured to be significant making it necessary to average drag
in both time and space and furthermore to make sure that areas of the model with surface
imperfections or pressure taps should not be present in the measurements.

Results for lift obtained from airfoil and wall pressure respectively were in very good agreement
for attached flow but were slightly different in the post stall making it necessary to know the
measurement method when evaluating wind tunnel results.
6
By comparing results from different model chords at the same Reynolds number, 3x10 , it was
possible to verify the applied wall corrections. The lift curve slopes as well as maximum lift and
minimum cd values were consistent. However, the verification could only partially be
extrapolated to Re = 6 x10 6, where a small difference in lift curve slope was found.

A Benchmark of LSWT results against other wind tunnels put the verification of wall corrections
done in this paper into perspective. The small uncertainties that remained from this study should
be looked at together with the variation in results from different wind tunnels. Clearly, the
uncertainty from comparing results between different wind tunnels is more significant that the
remaining uncertainty from wall corrections.

References
1. Abbot IH, von Doenhoff AE, Stivers LS. Summary of airfoil data. 1945 NACA Rep.824
2. Timmer WA, van Rooij RPJOM. Summary of the Delft University Wind Turbine Dedicated
Airfoils. 2003 J. Solar Energy Engineering Vol. 125, Issue 4, pp. 488-496
3. Fuglsang P, Bak C. Development of the Risø wind turbine airfoils. 2004 J. Wind Energy Vol.
7 Issue 2 (p 145-162)
4. Fuglsang P, Development of the LM low speed wind tunnel. 2004. LM Glasfiber Internal
Report, Version 1.00, 31/08 2006.
5. Timmer WA, van Rooij RPJOM. Some aspects of high angle-of-attack flow on airfoils for
wind turbine application. 2003 Proc. European Wind Energy Conference, Copenhagen, July
2001.
6. Papenfuss HD. Aerodynamic Commissioning of the new Wind tunnel at LM Glasfiber A/S
(Lunderskov) 2006 LM Glasfiber Internal Report, June 2006.
7. Ewald BFR (Editor) Wind Tunnel Wall Correction. 1998 AGARD-AG-336.
8. Fuglsang P, Bove S LSWT Post processing – Function description. 2007 LM Glasfiber
Internal Report, Version 1.30, 24/10 2007.
9. Bove S. NACA 64-618 60 Version 2. 2007 LM Glasfiber Internal Report
10. Bove S. NACA 64-618 90 Version 2. 2007 LM Glasfiber Internal Report
11. Würz W. Vetter C. Wind Tunnel Measurements of the LM NACA 64-618 Airfoil. . 2006
Report, IAG Universität Stuttgart

11

You might also like