Aeroacoustic Optimization of

Wind Turbine Airfoils by

Combining Thermographic and
Acoustic Measurement Data
C. Dollinger, M. Sorg, P. Thiemann; University of Bremen
Bremen Institute for Metrology, Automation and Quality Science (BIMAQ) C. Dollinger

External Article English

Introduction for the acoustic-aerodynamic optimization of rotor blades

for wind turbines are presented. The method combines
The noise emissions of wind turbines lead to extensive thermographic data for turbulence boundary layer analysis
problems for manufacturers as the approval process and with the measurement data of an acoustic camera. Thus,
the development site are influenced. Against this back­ the amplitudes and frequencies of the three-dimensional
ground the industry requires immediate action. Due to acoustic emission field can be associated with the geome­
legal requirements and acceptance problems of wind tur­ try of the rotor blade and the inflow-conditions.
bines in the population, acoustic emissions are a signifi­ Thermographic measurement systems detect the temper­
cant issue for manufacturers. Over the last decade the ature distribution of the rotor blade surface. The surface
average rotor diameter of onshore wind turbines increased temperature is influenced by a variety of parameters and
from approximately 27  m to 82  m. In 2011 a maximum can be correlated to the air flow.
diameter of 126 m was achieved [1]. With increasing rotor
diameter, the noise emission increases as well. It acts as an Test Set-up
aerodynamic brake and significantly reduces the potential
energy yield. Many wind turbines are planned with techni­ The flow tests were conducted at the Deutsche WindGuard
cal compromises favoring the reduction of noise emissions Aeroacoustic Wind Tunnel in Bremerhaven. Laminar air
over power output. The acoustic-aerodynamic optimiza­ flows at speeds of up to 360 km/h and chord-Reynolds
tion of rotor blades could reduce these negative effects of numbers of up to 6 million can be generated in this acous­
the technical compromises and increase the efficiency. tically optimized wind tunnel. The relatively large wind
With current methods, a fast three-dimensional mapping tunnel is designed for acoustic and aerodynamic research
of the noise sources with respect to the flow conditions on and allows testing of large models and full-sized wind tur­
the surface of rotor blades is implemented satisfactorily bine airfoil segments. Different velocities up to 100  m/s
neither in wind tunnels nor for wind turbines. and positive as well as negative angles of attack were
In this article first results of a combined measuring method adjusted to investigate several different operating condi­

DEWI MAGAZIN NO. 43, AUGUST 2013 61

Fig. 1: Geometrical arrangement of the microphone array in the wind tunnel wall
on the suction side. Fig. 2: Lift CL over the angle of attack (α) in simulation
with XFOIL and measured in the wind tunnel.

Fig. 3: Thermographic images aligned to the geometry of the

airfoil with turbulent boundary layer transition on a)
the suction side and b) on the pressure side of the
airfoil.

Fig. 4: Position of the calculated transition over

the angle of attack (α) compared to an
XFOIL-simulation a) on the suction side (Top
Xtr) and b) on the pressure side (Bot Xtr) of
the airfoil.

Fig. 5: Results of an acoustic measurement of

the same airfoil. Airfoil a) untreated and
b) with badly applied serration (sawtooth
geometry) at the trailing edge.

62 DEWI MAGAZIN NO. 43, AUGUST 2013

tions of the rotor blade. The wind tunnel is of the closed the temperature of the airfoil-surface. Therefore, an exter­
return type and can be operated either with an open or a nal heating of the flow is not required.
closed test section. The airfoils are mounted vertically Knowing the position of the transition between laminar
between two turntables. The Lift force can be measured flow to turbulent boundary layer can be used in the design
directly through force measurements on the turntables of new airfoils [4] and in flow control of the position of the
and indirectly through pressure measurements both on transition point by airfoil add-ons (e.g. vortex generators)
the wind tunnel sidewalls and on the models surface. A [4,5]. In addition to the influence related to trailing edge
combination of a six-component force balance attached to noise, the location of the transition has a direct effect on
the base turntable, and a two-component force balance the drag of an airfoil. The turbulent boundary layer located
attached to the upper turntable allows a direct measure­ behind the transition is associated with a higher skin fric­
ment of the total wind forces acting on the airfoil. A total tion than the laminar boundary layer in front of the transi­
of 64 independent pressure sensors measure the pressure tion [4].
signature on the test section walls. 32 pressure taps in As no airfoil surface pressure measurements were con­
each of the test section sidewalls allow an indirect mea­ ducted in the tests, only a comparison of the measured
surement of the lift forces produced by the airfoil. and calculated position of transition to an XFOIL-simulation
Furthermore, a wake rake with 52 total head pressure sen­ is possible. Fig. 2 shows the lift of the used airfoil over the
sors and 12 static pressure sensors is available. Different angle of attack. XFOIL [6] is a very fast computer simula­
airfoils with a chord length of 800  mm and a span of tion tool for design and optimization of airfoils. The fast
1250 mm were used. simulation comes at the expense of a less precise predic­
The thermographic measurements were accomplished tion of the aerodynamic characteristic [4]. Compared to
with two different thermographic imaging systems. On the the measurements in the wind tunnel, there is a slight
suction side of the airfoil a 640 x 480 pixel microbolome­ deviation between the simulation results and the mea­
ter-focal-plane-array with a wavelength range from surements (Fig. 2). Similar to the lift results, transition
7.5  to  14  µm and a temperature resolution of at least prediction models in XFOIL may vary to measured posi­
0.03 K is used. The thermographic sensor on the pressure tions of the transition.
side is a 640 x 512 pixel InSb-focal-plane-array. The sensor The thermographic images on the suction and the pres­
is sensitive in the wavelength range from 2 to 5 μm and is sure side are used to calculate the position of the turbu­
cooled by a Stirling cycle. The temperature resolution of lent boundary layer transition by an image processing
the cooled array is below 0.025 K. Both cameras observe algorithm. Based on the geometric arrangement of the
the airfoil through Germanium lenses in the wind tunnel thermographic cameras in the wind tunnel and the geom­
walls. IR radiation with a wavelength between 2 and 14 µm etry of the airfoil, the thermographic data is aligned to the
can pass the lenses. shape of the airfoil (Fig. 3). A subsequent edge detection
An acoustic array with 40 MEMS (micro-electro-mechani­ combined with information of the temperature gradient at
cal system) Microphones was used to measure the noise the point of transition allows determining the chord-wise
emitted by the airfoils. The microphone array is mounted position of the transition for each angle of attack. Fig. 4
in the wind tunnel wall (in flow) on the suction side of the shows measurement and simulation results with a chord-
airfoil in a random geometrical arrangement (Fig. 1). Reynolds number of 4 million and angles of attack between
The overall dimension of the array is 2200 x 1100 mm and -5° and 15°. On the left side the results for the suction side
it is placed in a distance of about 1300 mm to the airfoil. and on the right side the results for the pressure side are
The data acquisition is achieved with a 24 bit multichannel shown.
frontend with a simultaneous sampling rate of 48 kHz. The measured and calculated position of the transition on
the suction side for small angles of attack (in the region of
Thermographic Measurements linear lift) is almost identical to the simulation results.
Near and above CL max there is a significant difference.
The turbulence boundary layer analysis with a thermo­ Thus, the position of the transition in the region near CL max
graphic system is based on the heat exchange by convec­ is overpredicted by the XFOIL-simulation. Due to a limited
tion, conduction and radiation between the airfoil and the field of view, based on the angle of aperture of the ther­
surrounding flow. Energy exchange by conduction and mography camera, the transition for higher angles of
radiation can be disregarded in steady state conditions attack on the pressure side is not detectable. The maxi­
[2,3]. Because of the enhanced exchange of energy (by mum visible position in the field of view is marked with the
convection) between flow and body in the region of turbu­ black stars in Fig. 4b. Apart from the deviating results
lent flow, an increase of the temperature compared to the above 2° angle of attack, the measurement is close to the
region with laminar flow can be detected by a thermo­ simulation. These results are reproducible and repeatable
graphic measuring system [2]. In trans- und supersonic for other airfoil segments.
flow the exchange of energy is sufficient to generate a Beside the detection of the turbulent boundary layer tran­
noticeable temperature gradient. For wind turbine airfoils sition, this thermographic method allows the verification
in subsonic flow either the airfoil or the flow has to be whether the airfoil is free of roughness elements on the
heated externally. It can be distinguished between an surface or the airfoil is dirty. For adequate pressure mea­
active (heated airfoil) and a passive (heated flow) opera­ surements in the wind tunnel, the dirty airfoil has to be
tion mode [3]. The air temperature in the provided wind cleaned.
tunnel is, due to effects of friction in the fans, higher than

DEWI MAGAZIN NO. 43, AUGUST 2013 63

Aeroacoustic Measurements the two measuring methods could be combined, yielding
space-resolved information about flow states and noise
Microphone arrays for the acoustic measurements enable emissions. Thus, the combined measuring method would
a three-dimensional mapping of sound sources in a short enable an acoustic-aerodynamic optimization of rotor
measuring time. Due to the different distances of a sound blades going far beyond the capabilities of current
source and the specific microphones in the array as well as approaches by simulations.
the pressure amplitude affected by the distance, the loca­
tion of the sound source can be determined [7,8]. The Acknowledgements
noise emission on wind turbine airfoils is caused by the The authors would like to thank the Deutsche WindGuard
interaction of upstream atmospheric turbulence with the Engineering GmbH in Bremerhaven for providing the
leading edge of the airfoil and by the interaction between acoustically optimized wind tunnel in the experiments.
a turbulent boundary layer and the trailing edge of the The article is based on results of the project MakOR. The
airfoil (trailing edge noise) [9,10]. Turbulent flow, passing research project is funded by the State of Bremen within
the trailing edge of an airfoil, is one of the main airfoil self- the Applied Environmental Research Program (AUF) and
noise mechanisms responsible for noise [11]. This trailing the European Regional Development Fund EFRE 2007-
edge noise, caused by the interaction between the turbu­ 2013.
lent boundary layer and the trailing edge, can be tonal or
broadband [10].
Fig. 5 shows acoustic measurements on an airfoil evaluat­
ed with a standard beamforming algorithm. Measured
frequencies are between 2000  Hz and 10000  Hz. Fig. 5a
shows results for the untreated airfoil. The expected noise
sources are the vortex generators at 25% chord on the top
and on the bottom of the airfoil. On the right side (Fig. 5b)
the same airfoil with a badly applied serration at the trail­ References
ing edge is shown. The serration causes a vertical noise [1] Fraunhofer Institut für Windenergie und Energiesystemtechnik
source. In normal case serrations are used to reduce trail­ (IWES): Windenergie Report Deutschland 2011.
ing edge noise, but a misalignment of the serration can [2] M. Malerba, M. Argento, A. Salviuolo, G.L. Rossi: A boundary layer
have the opposite effect. inspection on a wing profile through high resolution thermography
The used microphone array, mounted in the wind tunnel and numerical methods. WSEAS Transactions on Fluid Mechanics
wall, is still very sensitive to structure-borne noise and Volume 3, 2008.
flow noise. For the detection of trailing edge noise on [3] U. Desideri, G. Rossi, A. Giovannozzi, P. Maggiorana: High Resolution
unprepared airfoils, an acoustic shielding is required. Infrared Thermography for Airfoils Boundary Layer in Passive Mode.
Proccedings of ASME Turbo Expo 2004.
Conclusions and Future Investigations [4] P. Bæk, P. Fuglsang: Experimental Detection of Transition on Wind
Turbine Airfoils. Proceedings of EWEC European Wind Energy
Thermographic and aeroacoustic measurements on airfoils Conference 2009.
in a wind tunnel provide additional information concern­ [5] N. Součková, J. Kuklová, L. Popelka, M. Matějka: Infrared
ing turbulent boundary transition and trailing edge noise. Thermography as a Visualization Tool of Flow Control by Vortex
The thermographic method for turbulence boundary layer Generators. Colloquium Fluid Dynamics 2010.
analysis can be used to validate XFOIL and CFD simulations [6] M. Drela: Xfoil: An analysis and design system for low reynolds num-
in a wind tunnel test. The position of the transition can be ber airfoils. Low Reynolds Number Aerodynamics. Springer-Verlag,
detected without preparing the airfoil for surface pressure Lecture Notes in Engineering 54, 1989.
measurements. In addition flow control devices as vortex [7] D. E. Berg, J. R. Zayas: Aerodynamic and Aeroacoustic Properties of
generators or zigzag tapes can be checked regarding the Flatback Airfoils. 46th AIAA Aerospace Sciences Meeting and
functionality. Due to the boundary layer on the tunnel Exhibit, 27th ASME Wind Energy Symposium, 2008.
walls, acoustic measurements with a microphone array in [8] S. Oerlemans: Detection of aeroacoustic sound sources on aircraft
a closed test section are a challenging task. Current results and wind turbines. Ph.D thesis, University of Twente, 2009.
show the suitability of the microphone array. Due to [9] M. S. Howe: A review of the theory of trailing edge noise. Journal of
structure-borne noise and flow noise, the measurement of Sound and Vibration Volume 61, 1978.
trailing edge noise is not yet satisfactory. The acoustic [10] S. Oerlemans, P. Sijtsma, B. Méndez López: Location and quantifica-
shielding between wind tunnel wall and microphones is tion of noise sources on a wind turbine. Journal of Sound and
one of the next tasks in the project. Vibration 299, 2007.
Beyond that, future investigations concerning the combi­ [11] T. Geyer, E. Sarradj, C. Fritzsche: Measurement of the noise genera-
nation of both methods are planned. With the develop­ tion at the trailing edge of porous airfoils. Experiments in fluids
ment of appropriate evaluation-algorithms, the results of Volume 48, 2010.

64 DEWI MAGAZIN NO. 43, AUGUST 2013