Content uploaded by Thierry Maeder
Author content
All content in this area was uploaded by Thierry Maeder on Oct 29, 2015
Content may be subject to copyright.
UNCLASSIFIED
Executive summary
UNCLASSIFIED
Nationaal Lucht- en Ruimtevaartlaboratorium
National Aerospace Laboratory NLR
This report is based on an article published in AIAA Journal, Vol. 47, No. 6, June 2009.
Report no.
NLR-TP-2009-401
Author(s)
S. Oerlemans
M. Fisher
T. Maeder
K. Kögler
Report classification
UNCLASSIFIED
Date
August 2009
Knowledge area(s)
Aëro-akoestisch en experimenteel
aërodynamisch onderzoek
Descriptor(s)
wind turbines
microphone arrays
aerodynamic noise
Reduction of wind turbine noise using optimized airfoils and
trailing-edge serrations
Baseline SIROCCO Serration
Problem area
Wind turbine noise is a major
hindrance for the widespread use of
wind energy. For a modern large
wind turbine, aerodynamic noise
from the blades is the dominant
noise source.
Description of work
Acoustic field measurements were
carried out on a 94-m diameter
three-bladed wind turbine with one
standard blade, one blade with
trailing edge serrations, and one
blade with an optimized airfoil
shape. A large horizontal
microphone array was used to
locate and quantify the noise
sources in the rotor plane and on the
individual blades. The acoustic
performance of the different blades
was investigated as a function of
wind speed, observer position, and
blade azimuth.
Results and conclusions
Both modified blades show a
significant trailing-edge noise
reduction at low frequencies, which
is more prominent for the serrated
blade. However, the modified
blades also show tip noise at high
frequencies, which is mainly
radiated during the upward part of
the revolution and is most important
at low wind speeds due to high tip
loading. Nevertheless, average
overall noise reductions of 0.5 dB
and 3.2 dB are obtained for the
optimized blade and the serrated
blade, respectively.
Applicability
This study has demonstrated that
wind turbine noise can be halved
without adverse effects on the
aerodynamic performance.
UNCLASSIFIED
UNCLASSIFIED
Reduction of wind turbine noise using optimized airfoils and trailing-edge
serrations
Nationaal Lucht- en Ruimtevaartlaboratorium, National Aerospace Laboratory NLR
Anthony Fokkerweg 2, 1059 CM Amsterdam,
P.O. Box 90502, 1006 BM Amsterdam, The Netherlands
Tele
p
hone +31 20 511 31 13, Fax +31 20 511 32 10, Web site: www.nlr.nl
Nationaal Lucht- en Ruimtevaartlaboratorium
National Aerospace Laboratory NLR
NLR-TP-2009-401
Reduction of wind turbine noise using optimized
airfoils and trailing-edge serrations
S. Oerlemans, M. Fisher1, T. Maeder2 and K. Kögler1
1 GE Energy
2 GE Global Research
This report is based on an article published in AIAA Journal, Vol. 47, No. 6, June 2009.
The contents of this report may be cited on condition that full credit is given to NLR and the authors.
This publication has been refereed by the Advisory Committee AEROSPACE VEHICLES.
Customer NLR
Contract number ----
Owner NLR
Division NLR Aerospace Vehicles
Distribution Unlimited
Classification of title Unclassified
October 2009
Approved by:
Author
Reviewer Managing department
NLR-TP-2009-401
2
Contents
I Introduction 3
II Experimental Method 5
A Test Setup 5
B Data Acquisition 5
C Test program 6
D Phased-Array Processing 7
E Accuracy of Sources Localization and Quantification 7
III Results and Discussion 8
A Noise Sources in the Rotor Plane 8
B Noise Sources on the Individual Blades 9
IV Conclusions 13
References 13
Reduction of Wind Turbine Noise Using Optimized
Airfoils and Trailing-Edge Serrations
Stefan Oerlemans∗
National Aerospace Laboratory/NLR, 8300 AD Emmeloord, The Netherlands
Murray Fisher†
GE Energy, Greenville, South Carolina 29615
Thierry Maeder‡
GE Global Research, 85748 Munich, Germany
and
Klaus Kögler§
GE Energy, 48499 Salzbergen, Germany
DOI: 10.2514/1.38888
Acoustic field measurements were carried out on a 94-m-diam three-bladed wind turbine with one standard blade,
one blade with trailing-edge serrations, and one blade with an optimized airfoil shape. A large horizontal microphone
array, positioned at a distance of about one rotor diameter from the turbine, was used to locate and quantify the noise
sources in the rotor plane and on the individual blades. The acoustic source maps show that for an observer at the
array position, the dominant source for the baseline blade is trailing-edge noise from the blade outboard region.
Because of convective amplification and directivity, practically all of this noise is produced during the downward
movement of the blade, which causes the typical swishing noise during the passage of the blades. Both modified blades
show a significant trailing-edge noise reduction at low frequencies, which is more prominent for the serrated blade.
However, the modified blades also show tip noise at high frequencies, which is mainly radiated during the upward
part of the revolution and is most important at low wind speeds due to high tip loading. Nevertheless, average overall
noise reductions of 0.5 and 3.2 dB are obtained for the optimized blade and the serrated blade, respectively.
Nomenclature
D= trailing-edge noise directivity function
f= frequency
M= local blade inflow Mach number
N= number of measurements
P= rotor power
St = Strouhal number (f=U)
U= local blade inflow velocity
U10 = wind speed at 10 m height
= misalignment angle between array and wind turbine
= trailing-edge boundary-layer displacement thickness
"= standard deviation of the mean (standard error) (=
N
p)
= angle between blade chord line and source-observer line
= angle between blade flow velocity and source-observer
line
= standard deviation
’= angle between blade plane and plane containing chord
line and observer
= blade azimuth angle
I. Introduction
WIND turbine noise is one of the major issues for the wide-
spread use of wind energy. For a modern large wind turbine,
aerodynamic noise from the blades is generally considered to be the
dominant noise source, provided that mechanical noise is adequately
treated [1]. The sources of aerodynamic noise can be divided into
airfoil self-noise and inflow-turbulence noise. Airfoil self-noise is the
noise produced by the blade in an undisturbed inflow and is caused
by the interaction between the boundary layer and the trailing edge of
the blade. Self-noise can be tonal or broadband in character and may
be caused by several mechanisms, such as turbulent boundary-layer/
trailing-edge interaction noise (subsequently denoted as trailing-
edge noise), laminar boundary-layer vortex-shedding noise, trailing-
edge bluntness noise, or blade tip noise. Inflow-turbulence noise is
caused by the interaction of upstream atmospheric turbulence with
the blade and depends on the atmospheric conditions. It is an open
issue to what extent inflow-turbulence noise contributes to the
overall sound level of a wind turbine [2].
Because of the large number of applications (e.g., wind turbines,
airplanes, helicopters, and fans), the characteristics of airfoil noise
have been investigated extensively in both experimental and theo-
retical studies [3–13]. Both inflow-turbulence and self-noise mech-
anisms were considered and the dependence on parameters such as
flow speed, angle of attack, radiation direction, and airfoil shape
was characterized. These studies formed the basis of several
semi-empirical wind turbine noise prediction models, which were
validated by comparison with field measurements [14–20]. Because
the field results only provided the overall sound level of the turbine,
the relative importance of the different mechanisms was determined
mainly on the basis of the predictions. In some studies, inflow-
turbulence noise was regarded to be the dominant source [11,14–
16,18], whereas others considered trailing-edge noise to be dominant
[17]. In another case, the turbine noise in different frequency ranges
was attributed to mechanical noise, trailing-edge noise, tip noise, and
inflow-turbulence noise [19].
Presented as Paper 2819 at the 14th AIAA/CEAS Aeroacoustics
Conference, Vancouver, Canada, 5–7 May 2008; received 2 June 2008;
revision received 23 February 2009; accepted for publication 24 February
2009. Copyright © 2009 by the National Aerospace Laboratory/NLR.
Published by the American Institute of Aeronautics and Astronautics, Inc.,
with permission. Copies of this paper may be made for personal or internal
use, on condition that the copier pay the $10.00 per-copy fee to the Copyright
Clearance Center, Inc., 222 Rosewood Drive, Danvers, MA 01923; include
the code 0001-1452/09 $10.00 in correspondence with the CCC.
∗Research Engineer, Aeroacoustics Group, P.O. Box 153.
†Research Engineer, Advanced Aero & Acoustics Technologies, 300
Garlington Road.
‡Research Engineer, Alternative Energy and Environmental Technologies,
Freisinger Landstrasse 50.
§Research Engineer, Advanced Aero & Acoustics Technologies,
Holsterfeld 16.
AIAA JOURNAL
Vol. 47, No. 6, June 2009
1470
NLR-TP-2009-401
3
In a few studies, source location measurements were performed to
provide more direct information on the source mechanisms [21–25].
The results from [21–23] were obtained using an acoustic parabola or
a linear array of microphones and focused only on the horizontal
(downward) blade position ( 90 deg). In [24,25], a large two-
dimensional microphone array, positioned on the ground about one
rotor diameter upwind of the turbine, was used to localize the noise
sources in the complete rotor plane and on the individual blades for
two different wind turbines. It was shown that practically all recorded
noise was produced during the downward movement of the blades.
This strongly asymmetric source pattern, which causes the typical
swishing noise during the passage of the blades, was explained by
convective amplification and trailing-edge noise directivity. The
following directivity function for trailing-edge noise was used [26]:
D2sin2=2sin2’
1Mcos 4(1)
where is the angle between the blade chord line and the source-
observer line, ’is the angle between the plane of the blade and the
plane containing the chord line and the observer, is the angle
between the (inverted) local blade inflow velocity and the source-
observer line, and Mis the local blade inflow Mach number (see
Fig. 1 for definition of angles). The numerator in Eq. (1) describes the
directivity of high-frequency trailing-edge noise and indicates that
most of the noise is radiated in the direction of the airfoil leading
edge. It was analytically derived for edge noise from a semi-infinite
flat plate [6,27], but was also found to be valid for finite airfoils [9],
provided that the angle is not too close to 180 deg. In the limit for
low-frequency dipole noise, for which the acoustic wavelength is
much larger than the airfoil chord, the sin2=2term changes into
sin2[5,26]. Nevertheless, in [24], Eq. (1) was found to be valid for
the whole tested frequency range, including the low frequencies, for
which the acoustic wavelength was of the same order as the blade
chord. The denominator in Eq. (1) represents the fourth-power
convective amplification factor for trailing-edge noise [28] and
indicates that the noise source becomes louder when it is moving
toward the observer.
Many studies have addressed the reduction of airfoil or wind
turbine noise. Because inflow-turbulence noise and trailing-edge
noise both scale with approximately the fifth power of the local blade
inflow velocity [5,6,24], an obvious means for noise reduction is to
reduce the rotor rpm or rotor diameter. However, these measures also
reduce the power output of the turbine [1]. The same holds for
increasing the blade pitch angle (i.e., turning the blade leading edge
against the wind): this reduces the local angle of attack and therefore
the noise, but (due to the reduced lift) also the power production.
Thus, the challenge is to achieve a noise reduction without a
reduction in power output. With regard to tip noise, which depends
on the characteristics of the tip vortex, it has been demonstrated in
several studies that the tip shape can have a significant influence on
the noise from a wind turbine [1,6]. The importance of inflow-
turbulence noise depends on the structure of the atmospheric
turbulence and on the shape of the blades. It has been shown both
experimentally and numerically that inflow-turbulence noise levels
increase with increasing sharpness of the airfoil leading edge [10,13].
With regard to trailing-edge noise, a number of reduction concepts
have been investigated. After it had been shown theoretically that the
acoustic radiation efficiency of a trailing edge can be reduced by
serrations [29] (see Fig. 2), this concept was investigated in a number
of experimental studies on 2-D airfoils [30], model wind turbines
[31,32], and a full-scale wind turbine [23]. In [31], serrations were
applied to a 16-m-diam model wind turbine and, depending on the
flow conditions, overall noise reductions of up to 3.5 dB were
obtained. To prevent increased noise at high frequencies, it was
found to be critical to align the plane of the serrations with the
trailing-edge flow. In [23], serrations were applied to a 1 MW wind
turbine, and an overall noise reduction of 2–3 dB was obtained,
despite increased noise at high frequencies. It should be noted that the
measurements in [23,31] only focused on the horizontal (downward)
blade position ( 90 deg), which may give an incomplete picture.
An alternative concept for trailing-edge noise reduction is the
application of flexible trailing-edge brushes. The brushes align
automatically with the trailing-edge flow and have shown significant
noise-reduction potential in wind-tunnel tests on flat plates and on a
2-D airfoil [12,33]. However, a first attempt to apply this concept to a
full-scale wind turbine yielded a reduction of only 0.5 dB [34],
possibly because the improvised brushes were too short. Finally, it
has been shown in calculations and wind-tunnel tests on 2-D airfoils
that trailing-edge noise can be reduced by a modification of the airfoil
shape, without a loss in aerodynamic performance [35]. Note that in
the case of an acoustically optimized airfoil shape, trailing-edge
noise is reduced by changing the structure of the boundary-layer
turbulence, whereas serrations or brushes are meant to affect only the
scattering at the trailing edge. Thus, the effects of an optimized airfoil
shape and brushes or serrations are expected to supplement each
other.
source
observe
r
ϕ
θ
-U
ξ
trailing edge
chord line
Fig. 1 Definition of angles for trailing-edge noise source.
Fig. 2 Climber removing trips from serrated blade.
OERLEMANS ET AL. 1471
NLR-TP-2009-401
4
The present study concerns acoustic field measurements on a
2.3 MW, 94-m-diam wind turbine with one standard (baseline)
blade, one blade with an acoustically optimized airfoil shape, and one
standard blade with trailing-edge serrations. The tests were
performed in the framework of the European SIROCCO (Silent
Rotors by Acoustic Optimisation) project [34]. Building on the
results from earlier wind-tunnel studies on a model rotor [32], the
subject of the project was the design, testing, and full-scale validation
of quiet wind turbine blades, without a loss in power performance. In
an earlier stage of the project, acoustic field measurements on the
baseline turbine [25] indicated that trailing-edge noise from the outer
25% of the blades was the dominant noise source for this turbine and
that the three blades produced practically the same sound pressure
levels: the average overall sound pressure level (OASPL) for the
three blades differed by less than 0.15 dB and, for the two standard
blades that were used again in the present campaign, less than
0.05 dB.
Subsequently, optimized airfoil shapes were designed and as-
sessed through aerodynamic and acoustic wind-tunnel tests on 2-D
airfoils [34–36]. The principle of the airfoil design was to reduce the
dominant low-frequency (less than 1 kHz) trailing-edge noise peak
in the spectrum (which is due to the thick suction-side boundary
layer) by reducing the loading of the suction side, at the expense of
an increased pressure-side loading (which causes a slightly higher
noise level at less important medium frequencies of 1–3 kHz) [35].
The wind-tunnel tests showed 2–3 dB reduction in OASPL
(depending on lift coefficient) [34] and an improved aerodynamic
performance for the newly designed airfoils, even though severe
geometric and aerodynamic constraints had to be considered in the
design (to enable implementation in the existing blade structure).
The new airfoil was then incorporated in the design of the outer part
of the optimized blade (subsequently denoted as the SIROCCO
blade). The twist distribution of the SIROCCO blade was modified
such that the lift distribution was approximately the same as for the
baseline blade. In addition to the new blade design, the third rotor
blade was used to test a second noise-reduction concept, trailing-
edge serrations. The serrated blade had the same nominal geometry
as the baseline blade (including the twist distribution). From power
and loads measurements on the baseline and modified rotors, it was
found that the in-plane and out-of-plane blade loads on the serrated
blade (and, to a lesser extent, also on the SIROCCO blade) were
slightly higher than on the baseline blade, causing the aerodynamic
performance of the modified blades to be similar or slightly better
than the baseline blade [34].
The main goal of the present test campaign was to assess the
acoustic performance of the two modified blades versus the baseline
blade. To assess the effect of blade roughness due to dirt or insects,
the blades were tested in both clean and tripped conditions. A large
horizontal microphone array, positioned at a distance of about one
rotor diameter from the turbine, was used to measure the distribution
of the noise sources in the rotor plane and on the individual blades.
Because the array position was fixed and the wind direction varied,
both up- and downwind measurements were performed. In the
present paper, the acoustic array results are presented and analyzed.
The noise characteristics for the three blades are investigated as a
function of wind speed, rotor azimuth angle, and observer position
(upwind or downwind), for clean and tripped conditions. Section II
describes the test setup, test program, and the array processing
methods. In Sec. III, the results are presented and discussed. The
conclusions of this study are summarized in Sec. IV.
II. Experimental Method
A. Test Setup
The measurements were carried out in March/April 2007 on the
same General Electric 2.3 MW prototype test wind turbine that was
used in the baseline test campaign in 2005. It had a rotor diameter of
94 m, a tower height of 100 m, and was located on the Netherlands
Energy Research Foundation test site in the Wieringermeer (The
Netherlands). The turbine control system adjusted the rpm and blade
pitch angle, depending on the wind speed measured at the nacelle: for
higher wind speeds, the pitch angle was increased, reducing the local
angle of attack and thus the blade loading. The rpm increased up to a
certain wind speed, after which it remained constant. The turbine had
a yaw mechanism that automatically turned the rotor against the
wind.
To compare the blade performance for identical weather and
turbine conditions, the rotor consisted of one standard (baseline)
blade, one standard blade with trailing-edge serrations, and one
SIROCCO blade. The SIROCCO blade was nominally identical to
the baseline blade, except for the outer 30%, where it had a new
airfoil. The serrated blade had the same nominal geometry as the
baseline blade. The aluminum serrations, with a thickness of 2 mm,
were mounted to the outer 12.5 m of the blade on the pressure side.
The 2 mm step was smoothed using filler material over a few
centimeters of chord. Similar to [32], the length of the serrations was
about 20% of the local chord and varied as a function of radius: the
tooth length was about 10 cm at the tip and about 30 cm at the most
inboard position. A picture of the serrated blade is shown in Fig. 2.
Using templates for different radial stations, the plane of the
serrations was aligned with the flow direction at the blade trailing
edge (as determined from flow calculations). This trailing-edge flow
direction is constant in the variable-rpm region of the turbine. By
aligning the serrations with the flow, it was attempted to minimize
their aerodynamic impact and prevent increased high-frequency
noise by flow through the teeth. Before the acoustic measurements,
all blades were cleaned. To assess the effect of blade roughness due to
dirt or insects, the blades were tested with and without trips: in state 1
all blades were tripped and in state 2 all blades were clean. The 2-D
trips, with a thickness of about 0.4 mm and a width of 4 mm, were
installed close to the leading edge on both sides of the blade, from the
very tip to about half the blade span. The (variable) blade pitch angle
was the same for the three blades.
An acoustic array was used to locate and quantify the noise sources
on the rotor and on the individual blades. The acoustic array
consisted of 148 Panasonic WM-61 microphones, mounted on a
horizontal wooden platform of 16 18 m2, which was positioned at
a distance of about one rotor diameter from the turbine (Fig. 3).
Because the array position was fixed and the wind direction varied,
both up- and downwind measurements were performed. The
Panasonic microphones were mounted flush to the surface of the
platform, with the membrane parallel to the platform, and were
equipped with wind screens. As a reference, two calibrated LinearX
M51 microphones equipped with hemisphere wind screens were
mounted on the platform as well. To correct for the view angle of
about 45 deg (Fig. 3), the microphone array had an elliptic shape
(Fig. 4), rather than the conventional round array design. In this way,
the effective array shape, as seen from the rotor, is round, so that the
resolution with which the noise sources in the rotor plane are
localized is approximately the same in the horizontal and vertical
directions. The ellipse was slightly tilted to the right-hand side of the
rotor plane, to obtain maximum resolution on the side where the
blades move downward [for the standard array position (i.e., upwind
of the turbine)] and where maximum noise radiation was observed in
the 2005 campaign. The array had a high microphone density in the
center to ensure low side-lobe levels at high frequencies and had a
low-density outer part to obtain a good resolution at low frequencies
[37]. The distance and orientation of the array with respect to the
turbine were determined using a laser distance meter and a compass.
B. Data Acquisition
Acoustic data from the array microphones were synchronously
measured using the VIPER multichannel data-acquisition system
[38] at a sample frequency of 30.7 kHz and a measurement time of
30 s. The acoustic data were processed using fast Fourier transform
blocks of 1024 samples with a Hanning window and 50% overlap,
yielding 1800 averages and a narrowband frequency resolution of
30 Hz. A second-order 500 Hz high-pass filter (12 dB=octave [38])
was used to suppress high-amplitude pressure fluctuations at low
frequencies and thus to extend the dynamic range of the A/D
converter, so that low-pressure amplitudes at high frequencies were
1472 OERLEMANS ET AL.
NLR-TP-2009-401
5
included. The sound levels were corrected for the filter response and
for pressure doubling due to reflections at the platform. Before the
measurements, the sensitivity at 1 kHz was determined for all
array microphones using a calibrated pistonphone. The frequency
response of the Panasonic microphones was taken from previous
calibration measurements. The frequency response of the M51
microphones was taken from calibration sheets. No corrections were
applied for microphone directivity, because calibration measure-
ments showed that these effects amounted to less than 2 dB up to
20 kHz for angles smaller than 75 deg with respect to the microphone
axis. Phase matching of the microphones was checked using a
calibration source at known positions. A trigger signal from the
turbine (one pulse per revolution) was recorded synchronously with
the acoustic data to determine the location of the blades as a function
of time for the source localization on the individual blades
(Sec. II.D).
In parallel to the acoustic measurements, several parameters from
the turbine and two nearby meteorological masts were continuously
measured at a sample rate of 4 Hz or higher. These data included wind
speed, wind direction, temperature, power production, turbine
orientation (misalignment angle ), rpm, and blade pitch angle.
C. Test Program
During the 4-week test campaign, in total, more than 600
measurements were taken. Following the International Electro-
technical Commission norm for wind turbine noise measurements
[39], it was attempted to obtain measurements at wind speeds (at
10 m height) between 6 and 10 m=s. The wind speed at 10 m height
was calculated by multiplying the average wind speed measured at
the nacelle by 0.70 (according to the standard wind profile from
[39]). Because the array position was fixed and the wind direction
varied, both up- and downwind measurements were performed.
Measurements with a large misalignment angle (see Fig. 3) were
excluded from the analysis, because for very oblique view angles, the
array resolution becomes poor. On the basis of the turbine
operational data, the most stable measurements (i.e., small variation
in wind speed, rpm, pitch angle, and turbine orientation) were
selected for further analysis.
An overview of the selected measurements is given in Table 1.
Because of unpredictable weather conditions, it was not possible to
obtain measurements in each wind-speed bin, and for state 1, only
downwind measurements were done. Because the clean rotor is
considered to be most representative for the rotor during normal
operation, and because the upwind measurements covered all
relevant wind speeds, the focus of this paper will be on state 2a.
The average turbine and weather conditions for the different rotor
states are listed in Table 2 (equal weights per wind-speed bin). As
mentioned in Sec. II.A, the blade pitch angle (not to be confused with
the twist distribution) is zero at low wind speeds and increases for
higher wind speeds. To give an impression of the variation of the
parameters within each state, this table also shows the standard
deviation for each value defined as
1
N1X
N
i1xi
x2
v
u
u
t
Wind
Platform 45°
Tu rb i n e
ψ
Platform
Turbine
α
Wind
Turbine
Platform
Fig. 3 Side view (left), front view (middle), and top view (right) of the test setup. The array microphones were mounted on the platform in an elliptic
shape for optimum resolution.
Table 1 Number of selected measurements per wind-speed bin for each rotor state
6m=s7m=s8m=s9m=s10m=s
State 1 (tripped rotor; array downwind) 8 8 8 0 0
State 2a (clean rotor; array upwind) 7 8 8 8 6
State 2b (clean rotor; array downwind) 8 8 8 0 0
Table 2 Average weather and turbine parameters for each rotor statea
U10,m=sP, MW rpm , deg Blade pitch, deg
State 1 6.9 1.6 (0.1) 14.5 (0.2) 204 (12) 0.0 (0.1)
State 2a 8.1 2.1 (0.1) 14.9 (0.0) 2(4.4) 5.1 (0.6)
State 2b 6.9 1.6 (0.1) 14.6 (0.1) 183 (2.4) 0.0 (0.0)
aThe standard deviation is indicated between parentheses.
-10
-6
-2
2
6
10
-8 -4 0 4 8
Y [m]
X [m]
Fig. 4 Layout of array microphones. The rectangle indicates the
platform dimensions.
OERLEMANS ET AL. 1473
NLR-TP-2009-401
6
with
x1
NX
N
i1
xi
Note that the power, rpm, and blade pitch angle are not randomly
distributed around the mean value, but depend on the wind speed
according to the turbine control system. Therefore, the standard
deviations for these parameters are based on linear curve fits through
the measured data as a function of wind speed. Because the turbine
had an automatic yaw system, the yaw angle (i.e., the difference
between the wind direction and the turbine orientation) was assumed
to be zero.
D. Phased-Array Processing
The microphone array data were processed using two different
methods. With the first (stationary) method, noise sources in the
complete rotor plane were localized using conventional beamform-
ing [40]. Thus, noise from the rotor hub can be separated from blade
noise, and it can be seen where in the rotor plane the blade noise is
produced. The method shows the integrated effect of the three blades,
averaged over the complete measurement time of 30 s (i.e., several
revolutions).
The first step of this processing involves the calculation of an
averaged cross-spectral matrix, which contains the cross powers of
all microphone pairs in the array. To improve the resolution and to
suppress background noise (e.g., wind-induced pressure fluctuations
on the microphones), the main diagonal of the cross-power matrix
(i.e., the auto powers) was discarded. A spatial window was applied
to the microphone signals, which reduced the effective array size
with increasing frequency and which corrected for the variation in
microphone density over the surface of the array [37]. In this way, the
array resolution at low frequencies was improved, and coherence-
loss effects at high frequencies (due to propagation of the sound
through the atmospheric boundary layer) were suppressed.
Acoustic source maps of the rotor plane were produced by
electronically steering the array to a set of grid points and calculating
the noise radiated from each of them. The scan grid, with a diameter
of 140 m and a mesh width of 2 m, was placed in the rotor plane and
rotated in accordance with the orientation of the turbine (depending
on wind direction). The 4 deg tilt angle between the rotor axis and the
horizontal plane was also taken into account. The effect of sound
convection in the atmospheric boundary layer was taken into account
by assuming a constant wind speed between the scan location and
the microphones. This constant wind speed was calculated as the
average wind speed between the rotor hub and the array center, as
determined from the standard wind velocity profile in [39].
The narrowband acoustic source maps were summed to one-third-
octave bands, and the source levels were normalized to a constant
reference distance. The noise sources in the rotor plane were
quantified using a source-power integration method [37]. This
technique sums the source powers in (part of) the measured source
map and corrects the results with a scaling factor obtained by
performing a simulation for a monopole source at the center of the
integration region. The thus-obtained integrated sound pressure level
of the turbine, as measured at the array position, is similar to the
apparent sound power level defined in [39]. All spectra presented in
this paper are in one-third-octave bands. The accuracy of the
integration method is discussed in the next section.
The second processing method employed three rotating scan
planes to localize the (de-Dopplerized) noise sources on the three
individual blades [rotating source identifier (ROSI)] [41]. This
enabled a comparison of the noise from the different blades. The start
position of the scan planes was determined using a trigger signal from
the turbine that was recorded synchronously with the acoustic data.
Acoustic source maps of the different blades were produced by
electronically steering the array to a set of rotating grid points and
calculating the noise radiated from each of them. The three scan grids
were placed in the rotor plane at azimuthal positions corresponding
to the three blades. The blade grids ran from 15 to 60 m in the radial
direction, had a chordwise extent of 30 m, and had a mesh width of
1m.
Similar to the first processing method, the narrowband acoustic
source maps were summed to one-third-octave bands, and the source
levels were normalized to a constant reference distance. The ROSI
results show the noise sources on the individual blades, averaged
over a specified part of the revolution. To distinguish between the
noise production during the down- and upward movements of the
blades, separate ROSI scans were done for blade azimuth angles
from 0 to 180 deg and from 180 to 360 deg (with 0 deg as the upper
vertical blade position). To limit processing time, only the first
rotor revolution after the start of each acoustic measurement was
processed. The noise from the blades was quantified using a power
integration method for moving sound sources [42], which is similar
to the aforementioned integration method for the stationary rotor
plane. The thus-obtained integrated sound levels represent the
contribution of the different blades to the overall sound pressure level
of the turbine, as measured at the array position.
E. Accuracy of Source Localization and Quantification
The relative position and orientation of the acoustic array and the
wind turbine were determined using a laser distance meter and a
compass. Nevertheless, there are a number of uncertainties in the
localization method, which may cause a deviation between the
measured and actual source positions. First, the blades are not located
exactly in the rotor plane: the 4 deg rotor tilt angle is accounted for,
but not the rotor cone angle and the bending of the blades outside the
plane. Second, sound refraction by wind shear and sound convection
by wind gusts are not accounted for; a constant wind speed is
assumed. Third, the rotor rpm is assumed to be constant within
one revolution. Therefore, the accuracy of the source localization
technique was assessed by attaching a whistle to one of the blades for
a short period of time, at a position unknown to the acoustic test team.
After determining to which blade the whistle was mounted, the ROSI
source maps were used to estimate the exact whistle position (Fig. 5).
The thus-obtained source radius was found to deviate only 0.5 m
from the actual radius, which is considered to be accurate enough for
these tests. Figure 5 also illustrates that the scan resolution is
sufficiently high to determine accurate, integrated, blade noise levels.
From numerical simulations [43], it was found that as long as the
distance between adjacent scan points is smaller than the main lobe
width (i.e., the width at 3 dB below the peak level), the integrated
levels are accurate.
The acoustic source maps were quantified using the power
integration method described in the previous section. The accuracy
of this method in terms of absolute sound pressure levels was verified
by comparing the integrated rotor source maps with the measured
sound pressure levels at the array microphones. If all the noise
measured by the array microphones is due to the turbine rotor, these
spectra should coincide. Figure 6 shows the spectra measured by the
array microphones (array) and the reference microphone at the center
of the array (refmic) versus the integrated spectra for the rotor
(powint) and the three blades (ROSI). These spectra were averaged
over all measurements in state 2a, were corrected for pressure
Fig. 5 Acoustic source map for whistle measurement. The black dots
indicate the scan grid and the cross indicates the actual whistle position.
1474 OERLEMANS ET AL.
NLR-TP-2009-401
7
doubling by the array surface, and were normalized to the same
reference distance. It can be seen that the average spectrum of the
array microphones is practically equal to that on the reference
microphone. The small difference at low and high frequencies may
be due to wind noise on the array microphones at the edge of the
platform.
Figure 6 also shows that the integrated spectra for the rotor and the
blades have the same shape as the reference spectrum, but are about
3 dB lower over the whole frequency range. This discrepancy cannot
be explained by the fact that the array method is applied to incoherent
extended sources, because simulations for an incoherent line source
yielded accurate integrated levels [37]. However, in addition to
the aforementioned possible deviation between the rotor plane
and the actual blade position, the observed difference may be
partly explained by certain assumptions and simplifications in the
integration method, such as the use of a single monopole source at the
center of the integration region to determine the scaling factor for the
source powers. For a simulated, realistic, wind-turbine-rotor noise
source distribution, the difference between the actual and integrated
overall rotor noise level was about 1 dB [25]. The difference may also
be partly attributed to coherence loss at the array microphones, due
to propagation of the sound through the turbulent atmospheric
boundary layer. A similar effect has also been observed in open-jet
wind-tunnel tests [37]. However, coherence-loss effects typically
increase with frequency, whereas a more or less constant offset is
found here. Furthermore, coherence-loss effects would be expected
to increase with increasing wind speed, whereas the difference here
between the integrated rotor noise level and the level of the reference
microphone remained constant (within about 0.5 dB) for increasing
wind speed. An alternative explanation for the lower integrated
levels could be that the reference and array microphones pick up
some background noise (e.g., from the wind over the platform),
which is not present in the integrated rotor noise spectra (no
background noise measurements with a stopped rotor were per-
formed in the present test campaign).
The difference between the two integrated spectra may be due to
the fact that the ROSI spectrum is de-Dopplerized and the rotor
spectrum is not. For the down-going blade, de-Dopplerization results
in reduced frequencies and reduced sound levels [due to convective
amplification (see Sec. I)], and conversely for the up-going blade.
Furthermore, the different integration regions (complete rotor versus
outer part of the blades) will result in different scaling factors
(depending on, for example, the side-lobe behavior [37]), which may
lead to differences in the integrated spectra.
For the evaluation of the noise-reduction concepts in the present
study, the accuracy of the relative sound levels (i.e., level differences
between the different blades) was most important. This accuracy was
assessed in the baseline test campaign by comparing the individual
blade noise spectra for two consecutive revolutions: for each blade,
the overall sound pressure level (averaged over all selected
measurements) reproduced within 0.06 dB for the two consecutive
revolutions, and the differences between the blades reproduced
within 0.03 dB. It should be noted that due to the small difference in
the out-of-plane blade loads (see Sec. I), the bending may be different
for the three blades. Because the scan planes for all three blades are
placed in the rotor plane, this may affect the measured noise
differences between the blades. However, because the difference in
bending between the blades can be estimated to be quite small (less
than 0.25 m) on the basis of the load measurements and the spatial
array resolution perpendicular to the rotor plane is limited, the
systematical error in the noise difference can be determined to be less
than about 0.05 dB (from array simulations). This means that the
average level differences between the blades can be considered to be
accurate within 0.1 dB for the given weather conditions, turbine
operation parameters, and misalignment angle.
III. Results and Discussion
In this section, the results of the acoustic array measurements are
presented and discussed. First, the noise source distribution in the
rotor plane is analyzed for the up- and downwind array positions.
Next, the noise sources on the individual blades are investigated, to
assess the performance of the SIROCCO blade and the serrations as a
function of array position, wind speed, and rotor azimuth angle.
Because the clean rotor is considered to be most representative for the
rotor during normal operation, and because the upwind measure-
ments covered all relevant wind speeds, the focus will be on state 2a.
A. Noise Sources in the Rotor Plane
Each 30 s measurement resulted in acoustic source maps, showing
the noise sources in the rotor plane as a function of frequency. To
show the general trends, these maps were averaged over all selected
measurements in the respective rotor state (Figs. 7–9). Thus, these
maps show the average effect of all three blades over all revolutions.
The black circle indicates the 94-m rotor diameter and the X indicates
the center of the rotor plane. The range of the color scale is always
12 dB and the maximum is adjusted for each frequency band and
each rotor state. The purpose of these source maps is to show the
qualitative source characteristics; a quantitative comparison between
the different rotor states will be made in Sec. III.B. The rotation
direction is clockwise; note that the source maps in Figs. 7 and 9 are
mirrored to allow easy comparison with the upwind measurements
(i.e., an observer at the array position would see the rotor turn in the
counterclockwise direction). Similar to the previous results on the
baseline turbine [25], the upwind measurements (Fig. 8) indicate that
for an observer at the array position, most of the noise is produced by
160 315 630 1250 2500 5000
Frequency (Hz)
SPL (dBA)
array
refmic
ROSI
powint
5 dB
Fig. 6 Verification of power integrated method.
Fig. 7 Average stationary source maps for state 1 (tripped rotor,
downwind array position). The range of the color scale is 12 dB and the
maximum is adjusted for each frequency band.
OERLEMANS ET AL. 1475
NLR-TP-2009-401
8
the outer 25% of the blades during their downward movement. This
effect, which causes the typical swishing noise during the passage of
the blades, can be explained by convective amplification and trailing-
edge noise directivity [24], as described in Eq. (1). For the higher
frequencies, minor noise sources can also be observed at the tip of the
up-going blades and at the location where the blades pass the tower.
The nature of this tower source is hard to assess on the basis of the
present data, but it could originate from 1) reflection of blade noise on
the tower, 2) impingement of blade tip vortices on the tower, and/or
3) the upstream influence of the tower on the flowfield around the
blade.
For the downwind measurements (Figs. 7 and 9), nacelle noise
appears to be more pronounced than for the upwind array position. In
these plots, the nacelle source appears offcenter, because it is located
in front of the scan plane, which coincides with the rotor plane. Only
about 2.5 dB of the observed difference between the up- and
downwind nacelle noise levels can be explained by the smaller
distance to the array and the distance between the source and the scan
plane. The remaining difference may be explained by a combination
of two factors: First, mechanical noise generated inside the nacelle is
radiated mainly in the downwind direction, because the ventilation
opening is on the rear side of the nacelle. The relative nacelle noise
level in state 1 is lower than in state 2b, probably because of the
misalignment angle of 204 deg (i.e., a deviation of 24 deg with
respect to 180 deg) in state 1 (see Table 2). Second, on the basis of
the dependence in Eq. (1), the trailing-edge noise from the blades is
expected to be slightly higher on the upwind side than on the
downwind side (due to the wind speed and rotor tilt angle). This was
confirmed by comparison of the blade noise spectra in states 2a and
2b for the same wind-speed bins (see Sec. III.B.5). Despite the
(relatively) increased nacelle noise, the overall turbine noise is still
dominated by the noise from the blades.
Note that in Fig. 7, the source maximum for the down-going blade
has shifted anticlockwise (relative to the state 2 plots), which can be
explained by the convective amplification factor in Eq. (1) for the
average misalignment angle of 204 deg in this rotor state [25]. Also
note that in the downwind source maps the noise source at the tip of
the up-going blades is more prominent than in the upwind maps. In
the next section it will be shown that, in addition to a small directivity
effect, this difference is mainly due to the lower average wind speed
for these measurements (Table 2), which leads to a lower pitch angle
and higher tip loading (see also Sec. II.A).
B. Noise Sources on the Individual Blades
As mentioned in Sec. I, acoustic field measurements on the
baseline turbine [25] showed that the average OASPL for the
two standard blades that were used again in the present campaign
(i.e., the baseline and serrated blade) differed by less than 0.05 dB.
Furthermore, it was argued in Sec. II.E that the average level
differences between the blades, as measured with the microphone
array, are accurate within 0.1 dB. Thus, with the present test setup it is
well possible to assess the acoustic performance of the serrations and
the new airfoil. In this section, the possibilities and limitations of
acoustic measurements with a single microphone are first briefly
discussed. Then the array results are used to study the average blade
noise characteristics and the dependence on rotor azimuth, wind
speed, observer position (upwind versus downwind), and rotor state
(tripped versus clean blades).
1. Single-Microphone Analysis
In this section, the acoustic results of a single microphone
are analyzed to demonstrate the possibilities and limitations of
such measurements and to illustrate subjective onsite observations.
During the field tests, the three different blades could be clearly
distinguished by the difference in swishing noise produced during
Fig. 8 Average stationary source maps for state 2a (clean rotor,
upwind array position). The range of the color scale is 12 dB and the
maximum is adjusted for each frequency band.
Fig. 9 Average stationary source maps for state 2b (clean rotor,
downwind array position). The range of the color scale is 12 dB and the
maximum is adjusted for each frequency band.
-5
-4
-3
-2
-1
0
0 60 120 180 240 300 360
rotor azimuth angl e
∆ψ
∆ψ
∆
(°)
∆
∆
OASPL (dBA)
Baseline
Modi fie d
baseline
blad
e
SIROCCO
blad
e
serrated
blad
e
Fig. 10 Average sound pressure level on central array microphone as a
function of rotor azimuth, for the baseline rotor (2005) and the rotor with
modified blades (state 2a).
1476 OERLEMANS ET AL.
NLR-TP-2009-401
9
the passage of each blade. This can be illustrated by plotting the
OASPL measured on a single microphone (i.e., the reference
microphone at the center of the array) as a function of rotor azimuth
angle (Fig. 10). The overall levels in this figure were summed
between 250 and 800 Hz to focus on the low-frequency noise of the
down-going blade. The result is A-weighted. Furthermore, the levels
were averaged over all measurements and all revolutions in state 2a
(synchronization was done using the trigger signal from the turbine).
As a reference, the result for the test campaign on the baseline turbine
(2005) is shown as well. Note that the results from the baseline rotor
cannot be compared directly with those from the modified rotor, due
to the different meteorological conditions. However, the variation in
noise level between the three blades can be compared for the two test
campaigns. For the baseline rotor, three humps are clearly observed,
representing the swishing noise that is observed when the three
blades pass the 3 o’clock position (see also Fig. 8). The three blades
are practically equally noisy, and the amplitude variation (or swish)
during the passage of the blades is about 2.5 dB. For the modified
rotor, three humps with different amplitudes are observed, which can
be associated with the baseline blade, the SIROCCO blade, and
the serrated blade, respectively. Thus, it can be estimated that the
SIROCCO blade yields a reduction of more than 1 dB, whereas the
serrated blade is about 4 dB quieter than the baseline blade. However,
it should be noted that these values only pertain to the low-frequency
noise radiated from around the 3 o’clock position, whereas the blades
may also produce significant noise during the other part of the
revolution, as will be seen subsequently. Moreover, at each moment,
the single microphone picks up the noise from all three blades, and so
the contribution of each blade cannot be extracted from the single-
microphone results. Thus, although Fig. 10 confirms subjective
onsite observations, a dedicated array processing method (as
described in Sec. II.D) is required to obtain a clear picture of the noise
radiated by each individual blade during the complete revolution.
2. Average Blade Noise Characteristics
The source maps for the individual blades, averaged over one
complete revolution and over all measurements in state 2a, are shown
in Fig. 11. The source maps run from 15 to 60 m in radial direction
and have a chordwise extent of 30 m. The black line indicates the
outer 32 m of the blade (trailing edge on upper side). The range of the
scale is 12 dB and the maximum is adjusted for each frequency band,
so that the colors within one row can be compared directly. First, the
source maps show that the differences in source position for the
three blades are small. The source radial position, defined as the
radius at which the maximum level in the source map occurs, is
shown in Fig. 12 as a function of frequency for the three blades.
Because the mesh size of the scan grid was 1 m (Sec. II.D), these
source radial positions are multiples of 1 m. For all blades, the source
radial position increases with frequency, which can be understood
using the relation St f=U const. for the trailing-edge
noise peak [6,33]; for increasing radius, the local blade inflow
velocity Uincreases and the boundary-layer displacement thickness
decreases, so that the produced frequencies are higher [24,25].
Except for the lowest frequencies, for which the modified blades
have a lower source radius, the differences in average source radial
positions between the different blades are small. More important, the
source maps in Fig. 11 show that for low frequencies, both modified
blades are significantly quieter than the baseline blade, especially the
serrated blade. For high frequencies, however, both modified blades
are noisier than the baseline blade, especially the SIROCCO blade.
These trends are illustrated in Fig. 13, which shows the average
integrated spectra for the three blades. These spectra were obtained
by averaging the integrated spectra for all measurements in state 2a,
with equal weights per wind-speed bin. As mentioned in Sec. II.D,
these sound levels represent the contribution of the different blades to
the overall sound pressure level of the turbine, as measured at the
array position. The integrated spectra confirm the low-frequency
noise reduction and high-frequency noise increase for the modified
blades. For the serrated blade, the A-weighted sound pressure level at
high frequencies is even higher than at low frequencies. The reasons
Fig. 11 Average rotating source maps for individual blades in state 2a,
as a function of frequency. The range of the color scale is 12 dB and the
maximum is the same within each row.
32
36
40
44
48
160
315
630
1250
2500
5000
Frequency (Hz)
radius_max (m)
Baseline
Sirocco
Serration
Fig. 12 Average source radial position as a function of frequency for
the three blades in state 2a.
OERLEMANS ET AL. 1477
NLR-TP-2009-401
10
for the increased noise level of the modified blades at high
frequencies will be discussed in subsequent sections. Based on these
average spectra, for the upwind measurements on the clean rotor (i.e.,
state 2a), which are considered most representative for normal
operation and which cover all relevant wind speeds, average overall
noise reductions of 0.5 and 3.2 dB were obtained for the SIROCCO
and serrated blades, respectively. For the other two rotor states,
which covered only the lower wind speeds, average noise reductions
of 0.0 and 1.6 dB (state 1) and 0.2 and 1.2 dB (state 2b) were found
for the SIROCCO and serrated blades, respectively. The reasons for
the lower noise reduction in these states will be discussed in
subsequent sections. The measurement uncertainty of the afore-
mentioned average noise reductions will be discussed subsequently
in Sec. III.B.4.
3. Dependence of Blade Noise on Rotor Azimuth
To better understand the acoustic behavior of the modified blades,
it is interesting to look at the dependence of the blade noise on the
rotor azimuth angle (Fig. 14). This figure shows the overall source
maps for the different blades for 12 rotor azimuth intervals of 30 deg,
starting at 0 deg (12 o’clock). These overall source maps
(averaged over all measurements in state 2a) were obtained by
summing the source maps between 160 Hz and 5 kHz, after applying
A-weighting and a correction for array resolution (which depends on
frequency and blade position) to the levels. Thus, the OASPL of the
blade is equal to the sum of all scan levels in a given source map.
Similar to the previous blade source maps, the black line indicates the
outer 32 m of the blade. The range of the color scale is 12 dB, and the
maximum is the same for all maps, so that the colors can be compared
directly. This figure shows that during the downward movement of
the blades, both modified blades are substantially quieter than the
baseline blade. However, during the upward movement, both
modified blades are noisier. These observations are illustrated in
Fig. 15, which shows the integrated blade spectra for the down- and
upward halves of the revolution separately. The high-frequency
noise increase for the modified blades occurs mainly during the
upward part of the revolution, and for the serrated blade, the
increased noise of the up-going blade even dominates the overall
average spectrum.
4. Dependence of Blade Noise on Wind Speed
In addition to the rotor azimuth, the blade noise characteristics
were also found to depend strongly on the wind speed. This is
illustrated in Figs. 16 and 17, which show the integrated blade noise
spectra for the 7 and 10 m=sbins of state 2a, respectively (the state 2b
spectra in Fig. 16 will be discussed subsequently in Sec. III.B.5). The
7m=sblade spectra show a high-frequency hump around 1250–
1600 Hz, which is absent in the 10 m=sspectra. As seen in the
previous section, this high-frequency noise is mainly produced
during the upward movement and originates from the very tip of the
blade (Figs. 11–14). Because the average blade pitch angle was 0 deg
in the 7m=sbin and higher in the 10 m=sbin, this suggests that the
high-frequency noise at low wind speeds can be associated with the
increased tip loading as a result of the lower pitch angle (see Sec. II.A
for turbine operation details). This also explains the fact that the up-
going blades were noisier in states 1 and 2b than in state 2a (Figs. 7–
9), because the average blade pitch angle was lower than in state 2a
(Table 2). Apparently, this tip noise does not follow the trailing-edge
noise directivity function described in Eq. (1), because it is mainly
radiated during the upward movement of the blades. Figure 16 also
shows that the tip noise is much more prominent for the modified
blades than for the baseline blade, which is always dominated by
trailing-edge noise. Thus, the adapted pressure distribution on the
modified blades, possibly in combination with the slightly increased
70
75
80
85
90
95
160
315
630
1250
2500
5000
Frequency (Hz)
SPL (dBA)
Baseline
Sirocco
Serration
5 dB
Fig. 13 Average blade noise spectra for state 2a.
Fig. 14 Average overall source maps for individual blades in state 2a,
as a function of rotor azimuth. The range of the color scale is 12 dB and
the maximum is the same for all maps.
1478 OERLEMANS ET AL.
NLR-TP-2009-401
11
blade loading (Sec. I), changes the tip vortex characteristics such that
tip noise increases.
For the baseline and SIROCCO blades, the frequency of the low-
frequency trailing-edge noise peak is higher at 10 m=sthan at 7m=s,
which can be explained by the lower blade loading (due to the higher
pitch angle), which leads to a thinner suction-side boundary layer.
This is confirmed in Fig. 18, which shows the source radial positions
for both blades in these two wind-speed bins: for a given radius, the
trailing-edge boundary-layer thickness decreases for the higher wind
speed, and so the produced frequencies are higher. For both bins, the
source radial position of the SIROCCO blade is slightly lower than
for the baseline blade, which is consistent with the higher trailing-
edge noise-peak frequency and which suggests that the main
objective of the airfoil design [i.e., to obtain a thinner suction-side
boundary layer (see Sec. I)] has succeeded. If we suppress the
spectral influence of tip noise by considering only the downward part
of the revolution (Fig. 15), a slight trailing-edge noise increase is
observed between 1 and 3 kHz for the SIROCCO blade, which,
similar to the wind-tunnel results, can be attributed to the increased
pressure-side boundary-layer thickness [35]. However, even if we
consider only the downward part of the revolution, the average
reduction in OASPL for the SIROCCO blade in state 2a is only
1.0 dB, which is lower than the 2–3 dB found in the wind-tunnel tests.
The reasons for this discrepancy between wind-tunnel and field
results are not fully clear yet. Apart from blade quality, a possible
explanation could be that instationary inflow conditions in the field
lead to lift fluctuations well beyond the prescribed design lift range
[34].
In terms of A-weighted overall sound pressure levels (summed
between 160 Hz and 5 kHz), both modified blades were found to
reach maximum noise levels at a wind speed U10 of about 7m=s,
where the tip loading and therefore the tip noise are highest (the blade
pitch angle only starts to increase significantly for wind speeds
higher than 7m=s). The noise from the baseline blade, which is
dominated by trailing-edge noise, also peaks around 7m=s. The
corresponding noise reductions (Fig. 19) are lowest around 7m=s
and increase for higher wind speeds, for both the serrated and
SIROCCO blades. Thus, the results indicate that at high wind speeds,
the noise from the three blades is dominated by trailing-edge noise,
which is effectively reduced by the new airfoil shape and the
serrations. However, at lower wind speeds (increased tip loading due
to lower pitch angle), significant high-frequency tip noise is
generated by both modified blades during their upward movement,
which partly cancels the trailing-edge noise reduction. As a result, the
average noise reductions obtained in states 1 and 2b, which had lower
wind speeds and therefore lower average blade pitch angles
(Table 2), were lower than the 3.2 and 0.5 dB found in state 2a for the
serrated and SIROCCO blades, respectively.
To evaluate these average noise reductions, it is important to
understand the uncertainty levels associated with the measurement
method employed during this study. As argued in Sec. II.E, for the
weather conditions and turbine operation parameters in state 2a, the
measurement uncertainty of the average noise reductions is smaller
72
77
82
87
92
97
160
315
630
1250
2500
5000
Frequency (Hz)
SPL (dBA)
Baseline down
Sirocco down
Serration down
Baseline up
Sirocco up
Serration up
5 dB
Fig. 15 Average blade noise spectra for the upward and downward
part of the revolution in state 2a.
75
80
85
90
95
160
315
630
1250
2500
5000
Frequency (Hz)
SPL (dBA)
Baseline s2a
Sirocco s2a
Serration s2a
Baseline s2b
Sirocco s2b
Serration s2b
5 dB
Fig. 16 Average blade noise spectra for the 7 m=swind-speed bin of
state 2a and state 2b.
70
75
80
85
90
95
160
315
630
1250
2500
5000
Frequency (Hz)
SPL (dBA)
Baseline
Sirocco
Serration
5 dB
Fig. 17 Average blade noise spectra for the 10 m=swind-speed bin of
state 2a.
32
36
40
44
48
160
315
630
1250
2500
5000
Frequency (Hz)
radius_max (m)
Baseline_U7
Sirocco_U7
Baseline_U10
Sirocco_U10
Fig. 18 Source radial position as a function of frequency for the
baseline and SIROCCO blades in the 7 and 10 m=swind-speed bins of
state 2a.
-1
0
1
2
3
4
5
6
567891011
U10 (m/s)
∆
∆
∆
∆
OASPL (dBA)
Sirocco
Serration
Fig. 19 Overall blade noise reduction as a function of wind speed for
state 2a. The solid lines are third-order least-squares fits through the
measured data.
OERLEMANS ET AL. 1479
NLR-TP-2009-401
12
than 0.1 dB. For different turbine and meteorological conditions
(within the tested range), the uncertainty in the noise reduction can be
assessed from the scatter in Fig. 19: for the serrated and SIROCCO
blades, standard deviations of 0.6 and 0.4 dB were found, which
leads to standard deviations of the mean "for the average noise
reductions of 0.10 and 0.06 dB, respectively. Because these "values
are smaller than the average noise reductions of 3.2 and 0.5 dB, the
average reductions are significant for the tested range of turbine and
meteorological conditions.
5. Upwind Versus Downwind Measurements
To assess the effect of observer location (i.e., upwind versus
downwind array position), Fig. 16 compares the blade noise spectra
for the 7m=sbin in states 2a and 2b. The meteorological and turbine
parameters were similar for both cases. It can be seen that the level
of the low-frequency trailing-edge noise hump is lower on the
downwind side for all three blades. A partial explanation for this
could be that the downwind integrated blade noise levels suffer from
increased coherence loss (see also Sec. II.E), because the blade noise
propagates through the rotor wake. Indeed, the difference between
the integrated rotor noise level and the level of the reference
microphone was about 1 dB higher for the downwind measurements
than for the upwind measurements in the 7m=sbin. In addition, the
reduced downwind trailing-edge noise levels may be explained by
the dependence in Eq. (1), due to the wind speed and rotor tilt angle
(see also Sec. III.A). In contrast to the low-frequency noise, the high-
frequency tip noise peak has approximately the same level for the up-
and downwind measurements, which means that the relative
importance of tip noise is higher on the downwind side. In addition to
the tip loading effect discussed in the previous section, this increased
importance of tip noise on the downwind side is an additional
explanation for the lower noise reductions obtained in states 1 and 2b
(as compared with state 2a) and for the fact that the up-going blades
are noisier in states 1 and 2b than in state 2a (Figs. 7–9).
6. Baseline Blade Noise for Different Rotor States
Figure 20 shows the baseline blade noise spectra for the three
different rotor states. With regard to the clean rotor, a number of
differences can be observed between states 2a and 2b: First, the low-
frequency trailing-edge noise peak for state 2a has a higher frequency
and higher level than state 2b. This can be explained, respectively, by
the higher wind speed (i.e., lower blade loading and thinner boundary
layer) in state 2a and the difference in directivity for the up- and
downwind observer positions (see Secs. III.B.4 and III.B.5). Second,
state 2b spectrum shows a high-frequency tip noise peak, which is
almost absent in state 2a. As mentioned before, this can be explained
by the lower average wind speed (and thus higher tip loading) in
state 2b (Sec. III.B.4) and by the increased importance of tip noise on
the downwind side (Sec. III.B.5). Thus, to assess the influence of
tripping on blade noise, state 1 should be compared with state 2b. The
average conditions for these two states are practically the same,
except for the misalignment angle (Table 2). Figure 20 shows that the
low-frequency trailing-edge noise peak for state 1 has a higher level
and lower frequency than in state 2b. This suggests that the trip has
increased the trailing-edge boundary-layer thickness. Furthermore,
the high-frequency tip noise peak is slightly lower for the tripped
case. Thus, the results do indicate a small effect of tripping (0.6 dB
increase in OASPL for the present conditions), but for a conclusive
answer, measurements should be done with a clean blade and a
tripped blade on one rotor.
IV. Conclusions
Acoustic field measurements were carried out on a 94-m-diam
wind turbine, with one standard blade, one blade with an optimized
airfoil shape, and one standard blade with trailing-edge serrations.
The blade modifications had no adverse effect on their aerodynamic
performance. To assess the effect of blade roughness due to dirt or
insects, the blades were tested in both clean and tripped conditions. A
large horizontal microphone array, positioned at a distance of about
one rotor diameter from the turbine, was used to locate and quantify
the noise sources in the rotor plane and on the individual blades.
Because the array position was fixed and the wind direction varied,
both up- and downwind measurements were performed.
The acoustic source maps for the baseline blade showed that for
an observer at the array position, the dominant source was trailing-
edge noise from the outer 25% of the blade. Because of convective
amplification and directivity, practically all noise was produced
during the downward movement of the blade, which caused the
typical swishing noise during the passage of the blades. Both
modified blades showed a significant trailing-edge noise reduction
at low frequencies, which was more prominent for the serrated
blade. However, the modified blades also showed a noise increase at
high frequencies, which can be associated with the blade tips. This
high-frequency tip noise was mainly radiated during the upward
part of the revolution and was most important at low wind speeds
(i.e., high tip loading) and for the downwind array position.
Nevertheless, for the upwind measurements on the clean rotor,
which were considered to be most representative for normal
operation and covered all relevant wind speeds, average overall
noise reductions of 0.5 and 3.2 dB were obtained for the optimized
blade and the serrated blade, respectively. For both blades, the noise
reduction increased with increasing wind speed. The downwind
measurements on the clean and tripped rotors only covered the
lower wind speeds and showed less noise reduction. Comparison of
the noise from the baseline blade for clean and tripped conditions
suggested a noise increase of 0.6 dB due to tripping.
Acknowledgments
Financial support for this research was given in part by the
European Commission’s Fifth Framework Programme, project
reference SIROCCO (Silent Rotors by Acoustic Optimisation)
(ENK5-CT-2002-00702). Financial support was also given by
the Netherlands Organisation for Energy and the Environment
(NOVEM). The authors would like to thank the colleagues from the
University of Stuttgart and from the Netherlands Energy Research
Foundation (ECN) for their valuable contributions to the definition of
the tests and the interpretation of the results.
References
[1] Wagner, S., Bareiss, R., and Guidati, G., Wind Turbine Noise,
Springer–Verlag, New York, 1996.
[2] Guidati, G., Ostertag, J., and Wagner, S., “Prediction and Reduction of
Wind Turbine Noise: An Overview of Research Activities in Europe,”
AIAA Paper 2000-0042, 2000.
[3] Amiet, R., “Acoustic Radiation from an Airfoil in a Turbulent Stream,”
Journal of Sound and Vibration, Vol. 41, No. 4, 1975, pp. 407–420.
doi:10.1016/S0022-460X(75)80105-2
[4] Howe, M. S., “A Review of the Theory of Trailing Edge Noise,”
Journal of Sound and Vibration, Vol. 61, No. 3, 1978, pp. 437–465.
doi:10.1016/0022-460X(78)90391-7
[5] Blake, W. K., Mechanics of Flow-Induced Sound and Vibration,
Academic Press, New York, 1986.
[6] Brooks, T. F., Pope, D. S., and Marcolini, M. A., “Airfoil Self-Noise
and Prediction,”NASA Ref. Publ. 1218, 1989.
70
75
80
85
90
95
160
315
630
1250
2500
5000
Frequency (Hz)
SPL (dBA)
State 1
State 2a
State 2b
5 dB
Fig. 20 Average baseline blade noise spectra for different rotor states.
1480 OERLEMANS ET AL.
NLR-TP-2009-401
13
[7] Dassen, T., Parchen, R., Guidati, G., Wagner, S., Kang, S., and Khodak,
A. E., “Comparison of Measured and Predicted Airfoil Self-Noise with
Application to Wind Turbine Noise Reduction,”Proceedings of the
European Wind Energy Conference, Irish Wind Energy Association,
Slane, Ireland, Oct. 1997, pp. 422–428.
[8] Guidati, G., Bareiss, R., Wagner, S., Dassen, T., and Parchen, R.,
“Simulation and Measurement of Inflow-Turbulence Noise on
Airfoils,”AIAA Paper 97-1698, 1997.
[9] Hutcheson, F. V., and Brooks, T. F., “Effects of Angle of Attack and
Velocity on Trailing Edge Noise,”AIAA Paper 2004-1031, 2004.
[10] Oerlemans, S., and Migliore, P., “Aeroacoustic Wind Tunnel Tests of
Wind Turbine Airfoils,”AIAA Paper 2004-3042, 2004.
[11] Moreau, S., and Roger, M., “Competing Broadband Noise Mechanisms
in Low Speed Axial Fans,”AIAA Paper 2004-3039, 2004.
[12] Herr, M., and Dobrzynski, W., “Experimental Investigations in Low
Noise Trailing Edge Design,”AIAA Paper 2004-2804, 2004.
[13] Moriarty, P. J., Guidati, G., and Migliore, P., “Prediction of Turbulent
Inflow and Trailing-Edge Noise for Wind Turbines,”AIAA Paper 2005-
2881, 2005.
[14] Grosveld, F. W., “Prediction of Broadband Noise from Large
Horizontal Axis Wind Turbine Generators,”AIAA Paper 84-2357,
1984.
[15] Glegg, S. A. L., Baxter, S. M., and Glendinning, A. G., “The Prediction
of Broadband Noise from Wind Turbines,”Journal of Sound and
Vibration, Vol. 118, No. 2, 1987, pp. 217–239.
doi:10.1016/0022-460X(87)90522-0
[16] Hubbard, H. H., and Shepherd, K. P., “Aeroacoustics of Large Wind
Turbines,”Journal of the Acoustical Society of America, Vol. 89, No. 6,
1991, pp. 2495–2508.
doi:10.1121/1.401021
[17] Lowson, M. V., “Theory and Experiment for Wind Turbine Noise,”
AIAA Paper 94-0119, 1994.
[18] Fuglsang, P., and Madsen, H. A., “Implementation and Verification of
an Aeroacoustic Noise Prediction Model for Wind Turbines,”Risø
National Lab., Risø-R-867(EN), Roskilde, Denmark, 1996.
[19] Lowson, M. V., Lowson, J. V., and Bullmore, A. J., “Wind Turbine
Noise: Analysis of Results from a New Measurement Technique,”
AIAA Paper 98-0037, 1998.
[20] Moriarty, P., and Migliore, P., “Semi-Empirical Aeroacoustic Noise
Prediction Code for Wind Turbines,”National Renewable Energy Lab.,
Rept. TP-500-34478, Golden, CO, 2003.
[21] de Bruijn, A., Stam, W. J., and de Wolf, W. B., “Determination of the
Acoustic Source Power Levels of Wind Turbines,”Proceedings of the
European Wind Energy Conference, H. S. Stephens and Associates,
Bedford, England, U.K., Oct. 1984, pp. 889–894.
[22] van der Borg, N. J. C. M., and Vink, P. W., “Acoustic Noise Production
of Wind Turbines in Practice,”Proceedings of the European Wind
Energy Conference, Hellenic Wind Energy Association and European
Wind Energy Association, Oct. 1994, pp. 148–155.
[23] Hagg, F., van Kuik, G. A. M., Parchen, R., and van der Borg, N. J. C. M.,
“Noise Reduction on a 1 MW Size Wind Turbine with a serrated
Trailing Edge,”Proceedings of the European Wind Energy
Conference, Irish Wind Energy Association, Slane, Ireland,
Oct. 1997, pp. 165–168.
[24] Oerlemans, S., Sijtsma, P., and Méndez López, B., “Location and
Quantification of Noise Sources on a Wind Turbine,”Journal of Sound
and Vibration, Vol. 299, Nos. 4–5, 2007, pp. 869–883.
doi:10.1016/j.jsv.2006.07.032
[25] Oerlemans, S., and Schepers, J. G., “Prediction of Wind Turbine Noise
and Comparison with Experiment,”Proceedings of the Second
International Meeting on Wind Turbine Noise [CD-ROM], Inst. of
Noise Control Engineering, Merseyside, England, U.K., Sept. 2007.
[26] Brooks, T. F., and Burley, C. L., “Rotor Broadband Noise Prediction
with Comparison with Model Data,”AIAA Paper 2001-2210, 2001.
[27] Schlinker, R. H., and Amiet, R. K., “Helicopter Rotor Trailing Edge
Noise,”NASA CR-3470, 1981.
[28] Dowling, A. P., and Ffowcs Williams, J. E., Sound and Sources of
Sound, Ellis Horwood, London, 1983.
[29] Howe, M. S., “Noise Produced by a Sawtooth Trailing Edge,”Journal
of the Acoustical Society of America, Vol. 1, No. 1, 1991, pp. 482–487.
doi:10.1121/1.401273
[30] Dassen, T., Parchen, R., Bruggeman, J., and Hagg, F., “Results of a
Wind Tunnel Study on the Reduction of Airfoil Self-Noise by the
Application of Serrated Blade Trailing Edges,”Proceedings of the
European Wind Energy Conference, H. S. Stephens and Associates,
Bedford, England, U.K., 1996, pp. 800–803.
[31] Braun, K. A., van der Borg, N., Dassen, A., Doorenspleet, F., Gordner,
A., Ocker, J., and Parchen, R., “Serrated Trailing Edge Noise
(STENO),”Proceedings of the European Wind Energy Conference,
James and James Science, London, Oct. 1999, pp. 180–183.
[32] Oerlemans, S., Schepers, J. G., Guidati, G., and Wagner, S.,
“Experimental Demonstration of Wind Turbine Noise Reduction
Through Optimized Airfoil Shape and Trailing-Edge Serrations,”
Proceedings of the European Wind Energy Conference, WIP-
Renewable Energies, Munich, July 2001, pp 351–354.
[33] Herr, M., “Design Criteria for Low-Noise Trailing Edges,”AIAA
Paper 2007-3470, 2007.
[34] Schepers, J. G., Curvers, A., Oerlemans, S., Braun, K., Lutz, T., Herrig,
A., Wuerz, W., Matesanz, A., Garcillán, L., Fisher, M., Koegler, K., and
Maeder, T., “Sirocco: Silent Rotors by Acoustic Optimisation,”
Proceedings of the Second International Meeting on Wind Turbine
Noise [CD-ROM], Inst. of Noise Control Engineering, Merseyside,
England, U.K., Sept. 2007.
[35] Lutz, T., Herrig, A., Würz, W., Kamruzzaman, M., and Krämer, E.,
“Design and Wind Tunnel Verification of Low Noise Airfoils for Wind
Turbines,”AIAA Journal, Vol. 45, No. 4, 2007, pp. 779–792.
doi:10.2514/1.27658
[36] Herrig, A., Würz, W., Lutz, T., and Krämer, E.: “Trailing-Edge Noise
Measurements Using a Hot-Wire Based Coherent Particle Velocity
Method,”Proceedings of the 24th AIAA Applied Aerodynamics
Conference, San Francisco, AIAA Paper 2006-3876, June 2006.
[37] Oerlemans, S., Broersma, L., and Sijtsma, P., “Quantification of
Airframe Noise Using Microphone Arrays in Open and Closed Wind
Tunnels,”International Journal of Aeroacoustics, Vol. 6, No. 4, 2007,
pp. 309–333.
doi:10.1260/147547207783359440
[38] Holthusen, H., and Smit, H., “A New Data-Acquisition System for
Microphone Array Measurements in Wind Tunnels,”AIAA
Paper 2001-2169, 2001.
[39] Wind Turbine Generator Systems—Acoustic Noise Measurement
Techniques, International Electrotechnical Commission Norm 61400-
11, Geneva, 2002.
[40] Johnson, D. H., and Dudgeon, D. E., Array Signal Processing,
Prentice–Hall, Upper Saddle River, NJ, 1993.
[41] Sijtsma, P., Oerlemans, S., and Holthusen, H., “Location of Rotating
Sources by Phased Array Measurements,”AIAA Paper 2001-2167,
2001.
[42] Sijtsma, P., and Stoker, R. W., “Determination of Absolute
Contributions of Aircraft Noise Components Using Fly-Over Array
Measurements,”AIAA Paper 2004-2958, 2004.
[43] Broersma, L., “Acoustic Array Measurements—Airframe Noise and
Wind Turbine Noise,”M.S. Thesis, University of Twente, Enschede,
The Netherlands, Jan. 2006.
J. Wei
Associate Editor
OERLEMANS ET AL. 1481
NLR-TP-2009-401
14
