AIAA 2016-3021

Aeroacoustics Conferences
30 May - 1 June, 2016, Lyon, France
22nd AIAA/CEAS Aeroacoustics Conference

Flow topology and noise emission around straight,

serrated and slitted trailing edges using the Lattice
Boltzmann methodology

W.C.P. van der Velden∗, A.H. van Zuijlen† and D. Ragni‡,

Delft University of Technology, Kluyverweg 2, 2629 HT, Delft, the Netherlands

The current study analyzes the flow topology and acoustic emission around straight,
Downloaded by PURDUE UNIVERSITY on June 6, 2016 | http://arc.aiaa.org | DOI: 10.2514/6.2016-3021

sawtooth serrated and slitted-sawtooth serrated trailing edges using a NACA0018 airfoil at
zero degree angle of attack. By using this specific setup, pressure differences between both
airfoil sides are avoided so that the focus will purely lie on the mean and fluctuating velocity.
The flow field is analyzed by evaluating the fully explicit, transient, compressible Lattice
Boltzmann equation. Acoustic perturbations are directly obtained from the computational
measurement domain, as well as by means of a Ffowcs Williams and Hawking and Curle
integral solution. Noise reductions up to 6 dB are achieved for the sawtooth serrated edge
in comparison with the straight trailing edge case, while the slitted-sawtooth edge reaches
a maximum of only 5 dB. At low frequencies, the solid-sawtooths outperform the slitted-
sawtooths. The general trend in term of velocity results shows favorable streamwise, wall-
normal and spanwise quantities for the slitted-sawtooth design, while the solid-sawtooth
design shows less fluctuations in both streamwise, wall-normal and spanwise direction.
Especially the increase of local, wall-normal fluctuations over the slitted-sawtooth serrated
edge, enchances the generation of acoustic pressure waves, making it a less effective trailing
edge noise-suppression add-on.

Nomenclature
α Observer
p angle
β 1 − M02
δ Boundary layer thickness
δ? Boundary layer displacement thickness
θ Boundary layer momentum thickness
κ von Karman constant
ν Viscosity
Φ Acoustic pressure spectra
ρ Density
τ Relaxation time
ω LBM weight function
a Speed of sound
B Law of the wall constant
b Airfoil span
Cµ Subgrid scale model constant
Ci Bhatnagar-Gross-Krook collision term
ci Discrete velocity vector
d Height of tripping device
f Frequency
∗ PhD candidate, Faculty of Aerospace Engineering, Aerodynamics, W.C.P.vanderVelden@TUDelft.nl
† Assistant professor, Faculty of Aerospace Engineering, Aerodynamics
‡ Assistant professor, Faculty of Aerospace Engineering, Wind Energy

1 of 15

American Institute of Aeronautics and Astronautics

Copyright © 2016 by the American Institute of Aeronautics and Astronautics, Inc. All rights reserved.
 gi Displacement vector of the distributed particles
H Shape factor
l Airfoil chord
lz Spanwise correlation length
Li Acoustic pressure loading term
M Mach number
p Pressure
q Dynamic pressure
R Geometric observer distance
Re Reynolds number
St Strouhal number
T Temperature
t Time
u Velocity
ue Edge velocity
u∞ Free stream velocity
Downloaded by PURDUE UNIVERSITY on June 6, 2016 | http://arc.aiaa.org | DOI: 10.2514/6.2016-3021

x, y, z Streamwise, wall-normal and spanwise direction

x Observer position
y Source position
y + Normalized wall normal distance

I. Introduction
The trailing-edge noise of a wind turbine blade is currently one of the most dominant noise sources on
a wind turbine and, therefore, understanding and modeling of the physics associated with the generation
and propagation of noise is of paramount importance for the design of silent wind turbines.1 Brooks et al.2
defined the fundamental airfoil self-noise mechanisms associated with the trailing edge, such as the noise
produced by the transitional or turbulent boundary layer flow with the trailing edge or that due to vortex
shedding. In the case of the interaction between the boundary layer flow and the trailing edge, perturbations
of the unsteady surface pressure field, introduced and convected with the turbulent eddies, are scattered at
the discontinuity posed by the trailing edge. The acoustic radiation depends largely on the length scale of
the turbulent individual eddies.3 In the case of a developed turbulent boundary layer, the surface pressure
is only affected over a highly localized area by various turbulent eddy sizes.3 Due to the small length scale
and high convective velocity of the eddies, this situation is typically encountered at high frequency. At
high frequency, the directivity pattern of the acoustic radiation shows a bias towards the leading edge (i.e.
in upstream direction).2, 3 This asymmetry arises due to the fact that the acoustic wavelengths are much
smaller than the airfoil chord, known as non-compactness of the acoustic source. For convecting turbulent
boundary layers over sharp trailing edges, where the spanwise correlation associated with turbulent eddies
is by far smaller than the airfoil span, an appropriate length scale is the local boundary layer displacement
thickness δ ? .4
Several authors, for instance Amiet5 and Howe,6 have discussed trailing-edge noise in the light of incident
turbulent flow and diffraction theory respectively. Within this framework, the relevant characteristics for
noise radiation due to boundary layer interaction with the trailing edge are the auto-spectral density (ASD),
the spanwise correlation length (lz ) of the unsteady surface pressure, and its convective velocity, which are
all at least a function of frequency ω. Amiet5 and Howe6 assumed that the incident pressure wave on the
surface of the airfoil convects past the trailing edge, which represents an impedance discontinuity and at
which the fluctuations are scattered in the form of acoustic waves.
As the main noise source is trailing edge noise at low Mach number flows, different noise reduction
techniques have been investigated. To reduce trailing edge noise, Howe6 analyzed a flat plate, with the
trailing edge modified by the presence of serrations possessing a saw tooth profile. Howe developed an
theoretic model and concluded that the intensity of radiation at the trailing edge could be reduced by such
a modification, with the magnitude of the reduction depending on the length and spanwise spacing of the
teeth, as well as the frequency of the radiation. It was determined that the dimensions of an individual
serration should be at least of the order of the turbulent boundary layer thickness and that longer, narrower
teeth should yield a greater intensity reduction. However, the theory overstimates the noise reduction and

2 of 15

American Institute of Aeronautics and Astronautics

 does not reflect the characteristic assumptions on the characteristic behavior with respect to experiments
which may be due to assumptions on the characteristics of the flow field, such as assumption of frozen
turbulence and the correct modeling of sound radiation and scattering. The latter might be Green function
related.7 Therefore, more recently, Azarpeyvand et al.8 tried to improve Howe’s theory by carrying out
analytical investigations of serrations, slits and more complex combinations of both. It was found that larger
noise reductions were achieved with more, unconventional and complex periodic serrations to the trailing
edge. The slitted-sawtooth trailing edge add-ons were found to be the most effective one in terms of noise
reduction. A further improvement has been made by Lyu et al.,9 who developed a new model to predict
the sound radiated by serrated-trailing edges. This latter theoretical model showed closer resemblance with
experiments presented in the past.10–13
Various experiments were performed on different scales. Airfoils and flat plates were tested in the past
by Dassen et al.10 A maximum of 10 dB noise reduction for the flat plate and a 8 dB reduction for the
airfoil was observed, both only in the low frequency range. Parchen et al.13 continued the experimental
investigation on trailing edge serrations, but considered full wind turbine blades and wind tunnel models.
Slightly lower noise reductions were found as compared with the study from Dassen et al.10 This study
was continued by Oerlemans et al.,12 however only a 2 to 3 dB noise reduction at low frequencies was
observed. Both Parchen et al.13 and Oerlemans et al.12 reported noise increases at high frequencies. More
Downloaded by PURDUE UNIVERSITY on June 6, 2016 | http://arc.aiaa.org | DOI: 10.2514/6.2016-3021

recently, Gruber11 attempted to investigate the aeroacoustic effects of trailing edge serrations, slits and more
complex add-ons on airfoils. On average, 3 − 5 dB reduction at low frequencies was observed using sharp
sawtooth serrations, while at high frequencies, noise increases up to 5 dB were found. It was suggested that
the significant noise decrease could be ascribed to a significant reduction of phase speed near the sawtooth
edges, in combination with a slight reduction of pressure coherence along the edge. Gruber11 confirmed the
results from Azarpeyvand et al.8 and also concluded that the slitted-sawtooth trailing edge add-ons were
the most effective noise-suppression add-on, with large noise reductions in the low frequency area and minor
noise increases (< 1 dB) in the high frequency range. More recent experiments were performed by Arce -
León et al.14 on sawtooth, slitted-sawtooth and straight trailing edges on a NACA0018 airfoil. The study
concluded the effectiveness of using noise-suppression add-ons, even at zero angle of attack. However, the
sawtooth design outperformed the slitted-sawtooth with a maximum of 5 dB at low frequencies. Important
to note though are the variations in terms of serration thickness and bending of the slits with respect to
other studies.
Numerical solutions of the flow source field and acoustic propagation have also been performed in the
past.15–17 By doing simulations one has the advantage of solving both flow and pressure fields, and solving
the experimentally difficult to obtain, wall pressure fluctuations. The computations showed that, in presence
of serrations, flow coming from pressure and suction side behave differently with respect to the straight
trailing edge case. This leads to a small modification in the near wake of the airfoil. Furthermore, the flow
is found to be highly three-dimensional, with different kind of formations of horse-shoe vortices in the space
between the serrations, in combination with a mean motion through the teeth of the serration going from
suction to pressure side. Lower actual noise reductions were found in comparison with the theoretical model
of Howe.6
Although many attempts have been made with these noise-suppression add-ons, the effectiveness of
serrations on the overall noise reduction is still not yet fully understood and might deal with, for example, a
decrease in acoustical source terms, a shift of low frequency noise to higher frequencies or a change in sound
diffraction due to the complex geometry. Especially the three dimensionality of the flow rises questions, as
it is currently unclear how this directly effects the acoustic behavior of the flow. Hence, understanding the
physics by which trailing edge serrations reduce airfoil self-noise is therefore of major importance and the
main objective of the current study. If the physics were completely understood it could potentially lead to
improvements in serration design, and possibly the development of improved, alternative techniques based
on similar mechanisms and principles.
This study focuses on the flow topology and its actual noise generation mechanism around a sawtooth,
slitted-sawtooth and straight edged NACA0018 airfoil under zero angle of attack. The flow and pressure field
is analyzed by evaluating the fully explicit, transient, compressible Lattice Boltzmann equation. Acoustic
perturbations are directly obtained in the measurement domain, making it feasible to comment on the actual
noise reduction mechanism in close comparison with earlier presented studies. The far field acoustic data is
obtained by means of a Ffowcs Williams and Hawking18 and Curle19 integral solution. A similar methodology
has been validated against experiments before as presented by van der Velden et al.20 in the past.

3 of 15

American Institute of Aeronautics and Astronautics

 II. Methodology
A. Flow simulation
Instead of solving the discretized set of the Navier-Stokes equations, the discrete Boltzmann equations
are solved for simulating fluid flows.21 The LBM method starts from a mesoscopic kinetic equation, i.e.
the Boltzmann equation, to determine the macroscopic fluid dynamics. The commercial software package
Exa PowerFLOW 5.3b is used to solve the discrete Lattice-Boltzmann equations for a finite number of
directions. The discretization considers 19 discrete velocities in three dimensions (D3Q19) involving a third
order truncation of the Chapman-Enskog expansion, which has been shown sufficient to recover the Navier-
Stokes equations for a perfect gas at low Mach number in isothermal conditions.22
The kinetic equations are solved on a Cartesian mesh, known as a lattice, by explicit time-stepping and
collision modeling. The Lattice-Boltzmann equation may then be written as:
gi (x + ci ∆t, t + ∆t) − gi (x, t) = Ci (x, t), (1)
where the particle density distribution function gi can be interpreted as a histogram representing a frequency
of occurrence at a position x with a discrete particle velocity ci in the i direction at time t. ci ∆t and ∆t
Downloaded by PURDUE UNIVERSITY on June 6, 2016 | http://arc.aiaa.org | DOI: 10.2514/6.2016-3021

are space and time increments respectively. The collision term on the right hand side of the LBM equation
adopts the simplest and also the most popular form known as the Bhatnagar-Gross-Krook (BGK) form:23
∆t
Ci (x, t) = − [gi (x, t) − gieq (x, t)] . (2)
τ
This term drives the particle distribution to the equilibrium with a relaxation time parameter τ . The variable
gieq is the local equilibrium distribution function, relates the LBM to hydrodynamic properties and is essential
for the local conservation criteria to be satisfied. The equilibrium distribution of Maxwell-Boltzmann is
approximated by a 2nd order expansion valid for small Mach number:24
ci u (ci u)2 |u|2
 
gieq = ρωi 1 + 2 + + (3)
cs 2c4s 2c2s
where ωi are the weight functions related to the velocity discretization model24 and cs = √13 is the non-
dimensional speed of sound. The equilibrium function is related to the macroscopic quantities density ρ and
velocity u, which can be computed by summing up the discrete momentum of the particle distribution:
X X
ρ(x, t) = gi (x, t), ρu(x, t) = ci gi (x, t). (4)
i i

The single relaxation time used is related to the dimensionless kinematic viscosity:24
∆t
ν = c2s (τ − ). (5)
2
The subgrid scale model, further denoted as a Very Large Eddy Simulation (VLES), is implemented as a
viscosity model through the relaxation time τ to locally adjust the numerical viscosity of the scheme:25
k 2 /
τef f = τ + Cµ , (6)
(1 + η 2 )1/2
where Cµ = 0.09 and η is a combination of a local strain parameter (k|Sij |/), local vorticity parameter
(k|Ωij |/) and local helicity parameters. The model consists of a two-equation k −  Renormalization Group
(RNG) modified to incorporate a swirl based correction that reduces the modeled turbulence in presence of
large vortical structures.26 This VLES methodology is implemented as standard turbulence model in Exa
PowerFLOW 5.1b.
Fully resolving the near wall region is computationally too expensive for high-Reynolds number turbulent
flow with the lattice concept of the LBM scheme. Therefore, a turbulent wall model is used to provide
approximate boundary conditions. In the current study, the following wall-shear stress model based on the
extension of the generalized law of the wall model is used:24, 27
 +  +
y 1 y
u+ = f = ln + B, (7)
A κ A

4 of 15

American Institute of Aeronautics and Astronautics

 with  
dp
A=1+f . (8)
dx
This relation is iteratively solved to provide an estimated wall-shear stress for the wall boundary conditions
in the LBM scheme. A slip algorithm,24 a generalization of a bounce-back and specular reflection process, is
then used for the boundary process. A variable resolution with 10 refinement zones is used, where the grid size
changes by a factor of two for adjacent resolution regions. Due to the explicit time-stepping characteristics of
the LBM scheme, the time-step size can be increased with cell size in factors of two as well. Larger cells will
therefore not be evaluated every time-step. This gives rise to the notation of time-step equivalent number
of cells, which is the number of cells scaled to operation at the shortest time-step in addition to the total
number of cells.

B. Acoustic prediction
Due to the fact that the LBM is inherently compressible and provides a time dependent solution, the sound
pressure field can directly be extracted from the computational domain, provided that there is sufficient
Downloaded by PURDUE UNIVERSITY on June 6, 2016 | http://arc.aiaa.org | DOI: 10.2514/6.2016-3021

resolution to accurately capture the acoustic waves. Sufficient accuracy is obtained when considering at
least 12 − 16 cells per wavelength for the LBM methodology.26 In addition, as for most trailing edge noise
problems, an acoustic analogy is used to obtain the far-field noise. Therefore, the directly obtained sound
field from the simulation is compared with an acoustic analogy based on the simulation fluid dynamics.
To recover the acoustic far-field, the Ffowcs Williams and Hawkings18 (FW-H) equation is employed. The
time-domain FW-H formulation developed by Farassat known as formulation 1A,28 and extended based on
the convective form of the FW-H equation is used to predict the far-field sound radiation of the beveled
trailing edge in a uniformly moving medium.29
The formulations are implemented in the time domain using a source-time dominant algorithm also
referred to as an advanced time approach.29 The input to the FW-H solver is the time-dependent flow field
on a surface mesh provided by the transient LBM simulations. This surface mesh is defined either as a
solid surface corresponding to the physical body (further denoted as Curle’s analogy19 ) or as a permeable
surface surrounding the solid body (further denoted as FW-H analogy18 ). Both methodologies are used and
compared in this study. Hence, acoustic dipole sources Li are the only source term for the current analogy,
defined as:

Li = (p − p0 )ni (9)

with p − p0 the fluctuating pressure on the solid surface and ni the surface normal in the ith direction. To
determine the far field pressure spectra, the distance between the observer (x) and the source position (y),
R needs to be defined. It can be written as:
−M0 (x1 − y1 ) + R?
R= , (10)
β2
with p
R? = (x1 − y1 )2 + β 2 [(x2 − y2 )2 + (x3 − y3 )2 ], (11)
and q
β= 1 − M02 . (12)
Technically, R represents the effective acoustic distance rather than the geometric distance between the
source and the observer in terms of time delay between emission and reception. The unit radiation vector
then reads:
−M0 R? + (x1 − y1 ) x2 − y2 x3 − y3
 
R̂ = , , . (13)
β2R R R
With the source term Li and the observer distance from the source R defined, the following integral relation
will solved:29
Z " #
1 L̇i R̂i
4πp0A A(x, t) = dS
a0 g=0 R(1 − Mi R̂i )2
ret

5 of 15

American Institute of Aeronautics and Astronautics

 " #
Li R̂i − Li Mi
Z
+ dS (14)
g=0 R2 (1 − Mi R̂i )2 ret
" #
Li R̂i (Mi R̂i − M 2 )
Z
+ dS.
g=0 R2 (1 − Mi R̂i )3 ret

The subscript ret denotes the evaluation of the integrand at the time of emission. The acoustic probes are
equally distributed in a circle around the trailing edge, 1.5 chords away.

III. Test case

A NACA0018 airfoil with a chord of l = 0.2 m and straight trailing edge is considered as baseline model
in an undisturbed turbulence free (< 0.1%) flow field under zero angle of attack of 20 m/s (M = 0.06),
resulting in an chord based Reynolds number of Rel = 265, 000. Spanwise, cyclic boundary conditions are
applied with a modeled span of b = 0.4l. Transition is enforced by a zig-zag transition strip of height
dtrip = 0.003l and length ltrip = 0.015l on both sides at a downstream location of x = 0.2l. One zig-zag
Downloaded by PURDUE UNIVERSITY on June 6, 2016 | http://arc.aiaa.org | DOI: 10.2514/6.2016-3021

spans btrip ≈ 0.015l, resulting in a repetitive pattern of 27 triangles.

Sawtooth Slitted-Sawtooth
2h = 0.2l

d = 0.0025l

λ = 0.1l Y
0.4l
s=

l=0
.2 m

Z X

Figure 1. Airfoil and serration geometry and dimensions

Two trailing edge modifications are tested in this current campaign; sawtooth serrations and slitted-
sawtooth serrations. A sketch of the geometry is found in Fig. 1. The serration model considered in this
study has teeth of 2h = 0.04 m (2h = 0.2l) length and λ = 0.02 m (λ = 0.1l) width, resulting in a length-
width ratio of λ/h = 1. The dimensions of the slits are similar to the dimensions of the width of the needles
and defined by the spacing d = 0.0025l. The slits are cut all way through to the trailing edge of the airfoil.
Both serrations are of flat-plate type, with a constant 1 mm (tser = 0.005l) thickness throughout the entire
span and length, thus not changing the nominal thickness of the straight airfoil trailing edge. The first
serration was placed mid-span, with a total of 4 serrations being modeled with the chosen span.
The simulation domain size is a block of 12l in both streamwise and wall normal direction. Outside a
circular refinement zone of 10l diameter an anechoic outer layer is used to damp acoustic reflections. A total

6 of 15

American Institute of Aeronautics and Astronautics

 of 10 refinement regions are applied such that, near the boundary, the first cell is placed in the viscous sub
layer, i.e. at 0.00039l. This results in an average y + value over the airfoil surface trailing edge of 3. In
total, around 150 million voxels were used to completely discretize the problem, with a total of 65 million
time-step equivalent voxels. 0.1 seconds of simulation time (10 flow passes) took 6300 CPU hours on a Linux
Xeon E5-2690 2.9 GHz platform of 80 cores.
After the transient phase of 15 flow passes, sampling is started. The Courant-Friedrichs-Lewy (CFL)
number is dependent on the wave propagation velocity and smallest voxel size in this compressible simulation
and fixed to unity for each single simulation. Therefore, the physical time step is fixed at 1.3 · 10−7 s. Data
is sampled at 30 kHz (St = f l/u = 300) for 0.2 physical seconds at local lattice size, resulting in a recording
of 20 flow passes or 6000 samples. Fourier transformed data is obtained with a Hanning window with 50 %
overlap, to smooth out the spectra.

IV. Result and discussion

A. Acoustic emission
The acoustic waves, originating from the interaction between the flow and solid at the trailing edge, can
Downloaded by PURDUE UNIVERSITY on June 6, 2016 | http://arc.aiaa.org | DOI: 10.2514/6.2016-3021

be clearly visualized by means of dilatation snapshots. Dilatation, defined by ρ1 ∂ρ ∂t ,

30
is visualized in Fig. 2
around a chord based Strouhal number of St = f l/u∞ = 5 for the sawtooth trailing edge case.
An alternating pattern of positive and negative dilatation on the airfoil wall and in the wake is present due
to the presence of the instability in the turbulent boundary, generating pressure fluctuations and therefore
density changes, while growing over the airfoil surface and in the wake. It shows a comparatively short
wavelength compared to the order of the airfoil chord. Moreover, acoustic waves propagate radially outward
from the trailing edge and posses a wavelength larger by an order of magnitude when compared to the
convected scales in the wake. This large difference in wavelength between hydrodynamic and acoustics
perturbations at the same frequency arises due to large ratio of the speed of sound and the convective
velocity in the wake, or the low Mach number considered in this case. It should be noted that the acoustic
pressure on the suction and pressure side shows a phase opposition, which is reminiscent of a dipole, and
displays the directivity characteristics of trailing edge noise for the convected dipole case31 with higher
amplitudes for the upstream propagating waves. Both acoustic waves from both airfoil sides cancel out
each other upstream the leading edge, creating a zone of silence and resulting in a directivity of a compact,
dipole-like, behavior.

Figure 2. Snapshot of the dilatation field around the sawtooth serrated trailing edges band passed at St =
f l/u∞ = 5 (500 Hz)

7 of 15

American Institute of Aeronautics and Astronautics

 At a location directly positioned above the trailing edge, the power spectrum is extracted using the three
different methods described; direct probes in the computational domain, applying Curle’s anology with body
sources (airfoil) only and applying FW-H analogy using a porous funnel located around the airfoil and two
chords of the wake. The data presented in Fig. 3 is scaled with a Mach number of M = 1, an observer radius
of R = 1 m and a span of b = 1 m by means of:

R2
Φscaled = +10 log10 ( ). (15)
bM 5
Effectively, this leads to an amplitude scaling by the free-stream velocity to the power five. This is an
often used scaling for typical problems of trailing-edge scattered noise problems, where non-compact noise
sources are the main source of interest.3, 32, 33 Due to the cyclic boundary conditions and limited span,
the acoustic pressure field in the numerical solution contains contributions from mirrored coherent image
sources of the airfoil arriving through the cyclic domain boundaries to the microphone location. To correct
for this, in addition to previous scaling parameters, another correction has to be applied to the sound spectra.
Oberai34, 35 derived a three-dimensional, frequency dependent, correction for low Mach number flows, which
can be rewritten in a dB form as:
Downloaded by PURDUE UNIVERSITY on June 6, 2016 | http://arc.aiaa.org | DOI: 10.2514/6.2016-3021

f b2
Φcyclic = +10 log10 ( ). (16)
aR
This correction has been applied on the direct probes measured in the simulated domain. The results of
pressure spectrum (Φa a) as well as the actual noise reduction with respect to the straight trailing edge
(positive is reduction) is presented in Fig. 3.

110
Sawtooth FW-H
Sawtooth Curle 6
Sawtooth direct
Slitted-sawtooth FW-H
100 Slitted-sawtooth Curle
Slitted-sawtooth direct
Φaa [∆dB]
Φaa [dB]

Straight FW-H 4
Straight Curle
Straight direct
90
Sawtooth FW-H
2 Sawtooth Curle
Sawtooth direct
Slitted-sawtooth FW-H
80 Slitted-sawtooth Curle
Slitted-sawtooth direct
0
4 8 16 32 4 8 16 32
f l/u∞ f l/u∞

Figure 3. Observer and span normalized far-field sound pressure levels (left) and reduction (right) directly
above the trailing edge

The overall trends are similar for each geometry and for each aeroacoustic method; high sound pressure
levels at the low frequencies together with a decay towards the higher frequencies. The power spectra do not
show any significant peaks, indicating no distinctive shedding is present at the trailing edge. In general, the
noise being observed is of broadband type, where the dilatation field (Fig. 2) clearly pinpointed the source
region to the trailing edge of the airfoil. Both serrated trailing edges show noise reductions between St = 4
and St = 32, where the most significant noise reduction is being observed by sawtooth edges with a maximum
of 6 dB at the very low frequencies. The slitted-sawtooth serrations show less of a noise reduction (about
2 − 4 dB) at the very low frequencies, although it reaches a 5 dB reduction in the mid frequency range as the
noise reduction increases a bit until similar reductions as with the solid sawtooth serrations are observed.
After St = 8, the noise reduction reduces gradually until it reaches similar conditions as with the reference,
straight trailing edge case. Similar results were presented before in literature, for example by Arce - León et
al.14 Here, the cross-over frequency for where the slitted-sawtooth design matches the solid-sawtooth design
and the cross-over frequency where a noise increases with respect tot the straight trailing edge was observed
were determined at f l/u∞ = 13 and f l/u∞ > 34 respectively. This is in close agreement with the results

8 of 15

American Institute of Aeronautics and Astronautics

 presented in this study. Finally, integrating the noise reduction levels result in an overall sound pressure
level reduction of 4.1 dB for the solid-sawtooth and a 2.8 dB reduction for the slitted-sawtooth levels.
The variance between the different aeroacoustic methods is negligible in the mid frequency range (St =
4 − 12), while the direct method starts to deviate from the Curle and FW-H approach after St = 16,
indicating the noise plateau reached due to the lower cut-off frequency. This could be solved by running the
solver in double-precision mode.20
The dilatation plot presented earlier showed the appearance of a convected dipole behavior, radiated
towards the leading edge. To take a closer into the noise radiation direction process, a directivity analysis
is performed. An array of 360 microphones positioned in a circle at mid-span around the trailing edge are
used to record the acoustic pressure fluctuations derived from the FW-H integral solution method. Results
are depicted in three different frequency bands in Fig. 4.
105° 90° 75° 105° 90° 75° 105° 90° 75°
120° 60° 120° 60° 120° 60°
135° 45° 135° 45° 135° 45°

150° 30° 150° 30° 150° 30°

165° 15° 165° 15° 165° 15°

Downloaded by PURDUE UNIVERSITY on June 6, 2016 | http://arc.aiaa.org | DOI: 10.2514/6.2016-3021

1.6 1.2 2 1.6 1.5 1.3

0.8 0.4 1.2 0.8 1.1 0.9
±180° 0° ±180° 0.4 0° ±180° 0.7 0°
Φaa (α, ∆f )/Φaa (∆f ) Φaa (α, ∆f )/Φaa (∆f ) Φaa (α, ∆f )/Φaa (∆f )

-165° -15 ° -165° -15 ° -165° -15 °

-150° -30 ° -150° -30 ° -150° -30 °

-135° -45 ° -135° -45 ° -135° -45 °

-120° -60 ° -120° -60 ° -120° -60 °
-105° -90 ° -75 ° -105° -90 ° -75 ° -105° -90 ° -75 °

Sawtooth uf∞l = 2 − 8 Sawtooth uf∞l = 8 − 16 Sawtooth uf∞l = 16 − 32

Slitted-sawtooth uf∞l = 2 − 8 Slitted-sawtooth uf∞l = 8 − 16 Slitted-sawtooth uf∞l = 16 − 32
Straight uf∞l = 2 − 8 Straight uf∞l = 8 − 16 Straight uf∞l = 16 − 32

Figure 4. Directivity plot for the straight, sawtooth and slitted-sawtooth trailing edge models under different
Strouhal numbers

The behavior in terms of directivity is as expected; at low Strouhal numbers, a compact dipole source
arises from the trailing edge. Increasing the frequency leads to a tilted dipole, directed towards the leading
edge. When further increasing the frequency, non-compact behavior appears and not only source radiation,
but also source-body interaction, known as scattering, is captured and propagated to the far-field using the
integral solution method.
In between St = 2 − 8, the directivity trends are similar between the straight, slitted-sawtooth and
sawtooth trailing edge, although a large amplitude difference is observed, indicating the noise reduction for
the serrated trailing edges. At the mid frequencies, depicted in green, the results behave similarly in terms
of their directivity pattern and with identical noise reductions for both serrated trailing edges. At higher
frequencies small variations are present when comparing the directivity shape from the straight trailing edge
with the serrated trailing edge. It seems that in case of the serrated trailing edges, upstream traveling waves
are canceled out more effectively, resulting in a less upstream oriented directivity shape. Therefore, in the
current study, larger noise reductions with serrated edges are found in the upstream (135◦ − 150◦ ) direction,
as can been seen from Fig. 4. The physical explanation is tried to be deduced from the flow results, which
will be addressed in the next section.

B. Source field
The input for the acoustic results from last section were derived from a flow analysis using the Lattice
Boltzmann methodology around a NACA0018 airfoil with different trailing edges. The flow was tripped in
order to bypass the transition process. To visualize the transition process and the growth of the turbulent
boundary layer over the airfoil edge, an instantaneous view is displayed in Fig. 5. As the airfoil is placed at
zero angle of attack, the flow is symmetric on both sides of the airfoil and hence, no pressure and suction
side can be distinguished. The boundary layer convects over the sharp trailing edge, and a thick wake with
a small shedding component is visible.

9 of 15

American Institute of Aeronautics and Astronautics

Downloaded by PURDUE UNIVERSITY on June 6, 2016 | http://arc.aiaa.org | DOI: 10.2514/6.2016-3021

Figure 5. Instantaneous view of the flow around a NACA0018 airfoil with straight trailing edge

As discussed before, the hydrodynamic velocity fluctuations in close vicinity of the wall will affect the
wall pressure fluctuations, in which the wall will act as an effective acoustic dipole source. Although there
is no straightforward model to connect both physical quantities, it is interesting to look at different key
parameters in the boundary layer which could affect the scattering process at an edge. Three different points
along the serrated edge are selected; in the root, mid and tip and compared to values obtained from the
straight trailing edge. Wall-normal samples of the mean streamwise velocity and Reynolds shear stresses
are extracted and plotted in Fig. 6 while the Reynolds normal stresses are plotted in Fig. 7. All plots are
normalized according to the boundary layer parameters, measured in mid-span at the trailing edge of the
straight case. The boundary layer thickness is determined based on the variation of the spanwise vorticity.
This asymptote is reached at a thickness of δ = 9.7 mm, or 0.0485l. At this position, the edge velocity
ue = 0.94u∞ is obtained and further used for normalization. The displacement thickness and momentum
thickness at the trailing are δ ? = 0.0105l and θ = 0.007l respectively, resulting in a shape factor of H = 1.58.

1.25 1.25

1 1

0.75 0.75
y/δ

y/δ

0.5 0.5

0.25 0.25

0 0
0 0.2 0.4 0.6 0.8 1 -0.01 -0.0075 -0.005 -0.0025 0
u/ue u‘v‘/u2e

Figure 6. Mean velocity profile (left) and Reynolds shear stress (right) at different locations on the edge for
the straight (black), sawtooth (solid) and slitted-sawtooth (dashed) case

The mean streamwise flow in Fig. 6 display small variations for both serrated trailing edges when con-
sidering the point in the valley of the serration (red line) with respect to the straight trailing edge (black).

10 of 15

American Institute of Aeronautics and Astronautics

 The boundary layer upstream of the trailing edge seems not to be affected by shape of the edge. A more
turbulent profile is found downstream the sawtooth edge (green and blue line), where a compression of the
boundary layer shows a thinner boundary layer thickness. The slitted-sawtooth edge (dashed line) show close
resemblance with the solid-sawtooth (solid line) case, although small variations, with lower mean streamwise
velocity are present in the lower part of the boundary layer. This is likely because of the porous capabilities
(larger surface) of the edge enhancing the shear layer. Fig. 6 also represents the Reynolds shear stress,
commonly to give information of structures sweeping and ejecting in the boundary layer. This could then be
connected to wall pressure fluctuations, as is proposed by, for example Chong and Vathylakis.36 All stresses
in Fig. 6 nicely drop down to zero for values of y/δ = 0. The largest shear stresses are observed for upstream
points, denoting a gradual decrease of shear over the edge of the serration. The slitted-sawtooth enhances
mixing and therefore show higher shear stress amplitudes. A maximum is observed at about y/δ = 0.3 for all
cases, although with slitted-sawtooth edge it looks like the fluctuations are spread out between y/δ = 0.1−0.6
at the tip of the serration.
1.25 1.25
Downloaded by PURDUE UNIVERSITY on June 6, 2016 | http://arc.aiaa.org | DOI: 10.2514/6.2016-3021

1 1

0.75 0.75

y/δ
y/δ

0.5 0.5

0.25 0.25

0 0
0 0.005 0.01 0 0.005 0.01
v′v′/u2e w‘w‘/u2e

Figure 7. Reynolds normal stress profiles at different locations on the edge for the straight (black), sawtooth
(solid) and slitted-sawtooth (dashed) case

Reynolds normal stresses in wall-normal and spanwise direction at the edge location could also be con-
sidered as key parameters of assessing the effectiveness of the noise reduction process using serrations. The
general trend shows maximum amplitudes around y/δ = 0.3 for both wall normal and spanwise stresses, with
the largest amplitudes upstream along the edge. The characteristics at the upstream part of the serrated
edges are therefore likely more effective in scattering acoustic waves. The incoming boundary layer remain
unaffected in wall-normal direction, but shows some minor variations in spanwise direction. The slitted-
sawtooth enhances the mixing process at the current zero angle of attack case, as both normal stresses show
higher amplitudes compared to the solid-sawtooth edge results. This deviation increases when going further
downstream the edge, also observed by Arce - León et al.14 in their PIV study. For the current study, it is
therefore interested to take a closer look along this serrated edge.
This analysis is performed by the use of Fig. 8 and Fig. 9, using the axis system earlier defined in Fig. 1.
Mean and RMS velocity in streamwise, wall-normal and spanwise direction were extracted on a line over
the sawtooth edge at different wall-normal locations in the boundary layer. The streamwise mean values at
the different locations in the boundary layer are as expected; a gradually increase of u/u∞ over the edge
towards the tip of the serration, as well as an increase towards u∞ when moving away from the surface. As
expected, the opposite is found for the selected locations when considering the mean, wall-normal velocity.
In general, the flow moves towards the gaps between the serrations (negative velocity in this specific case).
Flow in a higher part of the boundary layer experiences a stronger downward movement of the flow at the
valley of the serration. Again, this has been observed before by PIV experiments.14 Over the serrated edge
in downstream direction, this effect is drastically reduced. Mean flow in the lower part of the boundary layer
experiences larger wall-normal variations; with wiggles denoting local stronger and weaker injections of the
flow. Exceeding y/δ = 0.5 this behavior is negligible. v/u∞ deviations between both the solid and slitted-
sawtooth edges are small, while the general trend shows that u/u∞ is reduced a bit towards the tip when

11 of 15

American Institute of Aeronautics and Astronautics

 applying slitted-sawtooth edges. The spanwise velocity, which is of major importance in the determination
of the flow angle over the edge, shows some interesting trends. At the valley and in the viscous sub layer
of the boundary, flow is moving toward the serrated edge (negative velocity in this specific case). Directly
after, the flow is pushed outwards showing the highest w/u∞ at 0.05l, or 25 % of the serrated edge, resulting
is a misaligning of the flow with respect to the edge. The mean spanwise velocity reduces when moving
further downstream with some small wiggles, while keeping a positive direction towards the gap between
the serrations. Further outside the boundary, the variations are slightly smaller and deviate between both
inwards moving and outward moving spanwise flow, resulting in respectively positively (aligned with edge)
and negatively (misaligned with edge) oriented flow angles over the serrated edges. This angle is drastically
improved when considering the slitted-sawtooth design, as can be seen from the lower amplitude of the mean
spanwise velocity in Fig. 8.
0.9 0 0.03

0.8
-0.02 0.02
0.7

w/u∞
-0.04
u/u∞

0.6
v/u∞
y/δ = 0.1 0.01
Downloaded by PURDUE UNIVERSITY on June 6, 2016 | http://arc.aiaa.org | DOI: 10.2514/6.2016-3021

0.5 -0.06 y/δ = 0.2

y/δ = 0.3
0.4 y/δ = 0.4 0
-0.08
0.3 y/δ = 0.5
y/δ = 0.6
0.2 -0.1 -0.01
0 0.05 √ 0.1 0.15 0.2 0 0.05 √ 0.1 0.15 0.2 0 0.05 √ 0.1 0.15 0.2
x2 + z 2 /l x2 + z 2 /l x2 + z 2 /l

Figure 8. Mean velocity at different wall normal locations over the edge of the sawtooth (solid) and slitted-
sawtooth (dashed) serration

Next, rms values are investigated, as presented in Fig. 9. As Fig. 7 already revealed, the maximum for
either streamwise, wall-normal and spanwise fluctuations were reached around y/δ = 0.3. This is also seen
from the rms plots over the entire edge (yellow lines). In combination with higher amplitude fluctuations
near the valley of the serration, which could also easily represents the conditions at the straight trailing
edge as concluded before, these are the general trends seen in Fig. 9. More specifically, the amplitude of the
local fluctuations attains higher values by using the slitted-sawtooth serrations. As these fluctuations are
key parameters of impacting the wall surface pressure fluctuations, it is likely that they would initiate more
effective acoustic waves, resulting in a lower acoustic reduction then its solid-sawtooth counterpart.
0.14 0.1 0.11
y/δ = 0.1
0.13 y/δ = 0.2 0.09 0.1
y/δ = 0.3
0.12 y/δ = 0.4 0.08 0.09
w′/u∞
u′/u∞

v′/u∞

y/δ = 0.5
0.11 y/δ = 0.6 0.07 0.08

0.1 0.06 0.07

0.09 0.05 0.06

0.08 0.04 0.05

0 0.05 √ 0.1 0.15 0.2 0 0.05 √ 0.1 0.15 0.2 0 0.05 √ 0.1 0.15 0.2
x2 + z 2 /l x2 + z 2 /l x2 + z 2 /l

Figure 9. RMS velocity at different wall normal locations over the edge of the sawtooth (solid) and slitted-
sawtooth (dashed) serration

The mean velocity values presented before can better be interpreted by using a vector plot. At a distance
of y/δ = 0.1, the vectors (not to scale) are plotted in each plane. The incoming flow is immediately altered
due to the serrated edge at the valley; the vectors show an increase in y and z direction, resulting in a more
perpendicular streamline over the edge. This is an unwanted effect, as the flow should be aligned as much
as possible with the serrated edge.6 This effect is mostly canceled out towards the tip of the serration. It
is clear to see that the slitted design enhances the streamlines to be more aligned to the ‘porous’ edge. The
wall-normal variation on the other hand is clearly not uniform over the serrated edge, but is shows large
variations over both edges. This behavior needs further study in a next campaign.

12 of 15

American Institute of Aeronautics and Astronautics

 0.075
Sawtooth
0.05
Slitted-sawtooth

z/l
0.025

0
0 0.05 0.1 0.15 0.2 0.25
x/l
0.02
0.01
y/l

0
-0.01
-0.02
0 0.05 0.1 0.15 0.2 0.25
x/l
Downloaded by PURDUE UNIVERSITY on June 6, 2016 | http://arc.aiaa.org | DOI: 10.2514/6.2016-3021

0.02
0.01
y/l

0
-0.01
-0.02
0 0.025 0.05 0.075
z/l

Figure 10. Vector plot for each plane around the serrated trailing edge at a wall-normal distance of y/δ = 0.1

Lastly, the local spinning motion due to the insert of a sawtooth trailing edge design is analyzed using
Fig. 11 by means of vorticity magnitude slices over the serration and in the near wake. At the first slice, close
agreement is found for the initial, spanwise uniform, boundary layer for both models. Over the serrations,
the vorticity magnitude is gradually reduced. Slightly stronger and local vortical structures are present
in between the solid-sawtooths, while the slitted-sawtooth enhances the mixing locally, resulting in a local
increase in vorticity at the tip of the serration. This results in a slightly thicker and dense wake downstream,
as can been seen from the last slice.

Figure 11. Mean vorticity magnitude planes at different locations along the sawtooth (left) and slitted-sawtooth
(right) serration

V. Conclusion
The trailing-edge noise of a wind turbine blade is currently one of the most dominant noise sources on a
wind turbine and, therefore, understanding and modeling of the physics associated with the generation and

13 of 15

American Institute of Aeronautics and Astronautics

 propagation of noise is of paramount importance for the design of silent wind turbines. In the past, using
both experiments and simulations, a serrated trailing edge has been shown to be efficient in reducing trailing
edge noise. Although many attempts have been made with these noise-suppression add-ons, the effectiveness
of serrations on the overall noise reduction is still not yet fully understood.
This study focused on the flow topology and noise emission around a sawtooth, slitted-sawtooth and
straight edged NACA0018 airfoil under zero angle of attack. The flow and pressure field was analyzed by
evaluating the fully explicit, transient, compressible Lattice Boltzmann equation. Acoustic perturbations
were directly obtained from the computational domain, as well as with two integral solution methods.
The three aeroacoustic methods were showing similar results in the frequency region of interest. Noise
reductions up to 6 dB are achieved for the sawtooth serrated edge in comparison with the straight trailing
edge case, while the slitted-sawtooth edge reaches a maximum of only 5 dB. At low frequencies, the solid-
sawtooths outperform the slitted-sawtooths. The noise reduction vanishes above f c/u∞ = 32.
The flow field is further analyzed by looking at both mean and fluctuating velocity in streamwise, wall-
normal and spanwise direction. The general trend shows favorable mean quantities for the slitted-sawtooth
design, as the angle over which the flow convects over the edge is more aligned. The solid-sawtooth experi-
ences a more perpendicular velocity vector near the valley of the serration, therefore increasing the source
Downloaded by PURDUE UNIVERSITY on June 6, 2016 | http://arc.aiaa.org | DOI: 10.2514/6.2016-3021

generation. The counterpart though is that, overall, the slitted-sawtooth serrated edge show higher fluctua-
tions over the mean values. Potentially, these parameters are of paramount importance during the generation
process. A frequency analysis study in near future will be able to assign these variations to the wall pressure
fluctuations, thereby directly discussing the acoustic performance.

Acknowledgments
This research is funded and supported by Siemens Wind Power A/S, Brande, Denmark.

References
1 Oerlemans, S., Sijtsma, P., and Lopez, B. M., “Location and quantification of noise sources on a wind turbine,” Journal

of Sound and Vibration, Vol. 299, No. 4-5, 2007, pp. 869–883.
2 Brooks, T., Pope, D., and Marcolini, M., “Airfoil self-noise and prediction,” Tech. rep., NASA Reference Publication

1218, 1989.
3 Blake, W., Mechanics of flow-induced sound and vibration, volumes I and II , Academic Press, 1986.
4 Spalart, P., “Numerical simulation of boundary layers: Part 1. Weak formulation and numerical method,” Tech. rep.,

NASA Reference Publication 882222, 1986.

5 Amiet, R., “Noise due to turbulent flow past a trailing edge,” Journal of Sound and Vibration, Vol. 47, No. 3, 1976,

pp. 387–393.
6 Howe, M., “Trailing edge noise at low Mach numbers,” Journal of Sound and Vibration, Vol. 225, No. 2, 1999, pp. 211–

238.
7 Pröbsting, S., Schneiders, J. F., Avallone, F., Ragni, D., Arce-León, C., and Scarano, F., “Trailing-edge noise diagnostics

with low-repetition-rate tomographic PIV,” 22nd AIAA/CEAS Aeroacoustics Conference, 2016.

8 Azarpeyvand, M., Gruber, M., and Joseph, P., “An analytical investigation of trailing edge noise reduction using novel

serrations,” 19th AIAA/CEAS Aeroacoustic Conference, Vol. 2009, 2013.

9 Lyu, B., Azarpeyvand, M., and Sinayoko, S., “A Trailing-Edge Noise Model for Serrated Edges,” 21st AIAA/CEAS

Aeroacoustics Conference, Vol. 2362, 2015, pp. 1–24.

10 Dassen, A., 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,” Proceeding of the European Union Wind Energy Conference and
Exhibition, pp. 800–803.
11 Gruber, M., Joseph, P., and Azarpeyvand, M., “An experimental investigation of novel trailing edge geometries on airfoil

trailing edge noise reduction,” 19th AIAA/CEAS Aeroacoustic Conference, Vol. 2011, 2013.
12 Oerlemans, S., Sijtsma, P., and Lopez, B. M., “Reduction of wind turbine noise using optimized aifoils and trailing-edge

serrations,” AIAA Journal, Vol. 47, 2009, pp. 1470–1481.

13 Parchen, R., Hoffmans, W., Gordner, A., and Braun, K., “Reduction of airfoil self-noise at low Mach number with a

serrated trailing edge,” International Congress on Sound and Vibration, 1999, pp. 3433–3440.
14 Arce-León, C., Avallone, F., Ragni, D., and Pröbsting, S., “PIV Investigation of the Flow Past Solid and Slitted Sawtooth

Serrated Trailing Edges,” 54th AIAA Aerospace Sciences Meeting, Vol. 1014, 2016.
15 Jones, L. and Sandberg, R., “Acoustic and hydrodynamic analysis of the flow around an aerofoil with trailing edge

serrations,” Journal of Fluid Mechanics, Vol. 706, 2012, pp. 295–322.

16 Arina, R., Rinaldi, R. D. R., Iob, A., and Torzo, D., “Numerical study of self-noise produced by an airfoil with trailing-edge

serrations,” 18th AIAA/CEAS Aeroacoustic Conference, Vol. 2184, 2012.

17 Sanjose, M., Meon, C., Masson, V., and Moreau, S., “Direct numerical simulation of acoustic reduction using serrated

trailing-edge on an isolated airfoil,” 20th AIAA/CEAS Aeroacoustics Conference, Vol. 2324, 2014.

14 of 15

American Institute of Aeronautics and Astronautics

 18 Ffowcs-Williams, J. and Hawkings, D., “Sound generation by turbulence and surfaces in arbitrary motion,” Philosophical

Transactions of the Royal Society od London, Vol. 264, 1969, pp. 321–342.
19 Curle, N., “The influence of solid boundaries upon aerodynamic sound,” Proceedings of the Royal Society of London,

Vol. 231, 1955, pp. 505–514.

20 van der Velden, W., Pröbsting, S., de Jong de Jong, A., van Zuijlen, A., Guan, Y., and Morris, S., “Numerical and exper-

imental investigation of a beveled trailing edge flow and noise field,” 21st AIAA/CEAS Aeroacoustics Conference, Vol. 2366,
2015.
21 Succi, S., The lattice boltzmann equation for fluid dynamics and beyond, Oxford University Press, 2001.
22 Frisch, U., D’Humiéres, D., Hasslacher, B., Lallemend, P., Pomeau, Y., and Rivet, J., “Lattice gas hydrodynamics in two

and three dimensions,” Complex Systems, Vol. 1, 1987, pp. 649–707.

23 Bhatnagar, P., Gross, E., and Krook, M., “A model for collision processses in gases. Small amplitude processes in charged

and neutral one-component systems,” Physical Review , Vol. 94, No. 3, 1954, pp. 511–525.
24 Chen, S. and Doolen, G., “Lattice Boltzmann method for fluid flows,” Annual Review of Fluid Mechanics, Vol. 30, 1998,

pp. 329–364.
25 Chen, H., “Extended Boltzmann kinetic equation for turbulent flows,” Science, Vol. 301, No. 5633, 2003, pp. 633–636.
26 Habibi, K., Gong, H., Najafi-Yarzdi, A., and Mongeau, L., “Numerical Simulations of Sound radiated from Internal Mixing

Nozzles with Forced Mixers using the Lattice Boltzmann Method,” 19th AIAA/CEAS Aeroacoustic Conference, Vol. 2143, 2013.
27 Crouse, B., Freed, D., Senthooran, M., Ullrich, S., and Fertl, S., “Analysis of underbody windnoise sources on a production

vechicle using a Lattice Boltzmann scheme,” SAE Technical Paper , Vol. 1, No. 2400, 2007.
28 Farassat, F. and Succi, G., “The prediction of helicopter discrete frequency noise,” Vertica, Vol. 7, No. 4, 1983, pp. 309–
Downloaded by PURDUE UNIVERSITY on June 6, 2016 | http://arc.aiaa.org | DOI: 10.2514/6.2016-3021

320.
29 Bres, G., Perot, F., and Freed, D., “A Ffowcs Williams-Hawkings solver for Lattice-Boltzmann based computational

aeroacoustics,” 16th AIAA/CEAS Aeroacoustic Conference, Vol. 3711, 2010.

30 Wagner, C., Huttl, T., and Sagaut, P., Large-Eddy Simulation for Acoustics, Cambridge, 2007.
31 Roger, M. and Moreau, S., “Extensions and limitations of analytical airfoil broadband noise models,” International

Journal of Aeroacoustics, Vol. 9, No. 3, 2010, pp. 273–305.

32 Ffowcs-Williams, J., “Hydrodynamic Noise,” Annual Review of Fluid Mechanics, Vol. 1, 1969, pp. 197–222.
33 Költzsch, P., Strömungsmechanisch erzeugter Lärm, Ph.D. thesis, Technische Universität Dresden, 1974.
34 Oberai, A., Roknaldin, F., and Hughes, T., “Trailing edge noise due to turbulent flows,” 02-002, Boston University, 2002.
35 Ewert, R. and Schröder, W., “On the simulation of trailing edge noise with a hybrid LES/APE method,” Journal of

Sound and Vibration, Vol. 270, 2004, pp. 509–524.

36 Chong, T. and Vathylakis, A., “On the aeroacoustic and flow structures developed on a flat plate with a serrated sawtooth

trailing edge,” Journal of Sound and Vibration, Vol. 354, 2015, pp. 65–90.

15 of 15

American Institute of Aeronautics and Astronautics