Technical ReportPDF Available

Investigation of Wind-Induced Noise to Optimize Masts


Abstract and Figures

In modern society the awareness of disturbing noise has increased. Thus it is of importance to find the cause of noise generation and prevent the designs which produces them. In this report the wind-induced whistling noise from two sailboat mast profiles have been evaluated. It has previously been observed that one of the profiles generates a whistling noise while the other one does not. The predetermined conditions were a free stream velocity of 20 m/s and a yaw angle of 33. The commercial software STAR-CCM+ has been used along with aero-acoustic features. The simulations showed that no whistling noise could be identified. Although several interesting features regarding the flow around the mast geometries have been discovered, no root cause of the whistling noise could be completely established. In order to fully understand the noise generation, additional work needs to be performed on the subject. Thus it is recommended to examine more yaw angles and velocities of the incoming flow. In this manner it is the authors believe that it would be possible to finally find the answer of what is causing the whistling noise.
Content may be subject to copyright.
Investigation of Wind-Induced Noise
to Optimize Masts
Project in Applied Mechanics
Department of Applied Mechanics
Gothenburg, Sweden 2017
Project in Applied Mechanics 2017
Investigation of Wind-Induced Noise
to Optimize Masts
Jonathan Fahlbeck
Johan Forsgren
Sankar Raju Narayanasamy
Martin Ottosson
David Winqvist
Department of Applied Mechanics
Division of Fluid Mechanics
Chalmers University of Technology
Gothenburg, Sweden 2017
Investigation of Wind-Induced Noise to Optimize Masts
Jonathan Fahlbeck, Johan Forsgren, Sankar Raju Narayanasamy, Mar-
tin Ottosson, David Winqvist
©Jonathan Fahlbeck, Johan Forsgren, Sankar Raju Narayanasamy,
Martin Ottosson, David Winqvist, 2017.
Supervisors: Dr. Hua-Dong Yao and Prof. Lars Davidsson, Applied Mechanics,
Examiners: Dr. Håkan Johansson and Dr. Valery Chernoray, Applied Mechanics,
Project in Applied Mechanics 2017
Department of Appied Mechanics
Division of Fluid Mechanics
Chalmers University of Technology
SE-412 96 Gothenburg
Telephone +46 31 772 1000
Cover: An isosurface of the λ2criterion with the velocity magnitude vizualised along
the surface, for mast geometry (b).
Gothenburg, Sweden 2017
In modern society the awareness of disturbing noise has increased. Thus it is of
importance to find the cause of noise generation and prevent the designs which
produces them. In this report the wind-induced whistling noise from two sailboat
mast profiles have been evaluated. It has previously been observed that one of the
profiles generates a whistling noise while the other one does not. The predeter-
mined conditions were a free stream velocity of 20 m/s and a yaw angle of 33. The
commercial software STAR-CCM+ has been used along with aero-acoustic features.
The simulations showed that no whistling noise could be identified. Although sev-
eral interesting features regarding the flow around the mast geometries have been
discovered, no root cause of the whistling noise could be completely established.
In order to fully understand the noise generation, additional work needs to be per-
formed on the subject. Thus it is recommended to examine more yaw angles and
velocities of the incoming flow. In this manner it is the authors believe that it would
be possible to finally find the answer of what is causing the whistling noise.
Keywords: Whistling, Noise, STAR-CCM+, Mast, FW-H, Aero-Acoustics, Vortex
First of all we would like to thank our supervisor Hua-Dong Yao. Without his
continuous support and advice, the overall understanding and progress within the
aero-acoustic field, would not have been attained. His happy spirit helped us through
many dark hours of computational errors.
The computer resources have been of utter most importance to secure results. A
special thanks is thus directed to Chalmers University of Technology.
Finally we would like to thank Seldén Mast. Without their initial idea and interest,
this project would not have been possible. William Holt deserves extra acknowl-
edgement for his commitment to the progress of the project.
Jonathan Fahlbeck, Johan Forsgren, Sankar Raju Narayanasamy, Martin
Ottosson, David Winqvist, Gothenburg, May 2017
List of Figures xi
List of Tables xi
Nomenclature xii
1 Introduction 1
1.1 Background ................................ 1
1.2 CaseDescription ............................. 2
1.3 ProjectGoal................................ 2
1.4 Limitations ................................ 2
2 Theory 3
2.1 AcousticTheory.............................. 3
2.2 ComputationalTheory.......................... 5
2.2.1 Basics of Fluid Dynamics . . . . . . . . . . . . . . . . . . . . 5
2.2.2 Basics of Computational Fluid Dynamics . . . . . . . . . . . . 6
2.2.3 Ffowcs Williams-Hawking formulation . . . . . . . . . . . . . 6
2.2.4 DNS, LES, RANS, DES . . . . . . . . . . . . . . . . . . . . . 7
3 Method and Computational Settings 8
3.1 Pre-processing............................... 8
3.1.1 Acoustic Domain Setup . . . . . . . . . . . . . . . . . . . . . . 8
3.1.2 Meshgeneration ......................... 9
3.1.3 FW-H Surface and Receivers . . . . . . . . . . . . . . . . . . . 11
3.1.4 Mesh Dependency Study . . . . . . . . . . . . . . . . . . . . . 11
3.2 Simulations ................................ 12
3.2.1 Boundary Conditions and Material Properties . . . . . . . . . 12
3.2.2 RANS Simulation . . . . . . . . . . . . . . . . . . . . . . . . . 12
3.2.3 DESSimulation.......................... 13
3.2.4 Stopping Criteria . . . . . . . . . . . . . . . . . . . . . . . . . 13
3.3 Post-processing .............................. 13
3.3.1 Fast Fourier Transform and Sound Pressure Level . . . . . . . 14
3.3.2 Overall Sound Pressure Level . . . . . . . . . . . . . . . . . . 14
3.3.3 λ2Criterion............................ 14
3.3.4 ContourPlots........................... 14
4 Results and Discussion 15
4.1 Velocity .................................. 16
4.2 Sound Pressure Level . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
4.3 Overall Sound Pressure Level . . . . . . . . . . . . . . . . . . . . . . 17
4.4 λ2Criterion and Vorticity. . . . . . . . . . . . . . . . . . . . . . . . . 18
4.5 AdditionalResults ............................ 20
4.6 General Discussion of the Results . . . . . . . . . . . . . . . . . . . . 21
5 Conclusion 24
Bibliography 25
List of Figures
1.1 Mast profiles, (a)has a whistling noise, (b)does not have a whistling
noise..................................... 1
3.1 Domain and the mast profile (b). .................... 9
3.2 View of a directed volume mesh with 5.6 million cells, for mast (b). . 10
4.1 Velocity contour plots. . . . . . . . . . . . . . . . . . . . . . . . . . . 15
4.2 A-weighted SPL. Data recorded in two receivers. See Figure 4.4 for
thelocation................................. 16
4.3 PSD of static pressure in the probes located close to the tips. Com-
parison between the two configurations. . . . . . . . . . . . . . . . . . 17
4.4 Overall Sound Pressure Level around the two different mast configu-
rations. Note that 0refers to downstream of the mast. . . . . . . . . 18
4.5 Isosurface of λ2=250000 1/s2, for mast configuration (a)(top) and
for mast configuration (b)(bottom).................... 18
4.6 Vorticity contour plots. . . . . . . . . . . . . . . . . . . . . . . . . . . 19
4.7 Mast configuration (a)........................... 20
4.8 Vorticity near the mast. . . . . . . . . . . . . . . . . . . . . . . . . . 20
List of Tables
3.1 Domain design parameters. . . . . . . . . . . . . . . . . . . . . . . . . 9
3.2 Material data for dry air at 15 °C and atmospheric pressure [12]. . . . 12
AA Aero-Acoustics
ASZ Acoustic Suppression Zone
CAA Computational Aero-Acoustics
CFD Computational Fluid Dynamics
CFL Courant–Friedrichs–Lewy
CTH Chalmers University of Technology
DES Detached Eddy Simulation
DNS Direct Numerical Simulation
FFT Fast Fourier Transform
FW-H Ffowcs Williams-Hawkings
LES Large Eddy Simulation
OASPL Overall Sound Pressure Level
PSD Power Spectral Density
RANS Reynolds-Averaged Navier-Stokes
SPL Sound Pressure Level
1.1 Background
In today’s society, people are much concerned about noise pollution. Hence indus-
tries strive to make products less noisy in order to comprehend with the awareness
and sensitiveness of their customers. The aim of this project is to investigate the
unwanted whistling noises created by sailboat masts when the boat is moored and
the sail is stowed inside the mast.
An annoying whistling sound may arise when the flow velocity and the yaw angle are
large enough in certain boat mast designs. The company Seldén Mast AB desires
to know the root cause of this whistling noise in order to enable them to design
the masts in accordance to prevent it. Seldén Mast provided Chalmers University
of Technology (CTH) with two mast segments, of which one has been experienced
to produce a higher noise in comparison to the other. In Figure 1.1 the two mast
profiles are illustrated. Since the observations of the whistling sound is based on
experiences rather than experimental data the conditions of when this behaviour
occurs is not fully known. It has been estimated to a free stream velocity of 20 m/s
and a yaw angle from 20to 45.
Figure 1.1: Mast profiles, (a)has a whistling noise, (b)does not have a whistling
The aim of this collaboration between CTH and Seldén Mast is to achieve a win-
win situation. By using computational fluid dynamics tools, the group at CTH
aimed to help Seldén Mast in their understanding on the whistling noise generation.
Thereby enabling the company to reduce unwanted sounds generated by the masts
and making necessary changes in the design. The proposed analysis of the mast
designs will hopefully help the company in making them more competitive. In
1. Introduction
addition to the assistance provided to Seldén Mast, the group at CTH intended to
improve their skills in project work as well as to enhance their knowledge in fluid
mechanics and aero-acoustics.
1.2 Case Description
The two mast configurations in Figure 1.1 are investigated regarding noise gener-
ation. Mast (a)has a cross sectional dimension of 148 mm ×266 mm and mast
(b)has a cross sectional dimension of 270 mm ×140 mm. As noise is generated
by turbulence, 3D computations are performed. The height of the two profiles are
set to 250 mm, and thus the height of the 3D domain. The flow velocity is 20 m/s
and the yaw angle 33°. The fluid is dry air at 15 °C and atmospheric pressure. As
3D computations are challenging, instability issues such as numerical errors, conver-
gence problem and flow field development are expected. Mesh generation is also an
area where challenges might be encountered.
1.3 Project Goal
Use acoustic wave modeling to explain the effects of the tip1geometry in a mast
regarding noise generation. The project should result in recommendations to consider
while designing a mast to prevent noise generation.
1.4 Limitations
In order for the project to be feasible within the time frame a number of bound-
aries and limitations are applied. The project was to be carried out from March to
May 2017. The limits are determined by either absolute factors, such as the amount
of time available, or in order to assure that an overall quality of the work is attained.
Since unsteady 3D computations was performed, which are both time and computa-
tionally demanding, time available and computational resources are of importance.
Hence it is hard to establish a fine balance between number of cells and computa-
tional time.
The field of aero-acoustics was new for the project group. In the narrow time frame
a large amount of the time available had to be used in order to fully understand the
underlying problems in the field of aero-acoustics. Thus the task is limited to inves-
tigate noise caused by turbulence. Other types of sounds are not to be investigated.
All the simulations are performed in the commercial software STAR-CCM+. The
software is also used in the analysis part of the project in combination with MAT-
1The tip geometry refers to the tips at the opening of the mast, see Figure 1.1.
2.1 Acoustic Theory
Acoustics refers to the science of sound and originates from the Greek word for
hearing [1]. Sound is of high significance in our every day life when it comes to
communication, awareness of our surroundings and orientation. In recent years the
awareness of sound and noise has increased. One needs to note that not all sound
is bad. We still want to hear a click when we turn the key in the door, music or
sounds alerted by warning devices. The sound which today is unwanted is the one
that is "out of place" and disturbing [2].
It is known that sharp bangs or long time exposure to loud noises can damage our
hearing. In fact noise can also cause other health issues or even death. For example,
flies that are exposed to sound levels around 160 dB die after a short period of time.
To get this noise level into perspective one can compare this with the noise produced
by a refrigerator ~50 dB, a chain saw ~100 dB or a shotgun ~140 dB. Humans who
spend much time in areas with loud noise usually have a higher blood pressure and
heart rate. Noise has also been proven to affect sleeping patterns and just simply
annoys people. All these facts have made companies and whole industries more
aware of the effects of noise. Today acoustics is an important aspect in product
development [3].
In order to understand how noise travels, one can explain it by looking at what
happens when you throw a rock into a lake. When the rock hits the water surface,
the rock will be slowed down. The kinetic energy of the rock then is transferred into
the water and causes ripples to form on the water surface. The ripples are gradually
transported outwards and heating the water slightly as they pass and fade away.
In the same way the noise travels from, for instance, a sudden clap of your hands.
Energy will spread from the clap in a series of sound waves which consists of regions
with increased pressure. The fluid particles will move closer together for a short
period of time as the wave passes. For a louder clap, the fluid particles move tighter
together and a higher pressure difference is created. Like the ripples on the water,
the sound waves of the clap will die away as the energy eventually is transferred into
heat. Since the losses are small, sound waves can travel a very long distance before
they fade away [2].
There are different ways by which noise or sound can be created. For instance, the
one mentioned above which was as a "clap". Others are frictional noise or vibrations
2. Theory
noise from structures. Since the cause of these sounds are fairly straight forward,
the industries ability to dampen these sounds has come pretty far. In recent years
the removal of the "out of place" sounds has therefore been focusing on other sources
of sound such as aero-acoustics. An example of the importance of aero-acoustics is
when you are in a car driving at high velocities. If you open the window the aero-
acoustic noise will be the dominating one. Aero-acoustic noise generation can in a
simple manner be explained by the flow interacting with geometrical irregularities
of a car, an airplane, etc. The interaction creates unsteady turbulent flows which
are often detached and thus in turn generates noise. The noise generation can be
divided into two phenomena: impulsive noise and turbulent noise [3][4].
Impulsive noise is generated by the movement of surfaces or a surface in a non-
uniform flow condition. The non-stationary load on the body causes pressure fluc-
tuations to occur, which are generated as sound. This noise can be estimated in a
fairly simple manner from aerodynamic simulations. Turbulent noise, on the other
hand is harder to predict but it is quite common and predominantly exists. Since
the turbulence is stochastic by nature, it has a broad frequency spectrum. The
turbulent noise therefore creates a broadband noise consisting of many frequencies.
Turbulent energy is most efficiently transferred into acoustic energy in the presence
of sharp edges. The sharp edge forces two flows of different velocities to have a
sudden blend. An example of this is when a flow crosses a bluff body. Unless the
body is aerodynamically shaped the fluid will separate from the body and cause a
wake to form behind the body. The wake will consist of fluid with low velocity. The
pressure difference between the still fluid and the fluid crossing the body will cause
them to blend. This causes strong local equalizing flow which in turn results in peak
pressures, i.e impulsive noise. Behind a bluff body the pressure will characteristi-
cally alternate to the different sides of the wake in order to comprehend with the
velocity of the moving flow on both sides of the wake. This effect is a Von Kármán
vortex street which refers to the fluid twirls of alternating directions forming behind
an object. Vortex shedding is the phenomenon of every fluid twirl "coming loose"
from the body [3][5].
As stated in previous paragraph, vorticity is generated at boundaries by the relative
velocity of two surfaces, such as fluid and wall. When a certain threshold is reached
the two-dimensional wakes formed by vorticity, shed from the bluff body surface.
This results in a transition into the third dimension. The feature explains the fact
that the vorticity can be stored as a vortex street, leading to conservation of vor-
ticity in a system. Since the vorticity is mainly created just downstream of a bluff
body the vorticity further downstream is merely a response of the rest of the fluid,
trying to adjust to the instability that the bluff body initialized. Vortices tend to
merge downstream of the bluff body, creating larger sections of rotating fluid with
lower velocity. As a result one could state that the vorticity generated immediately
downstream of the bluff body is of higher importance [6].
Since turbulent broadband noise always exists when turbulence is present, one can
state that Aero-Acoustic (AA) noise consists of a broadband noise. Sometimes it is
2. Theory
accompanied by a narrow-band impulsive noises. As previously stated, though the
impulsive noise is fairly easy to predict, in order to predict the turbulent noise one
has to estimate the turbulence. Hence, in order to fully estimate the AA noise one
has to use CFD-tools. This is a fairly recent methodology that has been evolved
with the utilization of computers. This field of science is called Computational Aero-
Acoustics (CAA) [3][4].
There are two main methodologies in CAA to compute an acoustic field. The first
one being "direct method", which is considered to be the most exact and would
be the equivalent to DNS in the CFD field. In the direct method the governing
equations for the flow and acoustic field are solved over the entire domain, from the
aerodynamic effective area to a far-field observer. The fact that the domain of inter-
est is very large, makes the direct method extremely expensive when it comes to the
number of cells and time steps. The more common way to perform AA computa-
tions is to use hybrid methods. The sound generation in the aerodynamic effective
area is decoupled from the transport of the sound to the far-field. In more simple
terms one can say that one method is used for the sound generation and another
method is used for the transport process. There are several methods available, both
for sound generation and the transport process [3].
Even though the possibility to fully estimate the aero-acoustic noises is fairly new
due to improvement in technology (computers) today, a lot of the ground work
was performed in the 1960s. Many of the equations still used today are based on
the acoustical transport techniques from half a century ago. The most frequently
mentioned ones are the Lighthill analogy and the Ffowcs Williams–Hawkings (FW-
H) equation. These are used as transport methods while performing calculations
with the hybrid techniques [3]. The FW-H equation is explained in the section
2.2 Computational Theory
In order of fully understanding CAA some basic principles need to be addressed. As
stated in section 2.1, CAA is a mixture of acoustic theory and CFD. In this chapter
the theory behind these aspects will be explained. The aim is that this chapter will
help the reader in the understanding of the final results.
2.2.1 Basics of Fluid Dynamics
The principle of fluid dynamics is that it describes a pattern of flows. In fluid
mechanics it is common to investigate a flow within a fixed control volume. The
flow is governed by the continuity (eqn 2.1), Navier-Stokes (eqn 2.2) and energy
(eqn 2.3) equations [7].
dt +ρ∂vi
= 0 (2.1)
2. Theory
dt =∂p
∂xj"µ vi
dt =σij
Here ρis density, vvelocity, xlength, ttime, ppressure, uinternal energy, q
conductive heat flux, fbody force and σthe stress tensor.
2.2.2 Basics of Computational Fluid Dynamics
In CFD the variables of a flow field are solved for inside a domain of interest using a
computational software. The domain is discretized into a number of finite volumes or
cells in which the governing equations are solved in an iterative process. Depending
on the type of flow, different numerical schemes are used to discretize the differential
2.2.3 Ffowcs Williams-Hawking formulation
This formulation is a more general form of Lighthill’s equation [8][9], where moving
walls are allowed to be present inside the domain. The formulation is based on the
assumption that a volume B(t)exists which is enclosed by the surface S(t). The
surface needs to be sufficiently smooth to allow the definition of a smooth function
x, t)such that
x, t) =
>0if x
= 0 if x
<0if outside B(t).
One can note that the Heaviside function of this function, H(h), will be zero inside
the volume B(t)and unity outside B(t). If the mass conservation equation and
momentum equation is multiplied by H(h), use Lighthill’s procedure for acoustic
variable, p0=pp0, apply Green’s theorem and use the free-space Green’s function,
equation 2.4 can be obtained [3].
x, t) = 2
4πr #τ=te
4πr #τ=te
4πr #τ=te
∂t Z
4πr(1 Mr)#τ=te
4πr(1 Mr)#τ=te
σij is the viscous stress, nis the normal to the surface, H=H(h),r=||x
is the body force, c0is the speed of sound in the fluid surrounding the listener, b
the velocity of the moving surface, Mr=b·
y)/rc0and te=tr/c0.
2. Theory
The two surface integrals in equation 2.4 are later used to set up the FW-H surface
integral as equation 2.5. See section 3.1.3 for further information.
x, t) =
∂t Z
4πr(1 Mr)#τ=te
4πr(1 Mr)#τ=te
dS (2.5)
The major models that are utilized for solving turbulence in CAA or CFD as a whole
are Direct Numerical Simulation (DNS), Large Eddy Simulation (LES), Reynolds
Averaged Navier-Stokes Simulation (RANS) and Detached Eddy Simulation (DES)
DNS solves all the turbulence scales, making it highly expensive. Thus it is not
suggested to be performed for objects with larger dimension with the computational
facilities currently available.
LES resolves all large turbulent scales, i.e, it is volume averaged. The principal
idea behind LES is to reduce the computational cost by modelling or exclude the
smallest scales, which are the most computationally expensive to resolve. This is
achieved via averaging across time and space, which effectively removes small-scale
information from the numerical solution.
RANS models all turbulence. It is time-averaged and different turbulence models
are applied. Thus making it faster than other methods to compute. The most com-
mon turbulence models are the kεor kω.
DES is a combination of advantages of both RANS and LES. In DES, the scales
close to the wall are solved using RANS (or unsteady-RANS), within the boundary
layer. Further away from walls are the scales resolved using LES. Thus DES behaves
as a hybrid RANS-LES model.
Method and Computational Settings
While performing the analysis of the two mast designs a methodology divided into
three stages was used. The different stages were Pre-processing, Simulation and
Post-processing. Most of the work was performed in the commercial software STAR-
CCM+. The work was initially performed on mast profile (a)(see Figure 1.1) due
the simplicity of the geometry. All through the process, thorough documentation of
the choices made were performed. This was later used when repeating the set up
for mast (b). In this fashion a precise comparison between the two designs could be
3.1 Pre-processing
When performing analysis using numerical methods, the quality of the final result
is highly dependent of the work done prior to pushing the "run button". A well
planned domain using a suitable mesh design to account for the important features
is therefore crucial. In this section the different parts of the Pre-process will be
3.1.1 Acoustic Domain Setup
The outer domain design was based on the mast mean diameter Dm. In Figure 3.1
and in Table 3.1 the design features are visualized. The principle of the design in
Figure 3.1 is that the white area (outer domain) is formed by a coarser mesh while
the gray area (inner domain) has a finer mesh. This feature was applied in order
to capture the small-scale flow structures downstream of the mast. The bold line
which is dividing the areas is a FW-H surface, see section 3.1.3. A RANS simulation
was performed and gave a good estimation of the size and location of the turbulent
flow field behind the mast. This simulation was used to define the gray shaded
refinement zone and the FW-H surface.
In general is it necessary to use a very large domain when numerically calculating
acoustics. This is because the pressure fluctuations that noise is built from are small
and therefore sensitive to disturbance. Any presence of reflected noise could affect
the result. The following methods were used to reduce the risk of reflecting noise
and numerical errors:
To reduce the reflections of sound with high frequency an increased cell size
near the boundary was used. This makes the mesh unable to capture the
3. Method and Computational Settings
sounds since approximately 20 cells per wavelength is needed to capture the
sound correctly [10]. This design also has the advantage that it reduces the
computational effort.
Regarding the sound with low frequency an acoustic Suppression Zone (ASZ)
was used around the boundary. ASZ dampens the acoustic waves before they
reach the outer boundaries to prevent them from reflect back into the domain.
The zone can therefore be thought of as a sponge absorbing the sound. The
zone size, l, could be estimated from the speed of sound, c0, and predicted
minimum frequency, according to equation 3.1.
c0= 340 m/s, fminimum = 500 Hz l=c0
= 0.68 m (3.1)
Table 3.1: Domain design parameters.
Parameter DmR L r w D d
Used size (D+d)/2 15Dm25Dm3Dm13Dm0.27 m0.14 m
(a) Domain seen from above (top) and from
the side (bottom).
(b) Illustration of mast profile (b),
with geometrical lengths.
Figure 3.1: Domain and the mast profile (b).
3.1.2 Mesh generation
In order to find a suitable mesh for performing the simulations, a couple of parame-
ters needed to be taken into account. Firstly, the mesh should not exceed more than
10 million cells. This was to ensure that the computational resources provided were
used effectively. Secondly, the mesh needed to be fine enough in the areas where
small-scale vortex shedding occurs. This is where sound generation is expected. The
main areas of interest were the region around the tips (Figure 3.2c) and the region
behind the mast (Figure 3.2b). The mesh in these regions was controlled by a wake
3. Method and Computational Settings
refinement operation. Here one could specify the direction of the anticipated flow,
to which the geometry will be exposed. The length, angle and the cell growth rate
of the zone were controlled. The mesh can be seen in Figure 3.2.
(a) Top view of the whole domain. (b) Zoomed on the wake region.
(c) Zoomed on the tips. (d) Cut plane around the mast.
Figure 3.2: View of a directed volume mesh with 5.6 million cells, for mast (b).
To properly resolve the acoustics, the cell size must be small enough. One good
practise in estimating the cell size is to use the fact that at least 20 cells per wave-
length is preferable [10]. With the speed of sound, estimated to be the standard
value at sea level, c0= 340 m/s, combined with an approximate value for what
the frequency of the noise might be, fwhistling 1000 Hz, the cell size can then be
calculated according to equation 3.2.
CAA =c0
= 0.017 m (3.2)
This cell size was desirable to transport the sound correctly in the inner domain.
In section 3.1.1 a couple of ways to minimize the risk of reflected noise are listed.
These strategies were included into the mesh.
In addition to the mesh design in regards where to have large/small cells, the actual
mesh design needed to be accounted for. In 3D simulations, as in this case, three
common mesh types are tetrahedral, hexahedral and polyhedral. Tetrahedral is the
least complex with four vertices, four faces and six edges. The other ones are more
3. Method and Computational Settings
complex and are built from several tetrahedral cells. This results in the tetrahedral
mesh being the fastest to generate, but lacks in accuracy when solving the actual
problem, since the error decreases as the number of cell faces increases. In acoustic
simulations, a large number of iterations is in general required in order to attain a
correct solution. Therefore, even though the mesh takes longer time to generate,
the polyhedral mesh was the most suitable candidate, due to a higher accuracy at
a shorter computational time.
The whole mesh generation process was performed in an iterative manner where
different mesh strategies were tested. Methods that were tested and later discarded
were for example mesh designs using volumetric control for the entire domain, an-
other one using surface controls. The finally chosen mesh generation method was a
directed mesh. The advantage of this strategy is that the mesh is easy to control
regarding number of cells. The mesh is also faster to compute in comparison to
the other used methods. The fact that this thesis aims to compare two different
geometries was an argument for the choice of an easily controlled mesh and thus the
directed mesh.
3.1.3 FW-H Surface and Receivers
In order to capture the noise generated by the mast, a permeable FW-H surface
was applied within the computational domain. The data from CFD computations
were then extracted along the FW-H surface as suitable source terms in the acoustic
equations [11], see section 2.2.3 and equation 2.5.
The FW-H surface was applied as an internal interface between inner and outer re-
gion within the fluid domain, see bold line in Figure 3.1. Once AA had been enabled
in the used software, the interface was treated as a permeable surfaces within the
"FW-H surface" menu.
A number of FW-H receivers (36) was adopted outside of the fluid domain to record
the acoustic data. The receivers were installed in a circular pattern, with a radius of
10 m, around the mast. Since the acoustic transport equations are decoupled from
the flow equations it is possible to place the receivers outside of the domain.
3.1.4 Mesh Dependency Study
A mesh dependency study is planned in order to establish mesh independent results.
It is desirable to find a mesh giving accurate results without wasting computational
The study was planned so that the general design of the mesh were unchanged
and the only difference was the number of cells. Three different meshes were cre-
ated for mast configuration (a)(see Figure 1.1) with 2.2, 4.4 and 7 million cells.
Probes were inserted into the domain to measure velocity and pressure to later be
able to compare the results. The plan was then to choose one mesh and apply this
3. Method and Computational Settings
set up also for mast configuration (b). Unfortunately problems were encountered
regarding finding a converged solution. A lot of time was spent on fixing this and
when a converged solution was obtained the time was too short to analyze all the
meshes. Therefore, a fully completed mesh dependency study was not performed.
For the final results, a 4.4 million cell mesh and a 5.6 million cell mesh were used
for mast configuration (a)and mast configuration (b), respectively. The difference
in cell number between the mast configurations is due to varying complexity in the
3.2 Simulations
This section consists of different settings and options used for the simulations in
STAR-CCM+. The settings and preferences used in all simulations are based upon
[10], discussions with supervisor and the project group.
3.2.1 Boundary Conditions and Material Properties
The names of the domain boundaries can be found in Figure 3.1. The Inlet bound-
ary was set as a velocity inlet with a laminar incoming flow.Sides and Outlet were
set as pressure outlet. Atmospheric pressure at sea level was prescribed for the pres-
sure outlets. The Upper and the Lower boundaries were set to periodic boundary
condition where the flow propagates from one boundary to the other.
The fluid is dry air at 15 C and atmospheric pressure (101325 Pa). Other material
properties under consideration are presented in Table 3.2. The ideal gas law is a
viable assumption for the cases investigated. It is a fair approximation, since the
regular air at non-extreme temperatures and pressure are used in these simulations.
Table 3.2: Material data for dry air at 15 °C and atmospheric pressure [12].
kg·KiP r
288 1.226 17.96 25.24 1006 0.7159
Ttemperature, ρdensity, µkinematic viscosity, κthermal conductivity, Cpspecific
heat capacity, P r Prandtl number.
3.2.2 RANS Simulation
Initially a RANS simulation was performed to understand the turbulent flow field
around the mast. This simulation was performed to get an initial solution, from
which a more comprehensive DES simulations could be performed, see section 3.2.3.
The RANS simulation was performed as a steady state simulation using segregated
flow solver but without any AA features. The SST kωmodel was used together
with all y+wall treatment. During the RANS, first order schemes were used so that
a solution for the flow field could be attained rapidly.
3. Method and Computational Settings
3.2.3 DES Simulation
DES and AA tools were utilized in order to find and determine the source of the
noise. The initial solution was obtained from the steady state RANS simulation. In
contrast to the RANS the DES was carried out with an implicit unsteady scheme
where a low CFL number was desirable. A Segregated flow solver was applied, just
as for the RANS. The utilized model was SST kωwith all y+wall treatment
enabled. In the DES, second order schemes were used to achieve a higher accuracy.
The convection was treated with a Hybrid-BCD scheme.
During DES, the number of iterations per time step was used as an important factor
to govern the over all convergence. It is critical to attain convergence within every
time step and then consequently convergence in time as well. The number of iter-
ations per time step was set to 10 due to promising convergence while monitoring
the residuals.
To develop the correct unsteady behaviour in the flow field the fluid needs to pass
through the domain several times. In practice this meant that two different options
could be used. Either the simulation needed to go on for a long time, which would be
expensive regarding computational power. Another alternative is to start at a large
time step and develop the field and then gradually decreasing it until a sufficiently
low time step is reached. The second option was used during this project. The time
step was first set to 10 ms and then decreased in several steps to a final time step of
0.1ms. The final time step gave a maximum CFL number of around 10. This was
considered the lowest possible time step due to the computational power available
and was therefore good enough for this project.
The enabled acoustic options were "FW-H unsteady", along with an Acoustic Sup-
pression Zone (ASZ). The FW-H was used with the "On-the-fly" model option which
provides noise prediction in the receivers in parallel with the CAA. The Inlet, Sides
and Outlet boundaries had the ASZ set to 0.68 m. At the other boundaries no
ASZ was applied. The ASZ was configured to suppress low frequencies since high
frequency is dampened by increasing cell size near the boundaries, see Section 3.1.1.
3.2.4 Stopping Criteria
The stopping criteria were based on residual plots, monitor plots and receivers in
the far field. It was of importance to let the simulation run for sufficiently long time
for the receivers to capture enough noise data.
3.3 Post-processing
In the Post Processing stage, the focus is to analyze the gathered data. This section
explains the methods of comparison when analyzing the wind induced noise level
for the different masts.
3. Method and Computational Settings
3.3.1 Fast Fourier Transform and Sound Pressure Level
For analyzing the aero-acoustics, a Fast Fourier Transform (FFT) algorithm was
used to process the signals recorded by the receivers. The receivers record the pres-
sure fluctuations over time and the FFT algorithm transforms the time dependent
signals from the time domain into the frequency domain. After the FFT the Sound
Pressure Level (SPL) was plotted as function of frequency to distinguish tonal peaks
in the signal. This operation was efficiently performed using an FFT tool in STAR-
CCM+. The SPL was calculated using the A-weighting function. This means that
low frequency noise is filtered out in the similar way as in the human ear.
In order to analyze if the noise recorded in the receivers was generated from vortex
shedding produced by the tips of the masts, probes were introduced in these areas
to monitor the static pressure. An FFT algorithm computation was then performed
on the pressure-time history from the probes. This procedure was similar to that
of the data obtained from the receiver. The difference is that instead of the SPL,
the Power Spectral Density (PSD) was computed. PSD describes the signal power
per unit frequency. To confirm if the sound originates from this region, the pressure
spectra from the receivers and that of the probes should be of similar pattern. If a
pressure peak is present in the SPL from the receivers, a similar peak should appear
in the PSD from the probe data. This would imply that the noise is generated in
the region of the probe.
3.3.2 Overall Sound Pressure Level
In order to visualize the direction in which the highest level of noise is produced,
the Overall Sound Pressure Level (OASPL) was computed in every receiver location.
The OASPL is based on the root mean square value of the pressure fluctuations.
MATLAB was utilized for these calculations.
3.3.3 λ2Criterion
The λ2criterion is a technique that was utilized to visualize the vortices in the
turbulent flow using isosurfaces. An isosurface is a surface in space where a certain
variable is constant, in this case λ2. This method utilizes the eigen value of the
strain rate tensor, Sand vorticity tensor, .λ2is then the median eigenvalue from
S2+ 2, i.e, λ1λ2λ3[13] [14]. Hence it was of interest to investigate λ2, as
the vorticity produced in the wake region is known to have a strong correlation to
the generation of noise.
3.3.4 Contour Plots
Velocity plots were utilized to visualize the flow field. It was important to analyze
the vorticity contour plots, as they depict the vortex shedding which is correlated
to noise generation.
Results and Discussion
This chapter presents the results and the discussion of those. To simplify the under-
standing of noise generation, the flow field is discussed first. Gradually the various
physical aspects of noise generation are presented and analyzed. The results are
based upon the 4.4 million mesh for mast (a)and a 5.6 million mesh for mast (b).
(a) Mast configuration (a). Top view (top) and side view
(b) Mast configuration (b). Top view (top) and side view
Figure 4.1: Velocity contour plots.
4. Results and Discussion
4.1 Velocity
To get a better understanding of the flow field around the masts, the velocity con-
tour plots are shown in Figure 4.1. As expected, the flow fields look similar for
both configurations. A wake pattern has been formed downstream of the mast and
shows a Von Kármán vortex street. The vortex street is caused by the unsteady
flow separation. From Figure 4.1 one could see that the vortices in the wake are
created from the bluff body and not from the tip geometry. Mast configuration (b),
which was not expected to create a high level of noise, also seems to create a larger
unsteadiness than mast configuration (a). This indicates that mast configuration
(b) should produce a higher level of noise than mast configuration (a), according to
the theory in section 2.1.
4.2 Sound Pressure Level
As stated in section 2.1, the noise consists of a turbulent and an impulsive part.
The turbulent noise is often referred to as broadband noise and consists of noise in
several frequencies. This project has been aimed to investigate the source of the
tonal noise which originates from the impulsive noise. In Figure 4.2 an FFT has
been performed on the pressure-time data from the receivers and is presented as
an A-weighted SPL. The spectra for both mast configurations show a broadband
noise without any obvious protruding peaks, indicating that no tonal noise can be
observed. The graphs show that the SPL for configuration (b)is higher than for mast
geometry (a). Although, the sound pressure levels are considered to be relatively
(a) Receiver at 90°.(b) Receiver at 270°.
Figure 4.2: A-weighted SPL. Data recorded in two receivers. See Figure 4.4 for
the location.
In order to evaluate the origin of the noise, probes were deployed in the domain for
data comparison with the receivers. The location of the probes can be seen in Figure
4.3a. In this case the receivers do not indicate any tonal noise generation. In order
to verify if this result is correct, it is possible to look at the pressure fluctuations
4. Results and Discussion
in the probes, see Figures 4.3b, 4.3c and 4.3d. Here it can be seen that no obvious
spectral peaks occur. Hence it is possible to state that the recordings in the receivers
are reliable and no tonal noise is created in any of the mast configurations.
(a) Probe locations. (b) Probe 1.
(c) Probe 2. (d) Probe 3.
Figure 4.3: PSD of static pressure in the probes located close to the tips. Com-
parison between the two configurations.
4.3 Overall Sound Pressure Level
The Overall Sound Pressure Level for both mast configurations can be found in
Figure 4.4. The Figure is oriented as such that the inflow comes from the left hand
side (180°). The Figure shows the OASPL in each of the 36 receivers deployed
around the mast. The shape of the noise profile is similar for both the cases. The
highest noise is found diagonally downstream of the mast. The lowest noise is found
upstream of the mast, which is reasonable as the mast is blocking the sound which
is created in the wake. It is also evident that the sound level is higher for mast
configuration (b).
4. Results and Discussion
Figure 4.4: Overall Sound Pressure Level around the two different mast configu-
rations. Note that 0refers to downstream of the mast.
4.4 λ2Criterion and Vorticity.
The λ2criterion is visualized for the two mast configurations in Figure 4.5. As
can been seen mast (b), contradictory to real world experience, generate an intenser
vortex shedding than (a)near the tips. Thus it could be claimed that mast configu-
ration (b)will generate a larger noise at this yaw angle. Also in this visualization it
can be seen that vortices seems to be created from the mast body and not from the
tips. This gives a hint that, for this case, the tips do not generate vortex shedding
and thus hardly a tonal noise, according to the theory.
Figure 4.5: Isosurface of λ2=250000 1/s2, for mast configuration (a)(top) and
for mast configuration (b)(bottom).
4. Results and Discussion
The vorticity contours, see Figure 4.6, show that the mast (b)generates stronger
vorticity than configuration (a)near the tips. As vorticity is a good indicator of
noise, one could claim that configuration (a)produces less noise compared to mast
(b). Also, it could be found that the generation of vorticity is largely governed by
the bluff body, rather than the tips at this angle of attack.
(a) Mast configuration (a). Top view (top) and side view
(b) Mast configuration (b). Top view (top) and side view
Figure 4.6: Vorticity contour plots.
4. Results and Discussion
By both the visualization of the vorticity in Figure 4.6 and the λ2-criterion in Figure
4.5 one can see that major vorticity occurs near the mast for the (b)case and not
in (a). This could be some indication of that vortex interaction with the tips of
configuration (b)will not create any tonal noise, since the interaction occurs but the
mast still not produces any tonal noise according to Figure 4.2. For configuration
(a)nothing about the correlation between vortex interaction with the tips and tonal
noise can be stated, since the vortices are completely separated from the tips.
4.5 Additional Results
While the project progressed, several different simulations were performed. Due
to the low rate of acoustic simulation experience, many of these simulations were
in purpose of practice. Consequently the results from the simulations have been
of varying quality. During this period of trial and error a few interesting findings
have been made. The results from using an initial coarser mesh with 2 million cells,
showed signs of producing more noise than the final result. It is of importance to
note that due to the mesh being more coarse, these results cannot fully be trusted.
Nevertheless a few interesting details can be observed. In Figures 4.7-4.8 a compar-
ison has been made between an initial result for a coarse mesh and the final result.
The sound pressure level for the two meshes can found in Figure 4.7. In Figures
4.8a and 4.8b the vorticity is plotted close to the tips for both meshes respectively.
As previously stated in section 2.1, noise and vorticity does in theory have a strong
correlation. It can be seen in Figures 4.7-4.8 that the higher sound pressure level
coincides with a higher degree of vorticity around the tips of the mast.
Figure 4.7: Mast configuration (a).
(a) Coarse mesh.
(b) Final mesh.
Figure 4.8: Vorticity near
the mast.
4. Results and Discussion
4.6 General Discussion of the Results
To summarize the results, it is found that mast configuration (b)generates a higher
noise level than (a). This is in contrast to what was expected from real life observa-
tions. According to Seldén Mast the noise from mast (a)should be louder than the
noise from mast (b). No significant tonal noise has been recorded during the simu-
lations. The only noise that has been captured is broadband noise. The captured
noise level has also been very low, approximately 20 dB.
It is of interest to evaluate why the sound levels found in the simulations do not
coincide with the expectations. A topic of interest is the angle of attack which is
evaluated. This angle is set to be 33°. As stated previously, it does not look like the
vorticity reaches the tips of the mast (a). From the observations made prior to the
project it was expected that the noise was occurring when the flow was coming in at
an angle of attack of around 20°- 45°. If another angle of attack had been examined,
the results might have differed.
A source of error could be that in real life the flow is never laminar. There is
always turbulence and fluctuations in the wind. In the presence of turbulent flow,
the separation of the flow occurs at a later stage from the surface of the bluff body.
This would perhaps convey the fact that the vortices would be created closer to the
tips of the mast. Hence the tips of the mast would be exposed to larger pressure
fluctuations and noise would be generated. The delayed separation in combination
with the fluctuating wind direction would imply that the noise generation in real
life is less sensitive to the angle of attack in regard with the simulated case.
It is of interest to discuss why there is a difference in result between the initial
solution with 2 million cells and the final result for configuration (a). It would have
been of great interest to perform a full mesh dependency study. Due to lack of time,
this could not be done and hence both results are interesting to discuss. The mesh
was in both cases performed in a very similar fashion and the only difference is the
number of cells, which were 2 million and 4 million. The difference in cell size is
therefore two times in volume, but regarding the length of one cell this results in a
difference of 21/3which is hard to visualize graphically. The first possible reason for
the difference in result between the two meshes is that the coarser mesh is simply
not fine enough and hence the faulty results are found. Another one could be that
due to the size of the mesh, the coarser one ran for a longer time. Hence the flow
could be picked up at a later time step. There is a possibility that the final mesh for
configuration (a)has not yet reached the time where the vortices occur close behind
the mast. Perhaps the most likely reason is that due to the perfect case of laminar
flow that hits the mast, the finer mesh resolves this very well and a large wake is
created that never gets close to the tips. The coarser mesh on the other hand does
not solve the flow as well and there is an error that causes the vorticity to occur
close to the mast. Interesting to note here is that to some extent the reason for
the vortices acting close behind the mast is of less importance. By using the result
from the coarser mesh we can discuss what would happen if the vortices actually
4. Results and Discussion
did effect the tips.
In section 4.5 it was shown that the initial results with a coarse mesh gave a higher
SPL than the final results. Here the correlation between SPL and vorticity could be
studied. The fact that more noise is created from the mast when there is a higher
level of vorticity around the tips is of interest. Once again it is of great importance
to remember that these results are not fully verified. But apart from the coarser
mesh, there is nothing intending that these results are untrustworthy. These results
would imply that the theory as well as the statement from Seldén are verified. That
mast configuration (a)is indeed creating a larger noise than configuration (b). This
can be seen if comparing Figure 4.7 and Figure 4.2. Presuming that this is correct
it is of interest to look at Figure 4.6b and compare that to Figure 4.8a. It seems
that in both cases the noise is generated by the vorticity created by the tips. It is
then interesting to note that configuration (a)seems to generate more noise. Hence
one could state that the design of mast configuration (b)helps to reduce the noise
From Section 2.1 it is know that sharp edges in general generate noise. This is due
to the sudden blend of fluids at different pressure levels. If the Figures 4.6b and
4.8a are compared it is possible to see that there is vorticity around the tips in
both cases. Note that the tips that are referred to in configuration (b)are the two
enclosing the smaller hole for an extra sail. By comparing the sound level of these
two cases, it can be seen that configuration (a)generates a higher noise. The main
difference in the tip geometry is that in configuration (a)the distance between the
tips is much larger than in configuration (b). From this one could claim that the
amount of space after the edge would effect the capacity of noise that can be created.
Another interesting aspect to investigate is to regard the area inside the mast, where
the two tips create an opening, as a resonance box. In music instruments, such as
guitars, the strings create the vibration and with the help of the resonance box the
tune is created. The empty space within the mast could possibly work as a resonance
box. This would imply that it would be of interest to have as little empty space as
possible within the mast in order to reduce the creation of noise. An example of
noise that is created in a similar fashion is when you blow into a bottle. The more
liquid that the bottle contains, the higher frequency the noise has. This is an effect
of the amount of open space in the bottle. A full bottle barely produces any noise.
This would therefore imply that it is of interest to make a sail, which when it is
folded inside the mast, will fill out the entire cavity all through the mast.
Since numerical methods have been used in this project there is always a chance that
there are some errors in the simulation set up. Especially so since the time frame
was short in regards to the prior knowledge within aero-acoustics of the group. Un-
steady 3D computations are demanding and require large resources, therefore the
computations were time consuming. It would for instance have been desirable to use
a larger domain than the one used in these simulations. Especially in the span-wise
direction, where the domain had to be limited in order to attain a number of cells
4. Results and Discussion
that was low enough. The mesh itself may have been too coarse to fully resolve
the correct aero-acoustic features. As a part of the simulations there was an aim
to conduct an extensive mesh dependency study. This was partly abandoned due
to difficulties in generating a mesh with sufficiently low number of cells. Thus the
mesh generation process became very time consuming and valuable time was taken
away from the simulations and post processing. In the start-up of the simulations
large convergence problems occurred. This was mainly due to difficulties when set-
ting up the physics for capturing the acoustics. Various boundary conditions and
properties were therefore tested in order to acquire a converged solution. It should
be emphasized that none of the group members had any previous experience in the
aero-acoustic field.
The project goal was to explain the effects of the tip geometry in a mast regarding
noise generation. It was concluded that the two mast geometries indeed do generate
noise. However, from the case examined it cannot be verified that any of the two
configurations creates a tonal noise. The results show that the vortex shedding was
mainly caused by the bluff mast body itself and not the tips. Aero-acoustic theory
suggests that the whistling should occur due to vortex shedding caused by sharp
edges. Since the vorticity is low in the tips-region, no major whistling should be
generated. It is the believe of the authors that other yaw angles might generate a
whistling noise caused by the tips. But as for as now, one could not determine the
root cause of the whistling noise.
During the project the experienced tonal noise could not be reproduced. No rec-
ommendations can therefore be made with certainty to Seldén Mast based on the
results. However some interesting aspects regarding the flow around the mast have
been identified and analyzed.
From this study alone, it cannot be concluded that mast (a)in general generates a
higher tonal noise than mast (b). In order to fully understand the noise generation
features it is thus recommended to conduct a more comprehensive investigation. A
profound study should include several yaw angles and varying free stream velocities.
A more extensive mesh dependency study would also be of interest. This is to ensure
that the results are reliable enough.
[1] Junker Miranda, U. (2009) Bonniers Uppslagsbok.©Bonnier Fakta AB.
[2] Goldsmith, . (2012).DISCORD: The story of noise Oxford University Press.
©Mike Goldsmith 2012.
[3] Albrecht Wagner, C. Hütti, T. Sagaut, P (2007).Large Eddy Simulation for
Acoustics.©Cambridge University Press 2007.
[4] Anselmet, P. Mattei, P.O. (2016).Acoustics, Aeroacoustics and vibrations
©ISTE Ltd 2016.
[5] D. Olivari. (2012), Von Kármán vortex shedding, Retrieved from
[6] Powell A.,Why Do Vortices Generate Sound?, ASME. J. Mech. Des.
1995;117(B):252-260. doi:10.1115/1.2836464.
[7] Lars Davidson, (2017) Fluid mechanics, turbulent flow and turbulence modeling.
[8] Lighthill, M. J. (1952). On sound generated aerodynamically. I. general the-
ory. Proceedings of the Royal Society of London. Series A, Mathematical and
Physical Sciences, 211(1107), 564-587.
[9] Kaltenbacher, M., Escobar, M., Becker, S., and Ali, I. (2010;2009;). Numerical
simulation of flow-induced noise using LES/SAS and Lighthill’s acoustic anal-
ogy. International Journal for Numerical Methods in Fluids, 63(9), 1103-1122.
[10] Axel Kierkegaard (2016). Best Practices for Direct Noise Calculations.
[11] Hua-Dong Yao Lars Davidson Lars-Erik Eriksson Shia-Hui Peng Olof Grun-
destam Peter E. Eliasson (2014), Surface integral analogy approaches for pre-
dicting noise from 3D high-lift low-noise wings, pp. 326-338.
[12] F.J. McQuillan, J.R. Culham and M.M. Yovanovich (1984). PROPERTIES
OF DRY AIR AT ONE ATMOSPHERE. Microelectronics Heat Transfer Lab
University of Waterloo Waterloo, Ontario
[13] Dong, Y., Yan, Y. & Liu, C. (2016), New visualization method for vortex struc-
ture in turbulence by lambda2 and vortex filaments, APPLIED MATHEMATI-
CAL MODELLING, vol. 40, no. 1, pp. 500-509.
[14] J. Jeong, F. Hussain (1995),On the identification of a vortex, J. Fluid Mech.
285 69–94.
ResearchGate has not been able to resolve any citations for this publication.
Full-text available
A non-invasive method to determine epidermal thickness from the intensity of transmitted visible light is proposed. A simple model for passing light through the epidermis layers and then measuring what is transmitted back out of the surface, is implemented by using the Montecarlo method. Simulations revealed that the transmittance is linearly dependent on the thickness of the epidermis and so demonstrate the viability of the proposed method.
The incident direction of earthquake waves has significance on the seismic responses of long lined tunnels. In this study, based on the equivalent node force method and together with the viscous-spring artificial boundary, the 3-D input method of SV and SH-waves with arbitrary incident angles is studied firstly. The input method is introduced within the framework of ABAQUS software and is verified by two numerical examples. Consequently, the proposed input method is employed to investigate the impact of the incident angles of SV and SH-waves on the seismic responses of a long lined tunnel. The numerical results indicate that the non-linear seismic responses of long lined tunnels are highly affected by the incident angles of S-waves. As the incident directions of SV and SH-waves being not perpendicular to the tunnel longitude, non-uniform motions appear along the tunnel longitude, as well as the longitudinal internal forces. The plastic deformation and strain components along the tunnel cross-section are also influenced by the incident angles of SV and SH-waves. It is thus important to include the incident angles of earthquake waves in mapping the seismic risk of tunnels.
High frequency mechanical impact treatment is observed to increase the fatigue strength of welded joints. This technique induces compressive residual stresses, increases the local hardness, and reduces the stress concentration by modifying the weld toe radius. The goal of this study was to investigate residual stresses induced by ultrasonic impact treatment in S355, S700MC, and S960 grades steel experimentally and numerically. Plate specimens were manufactured and treated with different treatment intensities i.e. vibration amplitudes of the Sonotrode. The indentation depths were measured by the aid of a laser scanner and residual stresses using X-ray diffraction technique. The effect of steel grade and treatment intensity on the induced compressive residual stress state was firstly studied experimentally. In addition, displacement controlled simulations were carried out to estimate the local residual stress condition considering the effect of different material models. Both the numerically estimated and experimentally measured residual stresses were qualitatively in good agreement. Residual stress state in S355 and S700MC can be estimated well using combined strain rate dependent material model. No significant effect of the treatment intensity is observed on the indentation depth and residual stress state for S355 grade steel. The indentation depth decreases with the increase in the yield strength of the steel.
Vibration attenuation performance of vibration reducing (VR) gloves is, invariably, measured at the palm using the method described in ISO 10819 (2013). The standardized method, however, does not address the glove performance in terms of fingers vibration, which is substantially different from that of the palm. An investigation was conducted to enhance the understanding of fingers vibration transmission properties of 4 different types of VR gloves (air, gel, hybrid and leather) using 4 male subjects in the laboratory. Three different finger adapters, each containing a miniature three-axis accelerometer, were designed and assessed for measurements of vibration responses at the fingers of the gloved hand. Vibration responses of the VR gloves were measured at the middle phalanges of the index and middle fingers using two Velcro finger adapters under the standardized vibration spectra and the spectra of 3 different hand tools, namely a road breaker, a nutrunner and an orbital sander. Vibration transmissibility of the gloves was also measured at the palm using the standardized palm adapter. The frequency response functions (FRF) obtained for the fingers of the gloved hand were used to estimate vibration transmission performances of the gloves at the index and middle fingers. The results show that only the air and hybrid VR gloves provide some attenuation of vibration transmitted to the index finger compared to the bare hand, while nearly all the gloves amplify vibration of the middle finger. The vibration attenuation was evident at the palm for the air and hybrid gloves above 40 Hz. The gel and leather gloves revealed nearly unity transmissibility in most of the frequency range. Only the hybrid glove passed the criteria of ISO 10819 (2013) to be designated as an anti-vibration (AV) glove, although it amplified vibration transmitted to the middle finger, particularly in the H-frequency range (200–1250 Hz). Fairly good agreements (−11%–13%) were obtained between the measured and estimated fingers vibration responses for all the gloves using the FRF method under the vibration spectra of the selected tools.
The effectiveness of high-frequency mechanical impact (HFMI) is considered to rely on the existence of compressive residual stresses. To determine when residual stress relaxation occurs, and what the resulting influence on fatigue improvement is, local stress-strain response in as-welded and HFMI-treated weld toes was modelled under different peak stress conditions. Then, effective notch stress analysis was used to correlate these results with available experimental observations. The simulations showed that high stress ratios and compressive peak stresses were critical with respect to residual stress relaxation, as expected. A compressive peak stress of 0.6fy (nominal yield strength) resulted in full residual stress relaxation. The relative fatigue damage calculations and the notch stress analysis indicated, however, that fatigue improvement could be expected even after significant residual stress relaxation. Based on this and previously observed benefit for high stress ratios, an increase in maximum allowable stresses for HFMI-treated welded steel joints is suggested. The maximum stress ratio is proposed to be increased from R = 0.52 to R = 0.7, and the maximum stress range to limit compressive stresses is proposed to be increased from ΔSmax = 0.9fy to ΔSmax = 1.2fy, which corresponds to Smin = −0.6fy for stress ratio R = −1.
Fatigue improvement from high-frequency mechanical impact (HFMI) is considered to rely on compressive residual stresses, improved weld toe geometry and localized strain hardening at the weld toe. Of these, the existence of compressive residual stresses is generally assumed critical for the effectiveness of the method. In this study, the influence of stress ratio and peak loads on residual stress relaxation and fatigue damage in as-welded and HFMI-treated S700 transverse attachments was investigated. Elastic-plastic stress-strain response for as-welded and HFMI-treated conditions was simulated considering the effects of initial residual stress distribution, local geometry and local material properties. Relative fatigue damage was estimated using the Smith-Watson-Topper parameter. Full residual stress relaxation was observed for a stress ratio of 0.5 and a compressive overload of 0.6 times the nominal yield strength. The fatigue damage assessment showed benefit from HFMI with respect to the as-welded state for all simulated load conditions. The results are consistent with experimentally observed behaviour. The stepwise analysis indicated that the remaining benefit after residual stress relaxation was due to geometry improvement and strain hardening.
This paper presents the strain rate behaviour in tension of the S355 structural steel using a modified Split Hopkinson Tensile Bar for the mechanical characterisation at high strain rates (300 , 500 and 850 ), and a Hydro Pneumatic Machine for intermediate strain rates (5 and 25 ). These data are collected with the intention of setting down the basis for the enhancement of a progressive collapse analysis, that is, by the way, a dynamic event. Results show that the structural steel S355 is strain rate sensitive, keeping its strain hardening capacity with increasing strain-rates. Also the strain energy and the ductility show a marked influence to increasing strain rates. Two of the most used constitutive laws (Johnson–Cook and Cowper–Symonds) are considered and the accuracy of these models is demonstrated. The results can be of great interest for the assessment of robustness in structures where progressive collapse may be triggered by the failure of parallel steel members like, e.g., in the case of cable-stayed or suspension bridges and of large-span 3D trusses subjected to high strain-rate events.
Post-treatment techniques like the high frequency mechanical impact treatment (HFMI) exhibit a significant fatigue life enhancement of welded joints. The effectiveness of this mechanical impact treatment is primarily based on the combination of three effects: the local hardness increase, compressive residual stresses and reduced notch stress concentration at the weld toe. The goal of the present study was to develop a computationally efficient approach for predicting residual stresses induced by the HFMI process on steel specimens. For that purpose, explicit simulations of this post weld treatment technique were performed utilizing the software package ABAQUS. Although, the focus of this study is to find suitable process and material parameters as input for the numerical simulation. For this, the impact velocity, contact force and permanent indentation depth of a pneumatic HFMI-tool were measured. Concerning material modelling, an enhanced combined material model with strain rate dependency was applied. Furthermore, the simulated residual stress field was experimentally validated by X-ray diffraction and neutron diffraction residual stress measurement. The results of the simulations are in good agreement with the experimental results, showing that the material hardening model used for simulation has a high influence on the calculated residual stress values.
Post-treatment methods enhance the fatigue strength of welded high-strength steel joints significantly. In industrial applications, one commonly applied technique is the high frequency mechanical impact (HFMI) treatment. Attained local benefits increasing fatigue strength are the compressive residual stress state, the reduced notch effect at the weld toe, and additionally the local work hardening of the material. This paper presents the set-up of a closed simulation loop including structural weld simulation, numerical computation of the HFMI-process, and a numerical evaluation of the local fatigue life. A thermo-mechanical coupled weld process simulation using Sysweld is built-up to assess the course of residual stress and strain due to welding. The resulting local material behavior is transferred as mechanical cards to the solver Abaqus for the subsequent numerical simulation of the HFMI-process. Hence, major influences such as transient change of material parameters during welding and subsequent cool-down, effect of process dependent clamping conditions and changing contact characteristics are considered. Finally, a numerical evaluation of the local fatigue behavior by the local stress/strain approach and by crack propagation is performed by the aid of the software packages Femfat, nCode and Franc2D. A comparison of the simulated fatigue life with experimental test results proofs their basic applicability; but also numerical limitations of the presented simulation tools are determined. The major benefit of the established simulation chain is the opportunity to study different weld process and HFMI-treatment parameters in regard to fatigue strength without the need of comparably expensive fatigue tests.