Eth 7837 02 PDF

Diss. ETH No.
21645
Influence of the Fluid-Structure Interaction

on the Vibrations of Structures
A dissertation submitted to
ETH ZURICH
for the degree of
Doctor of Sciences
presented by
JOHANNES A. N. HENGSTLER
Dipl.-Ing., University of Stuttgart
born April 3rd 1982
citizen of Germany
accepted on the recommendation of

Prof. Dr. J. Dual, examiner
Dr. M. Sick, co-examiner
Prof. Dr. T. Staubli, co-examiner
2013
ii
Contents
Contents iii
Abstract v
Zusammenfassung vii
List of Figures ix
List of Tables xi
1 Introduction 1
1.1 Background and Motivation . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Fluid-Structure-Interaction of Water Turbines: Previous Research . . . . 2
1.3 Scope and Goal of the Present Work . . . . . . . . . . . . . . . . . . . . 3
1.4 Outline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2 System Description, Theory and Working Principle 7

2.1 Fluid-Structure Interaction . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.1.1 Compressible Fluid . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.1.2 Incompressible Fluid . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.2 Modal Testing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.2.1 Experiment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.2.2 Finite Element Simulation . . . . . . . . . . . . . . . . . . . . . . 19
2.3 Coupling Effects Caused by the Fluid-Structure Interaction . . . . . . . . 22
3 Rectangular Plate 33
3.1 Plate without clamping . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
3.1.1 Eigenfrequency in Air . . . . . . . . . . . . . . . . . . . . . . . . 35
3.1.2 Eigenfrequency in Water . . . . . . . . . . . . . . . . . . . . . . . 36
3.2 One- and Two-Sided Clamped Plate . . . . . . . . . . . . . . . . . . . . . 41
3.2.1 Eigenfrequency in Air . . . . . . . . . . . . . . . . . . . . . . . . 41
3.2.2 Eigenfrequency in Water . . . . . . . . . . . . . . . . . . . . . . . 42
4 Beam 45
4.1 Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
4.2 Material Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
iii
4.3 Eigenfrequency and Damping in Water . . . . . . . . . . . . . . . . . . . 46
4.3.1 Experimental setup . . . . . . . . . . . . . . . . . . . . . . . . . . 46
4.3.2 Two-Way Coupling Simulation . . . . . . . . . . . . . . . . . . . 47
4.3.3 Comparison of Experimental and Simulated Results . . . . . . . . 50
5 Circular Plate 55
5.1 Basics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
5.1.1 Imperfections . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
5.1.2 Mode Shapes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
5.2 Rotating Mode Shapes . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
5.3 Submerged Circular Plate . . . . . . . . . . . . . . . . . . . . . . . . . . 59
6 Experimental Rig 63
6.1 Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
6.2 Tests with Partially Filled Rig . . . . . . . . . . . . . . . . . . . . . . . . 64
6.2.1 Eigenfrequency and Damping in Air and Water . . . . . . . . . . 65
6.2.2 Influence of the Water-level . . . . . . . . . . . . . . . . . . . . . 68
6.3 Tests with Brimful Rig . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
6.3.1 Influence of the Water-level . . . . . . . . . . . . . . . . . . . . . 72
6.3.2 Influence of a Circular Water Flow . . . . . . . . . . . . . . . . . 72
7 Conclusions and Outlook 83

7.1 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
7.2 Outlook . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
Appendix A Photos of Experimental Setups Utilized for the Present Work 85

A.1 Experimental Setup for the Rectangular Plate: Photos . . . . . . . . . . 85
A.2 Experimental Setup for the Beam: Photos . . . . . . . . . . . . . . . . . 87
A.3 Experimental Setup for the Circular Plate: Photos . . . . . . . . . . . . . 88
A.4 Experimental Rig: Photos . . . . . . . . . . . . . . . . . . . . . . . . . . 89
References 93
iv
Abstract
All structures are vibrating which can be desired or undesired. For industrial applica-
tions, it is important to know the vibration spectra of a product already in the design
phase. While designing mechanical structures, design criteria and accurate simulation
tools are needed. Unfortunately it is often not enough, to study the dynamics of a struc-
ture itself because of the influence of a surrounding fluid on the dynamics. As soon as
the surrounding fluid influences the vibration behavior of a structure, the design process
becomes more difficult.
The aim of this thesis is to give a deep insight into the influence of coupling effects on
the dynamics of structures surrounded by a fluid. In view of the applications in water
turbines, the fluid is assumed to be water. The fluid influences the eigenfrequency and
damping of the structures. Nearby walls, a free water surface, and a flowing fluid are
boundary conditions which are considered here. By means of numerical simulations, the
eigenfrequencies and the damping factors are predicted and validated with experimental
data.
The eigenfrequencies of different structures like beams, rectangular plates, and circular
plates are investigated experimentally and compared to the results of literature values,
analytical calculations, and numerical finite element simulations. The structures are
immersed in water and excited with an electromagnetic coil to measure the frequency
response with a laser scanning vibrometer system. Structures have in water a lower
eigenfrequency and a higher damping than in air. Coupling effects like the presence
of nearby walls were analyzed in detail by varying the boundary conditions of different
plates. Nearby walls or a free water surface significantly influence the eigenfrequency of
a structure. Compared to the eigenfrequency of a structure in an infinite water domain,
a nearby wall lowers the eigenfrequency even further, while a free water surface increases
the eigenfrequency. The coupling effects which are responsible for the influence on the
eigenfrequency are explained by using a model with two identical circular plates with
water in between.
For the finite element simulations of the eigenfrequencies of the structures in water, the
commercial software ANSYS is used. The influence of the water on the eigenfrequency
of the structures was approximated by using acoustic elements. A typical application
of these acoustic elements is the dynamics of submerged structures. The governing
equation therefore is the wave equation.
v
Damping values are experimentally obtained from structures in air and in water. The
damping of plates increases for higher eigenfrequencies. Nearby walls increase the damp-
ing further. A circular water flow around a circular plate increases the damping addi-
tionally. Due to the computational effort, only damping values of experiments with an
immersed beam are compared with simulations. A numerical two-way coupling simula-
tion between computational fluid dynamics and structural analysis in the time domain
using the commercial finite element simulation software ANSYS is performed. For the
beam, the damping values of the first three bending modes are simulated and compared
with experiments.
When a fluid flows in a circular fashion near a circular plate, a new phenomenon has
been discovered. The eigenfrequencies of a circular plate split in the presence of a cir-
cular water flow near its surface. A circular plate is tested in an experimental rig where
an electric motor set the water coaxial to the circular plate in rotation. For different
rotational speeds the frequency response is measured with a laser scanning vibrometer
system. The split of the eigenfrequencies has an upward trend with an increasing ro-
tational speed. For a fluid-conveying pipe with periodic boundary conditions a similar
phenomenon exists. There, an analytical model is able to predict the split.
vi
Zusammenfassung
Alle Strukturen schwingen, ob erwunscht oder unerwunscht. Fur die industrielle An-
wendung ist es wichtig, das Schwingungsverhalten eines Produktes bereits in der Ent-
wicklungsphase genau zu kennen. Bei der Entwicklung von mechanischen Bauteilen sind
Auslegungsrichtlinien und genaue numerische Simulationen unerlasslich. Unglucklicher-
weise ist es oft nicht genug, das Schwingungsverhalten eines Bauteils fur sich alleine zu
betrachten, da dieses vom umgebenden Fluid beeinflusst werden kann. Und sobald das
Schwingungsverhalten eines Bauteils vom umgebenden Fluid beeinflusst wird, wird die
Produktentwicklung schwieriger.
Das Ziel dieser Arbeit ist ein tiefer Einblick in die Kopplungseffekte, die das Schwin-
gungsverhalten von Strukturen, die von Fluid umgeben sind, beeinflussen. Denn ein
umgebendes Fluid beeinflusst die Eigenfrequenz und die Dampfung von Strukturen. Im
Hinblick auf die Anwendung bei Wasserturbinen, wird als Fluid Wasser betrachtet. Nahe
Wande, eine freie Wasseroberflache und stromendes Fluid sind Randbedingungen, die
hier berucksichtigt werden. Mit numerischen Simulationen werden die Eigenfrequenzen
und die Dampfung vorhergesagt und ausserdem mit experimentellen Werten bestatigt.
Die Eigenfrequenz von verschiedenen Strukturen wie Balken, rechteckigen Platten und
Kreisplatten werden experimentell untersucht und mit Werten aus der Literatur, ana-
lytischen Berechnungen und numerischen finite Elemente Simulationen verglichen. Die
Strukturen sind dabei in Wasser und werden mit einer elektromagnetischen Spule ange-
regt, um die Frequenzantwort mit einem Laser Scanning Vibrometer System zu messen.
Strukturen haben in Wasser eine tiefere Eigenfrequenz und eine hohere Dampfung als in
Luft. Kopplungseffekte wie nahe Wande werden genauer untersucht, indem die Randbe-
dingungen fur verschiedene Strukturen verandert werden. Nahe Wande oder eine freie
Wasseroberflache haben einen starken Einfluss auf die Eigenfrequenz. Beim Vergleich der
Eigenfrequenz einer Struktur, die sich in einem mit Wasser gefullten, unendlichen Raum
befindet, ist die Eigenfrequenz bei einer nahen Wand noch tiefer und bei einer freien
Wasseroberflache hoher. Die Kopplungseffekte, die fur den Einfluss auf die Eigenfre-
quenz verantwortlich sind, werden anhand eines Modells mit zwei gleichen Kreisplatten
zwischen denen sich Wasser befindet, erklart.
Fur die numerische finite Elemente Simulation der Eigenfrequenzen von Strukturen in
Wasser wird die kommerziell verfugbare Software ANSYS verwendet. Der Einfluss des
Wassers auf die Eigenfrequenz der Strukturen wird mit Hilfe von akustischen Elementen
vii
angenahert. Eine typische Anwendung der akustischen Elemente ist die Dynamik von
Strukturen, die von Fluid umgeben sind. Die zu Grunde liegende Gleichung dabei ist die
Wellengleichung.
Dampfungswerte von Platten in Luft und in Wasser werden aus Experimenten be-
stimmt. Die Dampfung von Platten steigt fur Eigenfrequenzen hoherer Ordnung. Be-
dingt Durch nahe Wande steigt die Dampfung weiter. Rotierendes Wasser um eine
Kreisplatte erhoht die Dampfung zusatzlich. Bedingt durch den rechnerischen Aufwand
konnte nur die Dampfung eines Balkens, der von Wasser umgeben ist, simuliert wer-
den und dann mit den experimentellen Werten verglichen werden. Mit Hilfe der Si-
mulationssoftware ANSYS wurde eine numerische Zwei-Wege Kopplung zwischen einer
Stromungsmechanik-Simulation und einer Struktur-Simulation im Zeitbereich durch-
gefuhrt. Die Dampfungswerte der ersten drei Biegemoden des Balkens wurden simuliert
und anschliessend mit den experimentellen Werten verglichen.
Ein neues Phanomen wurde entdeckt, wenn Wasser in der Nahe einer Kreisplatte ro-
tatorisch stromt. Die Eigenfrequenzen einer Kreisplatte, die von rotierendem Wasser
umgeben ist, teilen sich auf. Dafur wurde eine Kreisplatte untersucht, die sich innerhalb
eines Versuchstandes befindet, bei dem das Wasser um die Platte herum gleichachsig
mit Hilfe eines Elektromotors in Rotation versetzt werden kann. Die Schwingungsant-
wort wurde hierfur bei verschiedenen Rotationsgeschwindigkeiten mit Hilfe eines Laser
Scanning Vibrometer Systems gemessen. Die Teilung der Eigenfrequenzen wird grosser
wenn auch die Rotationsgeschwindigkeit grosser wird. Fur ein mit Wasser durchstromtes
Rohr mit periodischen Randbedingungen existiert ein ahnliches Phanomen. Ein analy-
tisches Model kann diese Teilung vorhersagen.
viii
List of Figures
1.1 Undershot and overshot water wheel . . . . . . . . . . . . . . . . . . . . 1
1.2 Fourneyron turbine . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
2.1 Detailed view of an electromagnetic coil and an EMAT . . . . . . . . . . 13

2.2 Schematic diagram of a heterodyne interferometer . . . . . . . . . . . . . 14
2.3 Laser scanning vibrometer setup . . . . . . . . . . . . . . . . . . . . . . . 15
2.4 The signal of a Periodic Chirp in the time domain . . . . . . . . . . . . . 16
2.5 The signal of a Periodic Chirp in the frequency domain . . . . . . . . . . 16
2.6 Vibration measurement setup with a Lock-In amplifier . . . . . . . . . . 17
2.7 Schematic diagram of a Lock-In amplifier . . . . . . . . . . . . . . . . . . 18
2.8 Simulation of two CF circular plates . . . . . . . . . . . . . . . . . . . . 23
2.9 Four designs for the investigation of two identical CF circular plates . . . 25
2.10 The model of the coupling effects caused by the fluid-structure interaction 26
2.11 The mesh of the coupling effects caused by the fluid-structure interaction 27
2.12 Eigenfrequencies of the CF circular plates against different modes . . . . 28
2.13 Eigenfrequencies of the CF circular plates against different distances . . . 29
2.14 Simulation results of two CF circular plates in a radial symmetry plane
for 1-0 and 2-0 mode . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
2.15 Simulation results of two CF circular plates in a radial symmetry plane
for 5-0 and 6-0 mode . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
3.1 Dimensions and model of the FFFF rectangular plate . . . . . . . . . . . 34

3.2 Mode shapes of a FFFF rectangular plate . . . . . . . . . . . . . . . . . 35
3.3 Experimental amplitude spectrum of the FFFF rectangular plate in air . 36
3.4 The mesh of the FFFF rectangular plate . . . . . . . . . . . . . . . . . . 37
3.5 Experimental amplitude spectrum of the FFFF rectangular plate in water 38
3.6 Eigenfrequency of some modes for the FFFF rectangular plate . . . . . . 40
3.7 CFFF and CFCF rectangular plate and its dimensions . . . . . . . . . . 41
3.8 Mode shapes of a CFFF and CFCF rectangular plate . . . . . . . . . . . 42
3.9 Eigenfrequencies of a CFFF and CFCF rectangular plate in water . . . . 43
4.1 Beam and its dimensions . . . . . . . . . . . . . . . . . . . . . . . . . . . 45

4.2 The model of the cantilever beam . . . . . . . . . . . . . . . . . . . . . . 47
4.3 The mesh of the cantilever beam . . . . . . . . . . . . . . . . . . . . . . 48
4.4 Transient simulation of the beam . . . . . . . . . . . . . . . . . . . . . . 49
4.5 Squared residuals of the transient simulation . . . . . . . . . . . . . . . . 50
ix
5.1 Mode shapes of a FC circular plate . . . . . . . . . . . . . . . . . . . . . 58
5.2 Standing and rotating mode shape . . . . . . . . . . . . . . . . . . . . . 61
6.1 Experimental rig - technical drawing in 2D . . . . . . . . . . . . . . . . . 64

6.2 Schematic detailed view of the top part of the experimental rig. . . . . . 65
6.3 Dimensions and model of the FC vibrating plate . . . . . . . . . . . . . . 66
6.4 The mesh of the FC vibrating plate from the experimental rig . . . . . . 68
6.5 Eigenfrequency in water of the vibrating plate for different water-levels . 70
6.6 Damping in water of the vibrating plate for different water-levels . . . . . 71
6.7 Bode plot for the 2-0 mode with rotating water . . . . . . . . . . . . . . 73
6.8 Bode plot for the 3-0 and 4-0 mode with rotating water . . . . . . . . . . 74
6.9 Bode plot for the 5-0 and 6-0 mode with rotating water . . . . . . . . . . 75
6.10 Resonance frequency and damping with rotating water . . . . . . . . . . 76
6.11 A deflected fluid-conveying pipe with pinned ends . . . . . . . . . . . . . 78
6.12 Analytical solution of a fluid-conveying pipe for pinned boundaries . . . . 80
6.13 Analytical solution of a fluid-conveying pipe for periodic boundaries . . . 81
A.1 Detailed view of the experimental setup of the FFFF plate . . . . . . . . 85

A.2 Experimental setup for the vibration measurement of the FFFF plate . . 86
A.3 Experimental setup of the beam with a coil . . . . . . . . . . . . . . . . . 87
A.4 Experimental setup of the beam with an EMAT . . . . . . . . . . . . . . 87
A.5 Front view of the FC circular plate . . . . . . . . . . . . . . . . . . . . . 88
A.6 Top view of the FC circular plate . . . . . . . . . . . . . . . . . . . . . . 88
A.7 Detailed view of the glued magnet and one of the coils . . . . . . . . . . 89
A.8 View of the main part of the experimental rig . . . . . . . . . . . . . . . 90
A.9 Side view of the complete experimental rig . . . . . . . . . . . . . . . . . 91
A.10 Detailed view of the top of the experimental rig . . . . . . . . . . . . . . 92
x
List of Tables
2.1 Eigenfrequencies of two CF circular plates with water in between . . . . 25
3.1 Blevins table . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36

3.2 Eigenfrequencies in air of the FFFF rectangular plate . . . . . . . . . . . 37
3.3 Eigenfrequencies in water of the FFFF rectangular plate . . . . . . . . . 38
3.4 Eigenfrequencies in air of the CFFF and CFCF rectangular plate . . . . 42
3.5 Eigenfrequencies in water of the CFFF and CFCF rectangular plate . . . 42
4.1 Material properties of the beams . . . . . . . . . . . . . . . . . . . . . . 46

4.2 Eigenfrequency in air and water of the beams . . . . . . . . . . . . . . . 51
4.3 Damping in air and water of the beams . . . . . . . . . . . . . . . . . . . 52
6.1 Eigenfrequencies in air and water of the vibrating plate . . . . . . . . . . 67

6.2 Damping in air and water of the vibrating plate . . . . . . . . . . . . . . 67
6.3 Material properties of the pipe and the fluid . . . . . . . . . . . . . . . . 79
xi
xii
1 Introduction
1.1 Background and Motivation
In 1786 Du Buat [13] studied the oscillations of spheres in water and found out that
their vibrational behavior is influenced by the water. Since then, significant research
has been done to describe the vibrational behavior of immersed structures which are in
contact with fluids.
One important branch of this research is related to water turbines. The water wheel (fig-
ure 1.1) was invented in the third century BC, and it was not until 1826 that Fourneyron
invented a water turbine (figure 1.2) with a much higher efficiency than the water wheel.
Typically only some blades of a water wheel are in contact with water and drive the
Figure 1.1: Undershot and overshot water wheel. At the undershot water wheel, the
water flows below the wheel. It is the oldest type of wheel. At the overshot water wheel,
the water flows in a tube or flume above the wheel up to the top of the wheel. (source:
Wikipedia)
wheel. The advantage of the Fourneyron turbine was that all blades are continuously
in contact with the water to drive the wheel which leads to a higher efficiency. For his
design, he used an axial inlet and a radial outlet for the water. This design is known as
a radial runner. In 1849 Francis improved this design to the today widely used Francis
1
Chapter 1. Introduction
Figure 1.2: Side (left) and top view (right) of a Fourneyron turbine with its guide vanes
(s) and rotor blades (b). (source: Wikipedia)
turbine. The main difference to the Fourneyron turbine is that at the Francis turbine the
stator is outside and the rotor inside instead of the inverse alignment at the Fourneyron
turbine.
Today water turbines have become an essential part of our electrical power supply.
Francis turbines are of particular importance in stabilizing our power grid by providing
pumped-storage hydroelectricity. The stabilization is even more important because of
the increase in solar and wind power. Depending on the weather and the time of the
day, the solar and wind power have a big variation in their generation of electricity.
There exist two main design criteria for current radial runners. First they have to be
light. Light and thin walled turbines help to save material costs. Second due to cost effi-
ciency, smaller machines are requested with the same power. Using smaller machines, the
relative dynamic stress from the characteristic rotor-stator interaction increases which
can lead to fatigue cracks. For the design of radial runners, it is important to understand
the dynamics of the fluid, the dynamics of the structure itself, and their interaction. Due
to computational costs the full simulation of the interaction has not been possible in a
reasonable time, until now.
1.2 Fluid-Structure-Interaction of Water Turbines:

Previous Research
The simulation of the dynamics of a water turbine must be separated into two main
fields. The first is the simulation of fluid flow, for which the interested reader can find
2
1.3. Scope and Goal of the Present Work
a review in Keck and Sick [25] covering the past thirty years of research on numerical
flow simulation in hydraulic turbo machines. Second is the simulation of the structure
mechanical response. In this latter case, accurate simulation of the eigenfrequencies of
the turbine structure is important, including effects from the rotor-stator interaction.
In 1984, Dubas [14] presented a method to calculate the excitation frequencies from the
rotor-stator interaction. During the design process the coincidence between the excita-
tion frequencies and an eigenfrequency of the structure must be prevented. Otherwise
large vibrations are produced, resulting in fatigue cracks as described in one example
by Ohashi [36]. In 1997, Tanaka [44] presented an experimental study on a prototype
runner and a model runner. He investigated the frequency response of the runner in-
teracting with the stator for different rotational speeds. He found out empirically that
the eigenfrequency reduction of the runner between air and water is greater if a stator
is present instead of a water domain, which is much larger than the runner. Later,
Rodriguez et al. [37] and Liang et al. [32] made numerical simulations to estimate the
frequency reduction for a runner in a stationary infinite water domain.
1.3 Scope and Goal of the Present Work
For water turbines, it is difficult to get experimental data of the eigenfrequencies and
damping in-situ. Therefore the water turbines are normally tested in a basin with
water, where on five sides are walls and a free water surface at the top. The boundary
conditions have a big influence on the eigenfrequency and damping. Accurate simulations
are needed to estimate the influence of the boundary conditions quantitatively. To get
an accurate simulation tool, validations with experiments are mandatory.
In literature, as shown later, for structures like rectangular and circular plate, experi-
mental data of eigenfrequencies and damping in water exist, but there is still a lack of
experimental data to validate the prediction of the eigenfrequencies by numerical simu-
lations of immersed structures like plates with different boundary conditions like nearby
walls or a free water surface. Also the amount of experimental data of the damping of
immersed structures available in the literature is small.
The goal of this research is to provide experimental data of eigenfrequencies and damping
of different immersed structures. Especially nearby walls and a free water surface for
different distances to the plates are studied. The different boundary conditions are
3
Chapter 1. Introduction
studied independently of each other. Also the effect of a flowing fluid around a circular
plate is investigated.
Afterwards the experimental values are used for the validation of a combined numerical
simulation of the fluid and the structure. This has been done in this thesis for plates in a
structural simulation with a modal analysis using acoustic elements for the fluid and for
a beam with a transient two-way coupling between computational fluid dynamics (CFD)
and a structural analysis. Through this analysis, the effects of different situations, such
as nearby walls and flowing fluids, is now well understood. Speaking about fluids in this
thesis, always liquids like water are meant.
1.4 Outline
In chapter 2 the theory and working principles are described. In section 2.1 the theoret-
ical basics and formulas for the fluid-structure interaction are given. In section 2.2 the
methods are presented for the experiments and the simulations. In section 2.3 a numer-
ical study is presented to explain the boundary effects of an immersed solid structure.
From this study some basic coupling effects are derived.
In chapter 3 the vibration of a rectangular plate immersed in water is presented. Eigen-
frequencies of the rectangular plate from different experiments are compared with liter-
ature values. Numerical acoustic simulations have been performed to predict the eigen-
frequencies of different rectangular plates. The experimental values are used to validate
the results of the simulation.
In chapter 4 a numerical two-way coupling simulation between the fluid and a beam
is presented with the goal to predict the damping of the immersed beam additional
to the eigenfrequencies. The predicted values from the simulation are compared to
experimental data.
In chapter 5 the vibration characteristics of a circular plate is discussed. The focus is
on the effect of the circular plates periodicity on the eigenfrequency.
In chapter 6 the experimental rig used for further studies on a circular plate is pre-
sented. First the influence of different boundary conditions on the eigenfrequency and
damping are investigated. To predict the eigenfrequencies of the circular plate within
the experimental rig, numerical acoustic simulations are performed which are then com-
pared with the experimental values. Second the influence of a circular water flow on
4
1.4. Outline
the eigenfrequency and damping is investigated. The results are compared with the
literature.
In appendix A the interested reader can find photos of the different experimental setups.
5
6
2 System Description, Theory and
Working Principle
2.1 Fluid-Structure Interaction

A fluid-structure interaction in its general form can be described by the coupling of
the equation of motion of the structure and the equation of motion of the fluid. The
discretized equation of motion [11] of the structure can be written as:
M s u + D s u + K s u = F s (2.1)
where M s is the structural mass matrix, D s the structural damping matrix, K s the
structural stiffness matrix, F s the applied load vector, and u the nodal displacement
vector where the dot denotes the time derivative.
The common equations for the fluid are the Navier-Stokes equations which can be written
in the general form [27] as:

v
+ v v = p + T + f (2.2)
t
where v is the flow velocity vector, t the time, is the fluid density, p is the pressure, T
is the deviatoric stress tensor, and f represents body forces (per unit volume) acting on
the fluid. For the full description of the fluid flow, additional informations are needed
like the equation of state (for example p = c2 ) and the equation for the conservation of
mass, which can be written as:

+ (v) = 0 (2.3)
t
For the Navier-Stokes equations in its general form no general solution exists. Assump-
tions have to be made. This is shown in the following subsections.
7
Chapter 2. System Description, Theory and Working Principle
2.1.1 Compressible Fluid
A set of assumptions, used for example for ideal gases, are:

the fluid is compressible with pressure variations as density changes
the fluid is inviscid which means no viscous dissipation
the flow is irrotational
there is no mean flow of the fluid
the mean density and pressure are uniform throughout the fluid
With these assumptions one can derive the acoustic wave equation
1 2p
2 p = 0 (2.4)
c2 t2
p
where c is the speed of sound in the fluid medium with c = /0 , 0 the mean fluid
density, the bulk modulus of fluid and p the acoustic pressure. This equation is referred
as the lossless wave equation, because of the neglect of the viscous dissipation. For the
fluid-structure interaction, the equations (2.1) and (2.4) have to be combined as shown
for instance by Liang et al. [32]. At the end, the complete set of discretized equations for
the finite element for the fluid-structure interaction problem are written in assembled
form as:
" #( ) " #( ) " #( ) ( )

Ms 0 u Cs 0 u K s K fs u Fs
+ + = (2.5)
M fs M f p 0 Cf p 0 Kf p 0
where the indices f and fs denote the fluid and fluid-structure terms, respectively. Equa-
tion (2.5) is solved in ANSYS using the coupling between acoustic and structural ele-
ments.
Added Mass Effect
To estimate the fluid influence on the structures eigenfrequencies, some mass can be
added to the mass of the structure in equation (2.1). This added mass is approximately
the fluid mass which is displaced by the body motion. Iversen and Balent [22] made in
1951 an overview of the added mass factors published by different authors. But it was
clear that one added mass factor is only valid for one specific case at a specific frequency.
8
2.1. Fluid-Structure Interaction
In 1982 Brennen [10] wrote a review about the added mass and fluid inertial forces. He
had also a look on the effect of a nearby free surface and a nearby solid boundary. For
a nearby free surface, he found that the influence on the eigenfrequency is less than 5 %
for a submerged plate when the ratio of the submergence to the thickness of the plate is
greater than 4. This can be seen as fulfilled also by experiments and simulation presented
later in this thesis. He also observed an increasing influence for lower submergence to
thickness ratios. The increase ends when the submergence is equal or lower than the
thickness of a plate. Then the influence would decrease because less fluid is being
accelerated. As no investigations in this thesis have been done with a submergence
smaller than the thickness of a plate, this statement cannot be verified here. For a
nearby solid boundary, he found that the added mass is increasing progressively while
reducing the distance to the wall. For a narrow gap between the plate and the wall,
the fluid accelerations can be very large. In this thesis this behavior is also shown in
experiments as well as in simulation. Furthermore, for a narrow gap, Brennen postulates,
that the effect of the compressibility of the water on the added mass can be neglected
until the Mach number, which is the ratio of the typical fluid velocity to the velocity
of sound in the fluid, exceeds a value of at least 0.1. However the presence of large
quantities of bubbles in the water could significantly reduce the sonic velocity.
2.1.2 Incompressible Fluid
When the assumptions for deriving the acoustic wave equation are not applicable, other
assumptions like the one that the fluid is incompressible are needed. In this case the
Navier-Stokes equations in their general form can be reduced to

v
+ v v = p + 2 v + f (2.6)
t
where the left side of equation (2.6) is the inertia (per volume) which can be divided
into the unsteady acceleration term v/t and the convective acceleration term v v.
The right side of the equation is the divergence of the stress plus other body forces f .
The divergence of the stress is the sum of the pressure gradient p and the viscosity
2 v with as the dynamic viscosity. Also the equation for the conservation of mass
can be reduced to
v =0 (2.7)
9
For the simulation of the damping provided by the surrounding fluid of a structure it is
necessary to use equation (2.6) instead of the acoustic wave equation (2.4). This is used
in chapter 4.
2.2 Modal Testing
Modal testing is the characterization of the dynamical behavior of a vibratory structure.

The parameters of interest are mainly the eigenfrequency, the mode shape and the
damping. The determination of the parameters can be done by experiments or with a
numerical simulation.
2.2.1 Experiment
A good overview of the different aspects of the experimental modal testing is given by
Ewins [17]. In the following, the methods used in this work are described in more detail.
In principle there are four main parts necessary for a modal testing:
a structure which has to be analyzed (testee)
an excitation mechanism for the testee
a measuring device for the vibration
a tool for the signal processing and analysis.
The support of the testee, the excitation, the vibration measurement, and the analysis
will now be discussed in more detail.
Boundary Conditions
In general, the boundary conditions for a mechanical problem consist in either a pre-
scribed stress or prescribed displacements. For beams and plates in the following the
nomenclature used by Leissa [30] is introduced to classify the different boundary condi-
tions. Two simple boundary conditions for beams and plates applied at the edges are
used in this work:
clamped (C)
free (F)
10
2.2. Modal Testing
For a plate, the abbreviation for the classification is needed for each edge of the plate. As
a rectangular plate has four edges, also four letters are needed for the abbreviation. For
example, a FFFF rectangular plate is a plate where all four edges have free boundary
conditions. As a circular plate has only one edge but an additional support in the center
is possible, the abbreviation has two letters: the first one for the outer edge and the
second one for the center. For example a circular plate which is free at the outer edge
and clamped in the center is a FC circular plate.
The realization of the boundary condition within an experimental setup is sometimes
challenging. One intuitive problematic case is a completely free boundary condition (no
support) because this cannot be implemented in an experiment. Clamping is problematic
too. An approximation of the completely free boundary conditions is the supporting by
very soft springs which can be realized by soft rubber bands. To further minimize the
effect of the support they have to be applied at a location where the mode of interest has a
node. By measuring the six rigid body modes, it can be tested if the support is sufficiently
weak. Theoretically, the rigid body modes should have each an eigenfrequency of 0 Hz.
For the experiment, the support is alright if the eigenfrequencies of the six rigid body
modes are very low compared to the seventh mode. The low eigenfrequencies indicate a
small influence of the support.
For the clamping, it is more difficult to find a supporting base which is rigid enough,
because all structures have a compliance which is unequal to zero. Unfortunately it is
not as easy as for the completely free boundary condition to test if the supporting base
is rigid enough.
Excitation
There are several ways to excite a structure for modal testing. For example, they can
be divided into contact and non-contact methods, or high sensitivity at low and high
frequency. The frequency range of interest in this work is between 10 Hz and 1000 Hz.
This range is given by the eigenfrequency of the lowest modes of the tested structures.
As the lowest modes have the highest amplitudes, for water turbines, they are most
critical for fatigue cracks. To obtain the mode shapes of a structure by sequentially
measuring a large number of points, it is important to have a continuous and repeatable
excitation.
11
A method which is non-contact and optimal in the low frequency range is the excitation
of a plate with an electromagnetic coil where a small magnet is fixed on the structure.
If the structure is ferromagnetic no additional magnet is needed.
The best location for an excitation is the location where the desired mode has a maximal
amplitude which means an antinode. When measuring several modes, it might be im-
possible to find such a location. In this case a location has to be used where the modes
of interest have a least no nodes.
Electromagnetic Coil To excite the testee, a magnet is glued to the testee and a coaxial
coil nearby is driven by an alternating current at a desired frequency (figure 2.1(a)).
Depending on the polarity of the induced magnetic field, the magnet will be attracted
and repelled at the desired frequency. Every time the current is alternating the magnetic
field of the electromagnetic coil also changes the polarity. This induces a voltage which
opposes the applied voltage. The corresponding part of the electrical impedance of
the electromagnetic coil is proportional to the frequency. That is the reason why this
technique is more efficient in the low frequency range.
As the used signal generators in this work are not able to provide the necessary power
up to 10 W to drive the electromagnetic coil, a commercial audio amplifier is used as
an amplifier for the excitation signal coming from the signal generator. Designing the
coil, the resistance of the coil and the reactance at the desired frequency range have to
be considered because of the impedance match of the electromagnetic coil and the used
commercial audio amplifier. The audio amplifier has the highest power at an output
impedance between 4 and 8 .
Electromagnetic Acoustic Transducer (EMAT) If the structure is not ferromagnetic

and one would not want to apply any kind of additive to the structure, an excitation
with an electromagnetic acoustic transducer might be considered (figure 2.1(b)). This
technique is mainly used at ultrasonic frequencies. The advantage to excite a non fer-
romagnetic structure with a non-contact method is used for a comparison measurement
in chapter 4.
The two main parts of an EMAT are a magnet, which is usually a permanent magnet,
and an electromagnetic coil. For a successful excitation, the testee must be electrically
conducting. When the electromagnetic coil is placed near the testee and is supplied
by an alternating current at a desired frequency, an eddy current is induced in the
12
2.2. Modal Testing
(a) electromagnetic coil (b) EMAT
Figure 2.1: Detail view of a beam tip with (a) the glued magnet and the electromagnetic
coil and (b) the EMAT for the excitation.
surface layer of the testee. With a magnetic field which is perpendicular to the eddy
current, a Lorentz force will act on the testee perpendicular to the eddy current as well
as perpendicular to the magnetic field of the magnet. The interested reader will find
more detailed information for example in [12, 45].
Vibration Measurement
As the excitation before, the vibration measurement techniques can be also classified by
contacting or non-contacting methods. The most popular and widely used measurement
method is the one with an accelerometer which is applied directly onto a structure.
Because of the need of measuring a lot of points sequentially to get the mode shape,
measurement methods where a transducer with mass has to be applied at the structure
are not considered in this work.
Laser One of these non-contacting measurement methods is an optical method using a

laser. The Doppler effect is used to measure the velocity of the surface of a structure. For
the measurement, the laser beam from an optical cavity is split up by a beam splitter
into a reference beam and a probe beam. The probe beam will be focused onto the
surface of the testee. An important characteristic of the surface is a good retroreflective
behavior. The retroreflectivity can be improved by applying a thin retroreflective foil
(for example 3M Scotchlite High Gain Reflective Sheeting 7610) on the surface of the
testee.
13
Figure 2.2: Schematic diagram of a heterodyne interferometer with a Bragg cell as an

essential component. The laser has the frequency f0 , the Bragg cell shifts the frequency
with fB , and the Doppler frequency fD is proportional to the velocity of the object.
When the probe beam is reflected by the surface, the frequency of the reflected beam
is changed depending on the velocity of the surface. The reflected laser beam and the
reference beam are then superimposed. The interference of the reference and probe beam
is an intensity modulation with the Doppler frequency as the beat frequency. A detector
measures this beat frequency which is proportional to the velocity of the surface of the
structure. With this method one can identify the vibration of the surface but not its
direction. To overcome this problem a Bragg cell (figure 2.2) is used for the reference
beam which shifts the frequency of the beam f0 by fB = 40 MHz (the frequency of
the laser beam is about f0 = 474 THz). The signal at the detector than has a carrier
frequency fB which is modulated by the Doppler frequency fD . This Doppler frequency
is given by fD = 2v/, for example fD = 3.16 MHz for 1 m/s. With usage of this
Bragg cell also the sensitivity of the system is increased because well known frequency
modulation demodulation techniques can be used. The setup using a Bragg Cell is
called a heterodyne interferometer. The signal from the detector is processed with the
demodulator to a voltage signal which is directly proportional to the velocity.
The system used for this thesis is a Polytec scanning vibrometer PSV 200 from 1997
which was upgraded to a PSV 300 in 2003. The current components of the system are
schematically shown in figure 2.3. The scanning head OFV-055 is a modified single
point head OFV-303 with a Helium Neon Laser with a wavelength of = 633 nm.
The vibrometer controller OFV-3001S has the two velocity demodulators OVD-04 (high
frequency range - up to 1.5 MHz) and OVD-06 (low frequency range - up to 50 kHz).
14
2.2. Modal Testing
Vibration Analysis
Laser Scanning Vibrometer (LSV) The PC system PSV-PC F has an internal data
acquisition board PCI-6111 E (National Instruments) with a 12-bit analog-to-digital
converter which has a maximal bandwidth of 1 MHz. The internal function generator
has a maximal bandwidth of 512 kHz. On the PC the Polytec scanning vibrometer
software version 7.43 is installed. The junction Box VIB-Z-017 is used to connect the
components.
Figure 2.3: Laser scanning vibrometer setup
As all measurements in the following are in the low frequency range below 1000 Hz
only the OVD-06 demodulator was used, typically with a velocity range of 5 mm/s/V.
Because of the maximal number of fast Fourier transform (FFT) lines limited by the
system to 6400, the used frequency bandwidth was adapted for each measurement at
100 Hz, 200 Hz, 500 Hz or 1000 Hz depending on the maximal frequency of interest in
order to maximize the frequency resolution which is the quotient of the bandwidth
and the number of FFT lines. In the software, the parameters for the internal signal
generator were selected. A Periodic Chirp [9] was used as signal form, which is similar
to a frequency sweep. A Periodic Chirp with a frequency range from 0 Hz to 10 Hz and a
length of 4 s is shown in figure 2.4 in the time domain and in figure 2.5 in the frequency
domain. As the Periodic Chirp is periodic in time, leakage is not a problem.
15
1
Signal
1
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 2.2 2.4 2.6 2.8 3 3.2 3.4 3.6 3.8 4
Time [s]
Figure 2.4: The signal of one period of a Periodic Chirp in the time domain. The
length of this Periodic Chirp is 4 s and the frequency range is from 0 Hz to 10 Hz.
102
6
Amplitude
0
0 2 4 6 8 10 12 14 16 18 20
Frequency [Hz]
Figure 2.5: The signal of one period of a Periodic Chirp in the frequency domain. The
length of this Periodic Chirp is 4 s and the frequency range is from 0 Hz to 10 Hz.
The frequency range of the excitation signal was defined by the chosen frequency band-
width. To prevent aliasing the sampling frequency is 2.56 times the chosen bandwidth.
The data evaluation software processes the data from the demodulator and makes a
FFT from the signal which is displayed.
After measurement of many points on the surface, for a chosen frequency the software
calculates and displays an animation of the oscillation mode. With this visualization
one can assign the detected eigenfrequencies to the corresponding mode shape. One can
also see if the nodal lines of the mode shapes are fixed in space or moving which will be
important later on.
Lock-In Amplifier A Lock-In amplifier [43] is a device used to measure the ampli-
tude and phase directly without the need of a FFT. To get a better resolution of the
16
2.2. Modal Testing
amplitude- and phase-spectrum as compared to the vibration analysis of the laser scan-
ning system, vibration analysis has partly been done with the Lock-In amplifier. At
a desired frequency the amplitude and phase difference between the excitation signal
and the sensor signal are measured. The components of the setup are schematically
shown in figure 2.6. The Lock-In amplifier is a Dual Phase Lock-In Amplifier 5210 from
EG&G which belongs nowadays to Ametek Inc. As frequency generator, the Function
Generator DS345 from Stanford Research Systems is used, which excites the structure
and gives the reference for the Lock-In amplifier. On the PC the Software LabVIEW
2009 is installed. For the communication between the components, the GPIB interface
is used.
Figure 2.6: Vibration measurement setup with a Lock-In amplifier
Additionally, one needs an actuator and a sensor to have the full setup to measure the
vibration of a structure. In this thesis, for the actuation of the structure, an electro-
magnetic coil is used. As sensor the scanning head OFV-055 with vibrometer controller
OFV-3001S is used. Also possible is to use a second electromagnetic coil as a sensor.
With a program coded using LabVIEW, the excitation signal is open loop controlled
and the output from the Lock-In amplifier is stored. Typically one chooses a frequency
range of interest, a frequency resolution and a time delay for the sequential measurement
of each frequency of interest specified with the frequency range and the resolution.
17
The measurement routine for one frequency can be described as follows. The computer
program sets the frequency at the frequency generator. As the signal from the frequency
generator has not enough power, an audio amplifier is used to amplify the signal. The
electromagnetic coil excites the structure where a magnet is glued onto. The settling
time tsettling is choosen with tsettling = Q/f , where Q is the Q-factor. After the settling
time for the system, the laser vibrometer measures the vibration at a certain point on the
structure. The signal from the demodulator which is proportional to the velocity of the
vibration of the plate is the input signal (VSignal ) for the Lock-In amplifier (figure 2.7).
The transistor-transistor logic (TTL) signal from the frequency generator (VReference )
serves as reference . Within the Lock-In amplifier both signals are multiplied two times
Figure 2.7: Schematic diagram of a Lock-In amplifier where X is the out-of-phase part
of the signal and Y the in-phase part of the signal. With X and Y the amplitude as
well as the phase can be computed.
and low pass filtered while one time the reference signal is shifted by 90 .
The output of the Lock-In amplifier is the amplitude and phase-difference between the
sensor signal and the reference signal. These two values are stored within the program.
Afterwards, the program will set the next frequency for the evaluation. When the
frequency range has been measured, one gets a file with the frequencies, amplitudes
and phases for the selected range. With the curve fitting toolbox in Matlab, the two
curves were fitted to the characteristics of a single degree of freedom (SDOF) system
with the resonance frequency fres and the damping ratio . Note that is related to Q
by Q = 1/2. This is valid for a linear system where the eigenfrequencies are separated.
With the fitting function derived for one mode from equation (2.1)
s
a f2
|G| = (2.8)
2 2 f 2 + 2 (fres
2 f 2 )2
18
2.2. Modal Testing
where G is the transfer function, f is the frequency and a is a scaling factor, one can
get the resonance frequency and the damping. The same can be done for the phase
spectrum. The fitting function is there
2
(fres f 2)
= b + arctan( ) (2.9)
f
where is the phase of the transfer function and b is the phase value at the end of the
selected fitting range. For the split modes instead of a SDOF system, a two degree of
freedom system has been used.
2.2.2 Finite Element Simulation
There are several finite element simulation programs available for the investigation of
vibrations of structures submerged in water. Common programs especially for the fluid-
structure interaction are ANSYS, ABAQUS and COMSOL. Here ANSYS has been used.
During the course of this thesis, new versions of ANSYS have been released and used
for the present work. In the following the different methods and elements are described
which have been used in general for the simulations with ANSYS version 14. All simu-
lations have been done in Workbench, which is the graphical user interface for ANSYS
since version 11.
Geometry
From the ANSYS Workbench Platform the ANSYS DesignModeler has been used for
construction of geometry from the ground up.
Meshing
From the ANSYS Workbench Platform the ANSYS Meshing has been used for mesh
generation. For the rectangular parts, a mesh with cuboids has been used where for the
circular parts a structured prismatic mesh has been used with a spiderweb as base.
Elements
As ANSYS Workbench contains also an automatic mesher, the choice which type of
element is used, is made by the program itself.
19
Solid For all solid parts, the SOLID186, a 3D 20-Node Structural Solid Element, has
been used [3]. This element has a quadratic displacement behavior and each node has
three orthogonal translational degrees of freedom.
Acoustic To simulate the added mass effect of water there is no need to use a Com-
putational Fluid Dynamics (CFD) package. This kind of simulation can be done with
acoustic elements. The name of the used 3D Acoustic Fluid 20-Node Solid Element
is FLUID220 [3]. It has a quadratic pressure behavior. A typical application of this
element is the dynamics of submerged structures. It can be used for fluid media as well
as for the interface in fluid-structure interaction problems. Each element node has four
degrees of freedom: the pressure and the three orthogonal translational degrees. These
are only available at interface nodes. The governing equation for this element is the
3D wave equation (2.4). As the coupling equation (2.5) is unsymmetrical it makes it
unhandy to solve it because a special solver for unsymmetrical matrices has to be used.
But since version 14, ANSYS can make these matrices symmetrical and use then the
standard solver. The procedure is described by Sigrist [41].
In order to avoid contact elements between solid and acoustic, the whole model has to
be meshed as one. Therefore the model was fully meshed with 20-Node elements. As
the acoustic elements are not available in workbench directly the elements which have
to be acoustic elements have to be replaced by additional code written using ANSYS
Parametric Design Language (APDL). The following source code has to be used in a
command file within the workbench environment.
Source Code Comment

et,102,220,,2 Element definition for FSI interfaces present for modal
analysis (symmetric element matrix)
mp,dens,102,1000 Material definition for the density of fluid (water)
mp,sonc,102,1500 Material definition for the speed of sound in fluid (water)
cmsel,s,Fluid Select all elements from the predefined Fluid domain
nsle,s,1 Select all nodes which belongs to the Fluid domain
emodif,all,type,102 Changing of the element type
emodif,all,mat,102 Changing of the material type
allsel Select all
cmsel,s,if_node Select all nodes from the predefined interface between
structure and acoustic
sf,all,fsi Set a FSI-flag to all this nodes
allsel Select all
20
2.2. Modal Testing
For all ANSYS simulations with fluid in this thesis, the 3D Acoustic Fluid 20-Node Solid
Element has been used except for the transient two-way coupling between a structural
analysis and CFD presented in chapter 4. The coupling between the structure and the
acoustic is within one structural simulation and solved with equation (2.5).
It is recommended to use at least six to ten elements per wavelength [34]. This require-
ment has been fulfilled for all used models in this thesis.
Setup
During the setup one has to apply excitation forces, other external forces or boundary
conditions.
Boundary Condition Clamping of solid elements has been modeled by setting the three
displacement degrees of freedom equal to zero. The two used boundary conditions for
the acoustic elements were on the one hand setting the pressure equal to zero to model a
free water surface and on the other hand setting no boundary condition for the pressure
to model an infinite rigid wall. No pressure boundary condition is equivalent to an
interface where the pressure is translated to displacement and fixing the displacements.
Solver
As long as the simulation is done fully in the workbench environment the choice of the
right solver is made by ANSYS.
Modal Analysis To calculate the eigenfrequencies, a modal analysis is performed. This

is a fast way to characterize partly the vibration behavior of a structure with the disad-
vantage to get no information about the amplitude.
Harmonic Analysis To get an amplitude and phase spectrum, a harmonic analysis is

performed. In a defined frequency range the amplitude and phase of a defined number of
frequencies which are equally spaced in the range are calculated. To reduce the number
of frequencies which have to be calculated without loosing interesting amplitude and
phase information around the eigenfrequencies, the results of a modal analysis can be
used to cluster the frequencies which have to be calculated around the eigenfrequencies.
21
Transient Analysis The modal analysis as well as the harmonic analysis are calculated
in the frequency domain. To calculate the vibration behavior in the time domain, a
transient analysis is performed. For a two-way coupling between structural analysis
and CFD analysis, calculation in the frequency domain is not possible with ANSYS.
A transient two-way coupling analysis between structural analysis and CFD analysis
including viscosity is the only way to get information about the damping behavior of a
system with two-way coupling. The advantage of this method is that also the damping
induced by the fluid is included. The disadvantage is that it is very time consuming
as every time step for the structural analysis as well as for the CFD analysis has to
converge separately.
Result
To assign the eigenfrequencies from a modal analysis to a certain mode shape it is

appropriate to visualize the deformation of the plate for each eigenfrequency. Also the
shape of a structure for a simulated frequency of a harmonic analysis can be visualized
by plotting the deformation. A further output is a Bode diagram were the amplitude and
phase as a function of the frequency is plotted. In a transient analysis the deformation
of the structure can be visualized over the time.
2.3 Coupling Effects caused by the Fluid-Structure

Interaction: A Simple Example
To understand the physics of the coupling effects caused by the fluid-structure inter-
action, a simple example has been set up. Two identical circular plates with fluid in
between are taken. This example is according to Jhung et al. [24]. A critical review of
the publication has been done.
Jhung et al. studied the vibration of the two plates clamped at the outer edge (CF) while
the gap between them was filled with water. The circular plates are made of aluminum,
have a diameter of 300 mm and a thickness of 3 mm. For gap distances ranging between
10 mm and 60 mm, Jhung et al. identified the eigenfrequencies for the lowest modes.
They used an analytical method and finite element simulation. The analytical method
was based on the finite Fourier-Bessel series expansion and Rayleigh-Ritz method. For
the finite element simulation using ANSYS, four node elastic shell elements (SHELL63
22
2.3. Coupling Effects Caused by the Fluid-Structure Interaction
[3]) for the plates and an eight node displacement-based incompressible fluid element
(FLUID80 [3]) have been used. The discrepancy between the analytical method and the
finite element simulation for the lowest modes was below 3 %.
(a) out-of-phase mode (b) in-phase mode
Figure 2.8: Simulation of two CF circular plates using ANSYS. A symmetry plane
is introduced along the diameter. In between these plates is water. Both subfigures
are showing the 1-0 mode. Subfigure (a) shows the plates vibrating 180 out-of-phase.
Subfigure (b) shows the plates vibrating in-phase.
The modes consist of the plates vibrating either in-phase or out-of-phase (180 ). As the
ratio between the diameter and the wavelength of the fluid as well as the ratio between
the distance between the two circular plates and the wavelength of the fluid are both
much smaller than one, no fluid modes are present. Jhung et al. made their analysis
for a radially bounded fluid region as well as an unbounded one. To compare different
modes and boundary conditions, the eigenfrequency has been normalized by dividing
the eigenfrequency of the system with the fluid by the eigenfrequency without the fluid.
The following coupling effects which are related to the normalized eigenfrequency of two
identical circular plates with water in between, have been postulated by Jhung et al.:
I The fluid affects the out-of-phase modes more than the in-phase modes.
II The increase of the number of nodal diameters of the modes has more effect on
the eigenfrequency of the out-of-phase modes than on the in-phase modes.
III The decrease of the distance between the plates affects the eigenfrequencies signif-
icantly.
a) For the out-of-phase mode, the frequency has a downward trend for a de-
creasing distance.
23
b) For the in-phase mode, the frequency has an upward trend for a decreasing
distance.
A recent study by Askari et al. [4] revealed that for the explanation of the coupling effects
one has to consider not only the added mass of the fluid but also the fluid movement. For
the in-phase modes, the fluid has to move mainly in the direction which is perpendicular
to the plate. This movement is only little affected by the distance between the two plates.
However for the out-of-phase modes the fluid has to move mainly in the radial direction.
Decreasing the distance between the two plates will increase the maximum velocity
significantly. This consideration of the fluid movement is used for the explanation of
the coupling effect III. Furthermore Askari et al. try to explain the coupling effect II
by mentioning that the affected fluid region is larger for lower modes than for higher
modes. But this is only the explanation for an extension of the coupling effect II. This
extension would be that the coupling effect in general is decreasing for higher modes.
To summarize, there are three main aspects which can be used for the explanation of
the coupling effects:
The mass of the fluid which has to be moved
The velocity of the movement of the fluid
The influenced fluid region of the particular mode
To investigate these coupling effects, a more detailed parametric acoustic modal analysis
has been done here using ANSYS in the present study. The same dimensions and
properties have been taken as the ones used by Jhung et al. [24]. As the lowest modes of
water turbines have no nodal circles and the lowest modes have the highest amplitudes
which can lead to fatigue cracks, this study focused only on modes without nodal circles.
Additionally to the distances between the two plates used by Jhung et al., for the
distances between the two plates of 1 mm, 80 mm and 100 mm, the eigenfrequencies
were also calculated. Only a fluid radially bounded by a rigid wall was considered and
in addition the plate on top is replaced with a rigid wall as well as a free water surface.
As the focus here is set on the coupling effects, only small distances between the two
plates are relevant for the present work. The distance between the two plates should
be comparable or smaller than the shortest distance xd between two nearby antinodes
with opposite displacement. For a circular plate, this distance xd can be roughly ap-
proximated by xd = r/n, where r is the radius of the circular plate and n the number
of nodal diameters.
24
(1) in-phase mode (2) free water surface
(3) out-of-phase mode (4) wall
Figure 2.9: A schematic view of the radial symmetry plane of the four different cases
for the investigation of the dynamics of two identical CF circular plates with different
water-levels in between.
Mode 1-0 Mode 2-0 Mode 3-0 Mode 4-0

fres [Hz] fres [Hz] fres [Hz] fres [Hz]
in-phase
FLUID80 363.0 609.0 914.0 1279.0
FLUID220 358.0 604.1 910.2 1279.3
out-of-phase
FLUID80 150.0 371.0 674.0 1055.0
FLUID220 154.2 380.1 688.5 1073.3
Table 2.1: Eigenfrequencies from different simulations of two circular plates with 50 mm
water in between using ANSYS. Jhung et al. [24] has used the FLUID80 element for water
which is a displacement-based incompressible fluid element. The FLUID220 element
which is a pressure-based compressible fluid element is used in the present work. The
discrepancy between the simulations ranges between 0.02 % and 2.75 %.
The four cases shown in figure 2.9 are a view of the radial symmetry plane. As shown
in figure 2.10, only half of the plates and fluid are modeled and a symmetry plane is
introduced. Using the symmetry of the system, the computational effort is reduced and
the antinodes of the modes align all with the symmetry plane.
25
Figure 2.10: The model for the finite element simulation using ANSYS. Symmetry axis
along the diameter (red area) is used at the two circular plates with fluid in between to
reduce the computational effort and to align the mode to have all the antinodes at the
intersection. The translational degrees of freedom at the outer edge of the plates were
fixed (dark blue area) to have two CF circular plates. The fluid is bounded at the outer
edge.
For the mesh of the model, three elements are used for the thickness of the plates and
for the thickness of the fluid ten elements are used as shown in figure 2.11.
The comparison of the eigenfrequencies for the lowest modes without nodal circles of the
circular plates with 50 mm fluid in between of the finite element simulation by Jhung et
al. (with FLUID80) and the finite element simulation by the author of this thesis (with
FLUID220) is shown in table 2.1. The discrepancy between the two FEM simulations
ranges between 0.02 % and 2.75 %.
In figure 2.12 for discrete distances between 100 mm and 1 mm the normalized eigen-
frequency of each mode fi water /fi air is plotted for the modes with one to seven nodal
diameters but without nodal circles. In this case the reference eigenfrequency is the one
without any fluid, for example vacuum. It can be seen that in-phase modes (cases 1)
and free water surface (case 2), as well as out-of-phase modes (cases 3) and wall (case
4) behave in a similar manner, respectively.
A further symmetry can be introduced in the middle between the two circular plates. For
the out-of-phase modes, this symmetry is a plane of symmetry for the displacement and
26
Figure 2.11: For the mesh of the plates, three elements (SOLID186) for the thickness
were used. The fluid thickness was resolved by ten elements (FLUID220). For the mesh
6016 elements were used with a total of 27 125 nodes.
for the in-phase mode a plane of symmetry for the pressure. These symmetry conditions
can be replaced for the out-of-phase mode with a hard boundary condition, for example
a rigid wall, and for the in-phase mode with a soft boundary condition, for example a
free water surface. The consequence is that the eigenfrequency of a certain mode with a
wall on one side is equal to the eigenfrequency of the same mode which is out-of-phase
with the double water-level in between. The same is true for the eigenfrequencies of a
mode with a free water surface on one side and an in-phase mode.
For an increasing number of nodal diameters of the modes, the normalized eigenfrequency
increases but the gradient of the data points for the out-of-phase modes is higher than
for the in-phase modes. For a small height of the water the number of nodal diameters
will have almost no influence for cases 1 and 2.
In figure 2.13 the same data are plotted versus the distance between the two plates.
The curves show an asymptotic behavior to a value of the eigenfrequency which is valid
for large gaps. At a certain gap distance a further increase will have no influence to
the eigenfrequency. This gap distance has a downward trend for an increasing number
of nodal diameters of the modes. To explain this, not only the gap distance has to be
considered but also the shortest distance between the center of two regions with opposite
displacement behavior of one plate. Between these regions there is an exchange of fluid.
27
100 mm 80 mm 60 mm
1
0.8
fwater /fair
0.6
0.4
0.2
0
50 mm 40 mm 30 mm
1
0.8
fwater /fair
0.6
0.4
0.2
0
20 mm 10 mm 1 mm
1
0.8
fwater /fair
0.6
0.4
0.2
0
1 2 3 4 5 6 7 1 2 3 4 5 6 7 1 2 3 4 5 6 7
Mode n-0 Mode n-0 Mode n-0
In-phase mode Free water surface Wall Out-of-phase mode
Figure 2.12: Simulation of the eigenfrequencies of the CF circular plates which are
coupled by a bounded fluid. The eigenfrequencies of the circular plates are normalized by
the eigenfrequencies of the circular plates without any fluid. Here for different separation
distances the normalized eigenfrequencies are plotted against different modes with pure
nodal diameters n. The normalized eigenfrequencies increase for an increasing number
of nodal diameters n.
The influence on the eigenfrequency of the gap distance becomes small when the gap
distance is larger than the distance xd .
In the figures 2.14 and 2.15 one can see a corresponding view of pressure and inplane
velocity maxima of the symmetry plane of the result of a modal analysis for the 1-0, 2-0,
28
Mode 1-0 Mode 2-0 Mode 3-0

1
0.8
fwater /fair
0.6
0.4
0.2
0
1
0.8
fwater /fair
0.6
0.4
0.2
0
0 0.02 0.04 0.06 0.08 0.1 0 0.02 0.04 0.06 0.08 0.1 0 0.02 0.04 0.06 0.08 0.1
separation distance [m] separation distance [m] separation distance [m]
In-phase mode Free water surface Wall Out-of-phase mode
Figure 2.13: Simulation of the eigenfrequencies of the CF circular plates which are
coupled by a bounded fluid. The eigenfrequencies of the circular plates are normalized
by the eigenfrequencies of the circular plates without any fluid. Here for the modes with
pure nodal diameters ranging between 1-0 mode and 6-0 mode the normalized eigenfre-
quencies are plotted against different separation distances. The difference between an
eigenfrequency of an in-phase and an out-of-phase mode at the same separation distances
decreases for increasing the separation distances.
5-0, and 6-0 mode of two CF circular plates, using ANSYS. Between the plates, there
is 50 mm radially bounded water. The exaggerated deformation of the plates above and
below the water is shown with their maximum vertical displacement. For the water, the
maximum pressure field as color plot and the maximum velocity field as vector plot is
shown. The maximum pressure field and the maximum velocity field is 90 phase shifted.
This means that the maximum velocity occurs when the plates have no deformation and
the pressure field is zero. The color map of the pressure of the fluid is from negative
relative pressure to zero relative pressure to positive relative pressure from blue to green
to red. The length of the arrows is proportional to the fluid velocity.
29
Mode 1-0 out-of-phase
Mode 1-0 in-phase
Mode 2-0 in-phase
Figure 2.14: Simulation results of two CF circular plates in a radial symmetry plane. In
addition 50 mm radially bounded fluid is shown in between. The exaggerated maximum
deformation of the plates is shown. The fluid domain shows the maximum pressure field
as color plot and the maximum inplane velocity field as vector plot. The color plot shows
values from negative to zero to positive as blue, green, and red, respectively.
Special attention should be paid to the regions where the length of the arrows is not
negligible. They are for the 1-0 and 2-0 mode larger than for the 5-0 and 6-0 mode.
30
Mode 5-0 in-phase
Mode 6-0 in-phase
Figure 2.15: Simulation results of two CF circular plates in a radial symmetry plane. In
addition 50 mm radially bounded fluid is shown in between. The exaggerated maximum
deformation of the plates is shown. The fluid domain shows the maximum pressure field
as color plot and the maximum inplane velocity field as vector plot. The color plot shows
values from negative to zero to positive as blue, green, and red, respectively.
For the in-phase modes, the maximum velocity of the fluid will be approximately the
same for all modes with pure nodal diameters independent of the distance between the
31
two plates for the given period of one oscillation period. The main influence on the
eigenfrequencies is given by the mass of the fluid which has to be moved. For small
distances between the plates, there is only little mass which has to be moved and for a
larger distance between the plates there is more mass to be moved.
For the out-of-phase modes, the increasing maximum velocity of the movement of the
fluid which has to follow the movements of the plates for decreasing the distance between
the plates is the main effect for the decreasing normalized eigenfrequency.
To summarize the coupling effects induced by the fluid-structure interaction which have
an influence on the eigenfrequency of the plates the points postulated by Jhung et
al. have been adapted and extended as follow. The new parts are emphasized. In
various experiments and simulations for different structures these effects will be verified.
Extended Coupling Effects1

1. The fluid affects the out-of-phase modes more than the in-phase modes.
2. The increase of the nodal diameters of the modes has a larger effect on the eigen-
frequency of the out-of-phase modes than on that of the in-phase modes. The
coupling effect in general is decreasing for an increase of the nodal diameters.
3. The decrease of the distance between the plates affects the eigenfrequencies signif-
icantly.
a) For the out-of-phase mode, the frequency has a downward trend for a de-
creasing distance.
b) For the in-phase mode, the frequency has an upward trend for a decreasing
distance.
4. The fluid region which interacts with the vibration of the plates is limited. For an
increase of the nodal diameters of the modes, the affected fluid region is decreasing.
5. The behavior of a rigid wall (case 4) is similar to an out-of-phase mode (case 3)
and the behavior of a free water surface (case 2) is similar to an in-phase mode
(case 1).
1
with respect to the postulated coupling effects by Jhung et al. [24]
32
3 Rectangular Plate
When studying fundamental vibration effects, rectangular plates are in general taken as
a first choice, because of the possibility to find analytical solutions. Also when trying to
understand the influence of water on the vibration behavior several studies have been
made for rectangular plates with different clamping. Lindholm et al. [33] studied a fully
submerged CFFF plate, both from a theoretical and experimental point of view. They
investigated the influence of a nearby free water surface on the vibration behavior while
changing the water-level above the plate. In 1978 Marcus [35] simulated the vibration
of a submerged CFFF plate using acoustical elements for the fluid and compared the
results with experiments. Fu and Price [18] improved the theoretical model including a
free water surface. An analytical and experimental study of a CFCF plate was presented
by Haddara and Cao [19]. They evaluated the effect of the free water surface on the
eigenfrequency as well as on the damping ratio for several depths of submergence. Liang
et al. [31] used an empirical added mass formulation to estimate the eigenfrequency
of a submerged CFFF plate and compared it with numerical and experimental values
from the literature. Sinha et al. [42] suggested a formula for added mass based on
experimental and analytical studies of a submerged CFFF plate. They also studied
the damping behavior of the plate experimentally. In 2007 Vu et al. [47] made an
experimental study on CFFF and CFCF plates to obtain the influence of the water on
the eigenfrequency of the plate as well as on the damping.
3.1 Plate without clamping
As mentioned in subsection 2.2.1, the experimental realization of a plate without clamp-

ing is more easy than a plate with fixed clamping. Hence, for a preliminary test, a FFFF
plate made of stainless steel, which is 300 mm by 200 mm large and has a thickness of
5 mm (figure 3.1(a)), has been taken to validate as a first test the simulation of the
eigenfrequencies of structures submerged in water. The values are compared to analyt-
33
Chapter 3. Rectangular Plate
ical values from the literature as well as to experimental values. To come as close as
possible to the ideal case of a plate without clamping, in reality the plate was hung at
the four corners with rubber bands. A drawback of this clamping position is that a lot
of mode shapes have their maxima at the corners, as mentioned in 2.2.1. In the finite
element model the rubber bands were approximated with springs with a spring constant
of 16 000 N m1 . To excite the plate, an electromagnetic coil driven by a frequency gen-
erator was used. A magnet was fixed on the plate to transfer the forces to the plate.
The excitation was placed at the longer edge at about one third of the distance. The
position was selected by searching for a location where a minimal number of the first
modes have a nodal line close-by.
(a) Dimensions of the FFFF rectangular plate (b) Model of the FFFF rectangular plate
Figure 3.1: The used FFFF rectangular plate. Subfigure (a) dimensions for the vi-
bration measurement and subfigure (b) model of the FFFF rectangular plate with fluid
around. To get the most accurate model of the experimental setup, the plate is held by
four springs which are placed in the corners of the plate.
The vibration measurement of the plate has been done with the laser scanning vibrometer
system. For the excitation signal, a periodic chirp was chosen. The tests were made in
a box with the dimensions 450 mm, 320 mm and 300 mm. The mode shapes of the first
nine eigenmodes can be seen in figure 3.2. These mode shapes are the result of the finite
element modal analysis in air.
34
3.1. Plate without clamping
Figure 3.2: The mode shapes of the first nine eigenmodes of a FFFF rectangular plate.
These mode shapes are the result of the finite element modal analysis in air. The color
represents the displacement of the mode shape perpendicular to the plate. The color
map is from negative displacement to zero displacement to positive displacement from
blue to green to red.
3.1.1 Eigenfrequency in Air
The experimental averaged amplitude spectrum of the FFFF rectangular plate in air
measured with the laser scanning vibrometer system is shown in figure 3.3 between
100 Hz and 1600 Hz which includes the first nine eigenmodes. In table 3.2 the measured
eigenfrequencies of the first nine eigenmodes of a FFFF rectangular plate are shown
together with the values by Blevins [7] and the result of the finite element simulation.
The analytical values of the eigenfrequencies have been calculated with formula (3.1)
s
2 Eh2
fi = i 2 (3.1)
2a 12(1 2 )
35
where fi is the frequency, 2i is the dimensionless frequency parameter shown in table 3.1
which is dependent on the aspect ratio of the plate, a is the length of the longer edge
of the plate, E is the Youngs modulus, h is the plate thickness, is the Poissons ratio
and is the density. By tuning the Youngs modulus and the density of the plate, the
simulated eigenfrequencies of the plate in air are fitted to the ones from the measurements
of the plate in air. The following values have been used: a Youngs modulus of 210 GPa,
a density of 7850 kg/m3 and a Poissons ratio of 0.3
105
Mode 1-1
8 Mode 3-0
Amplitude [m/s]
Mode 2-0
6
Mode 0-2 Mode 3-1
4
Mode 2-1
Mode 2-2
2 Mode 1-2 Mode 4-0
0
100 200 300 400 500 600 700 800 900 1000 1100 1200 1300 1400 1500 1600
Frequency [Hz]
Figure 3.3: Experimental averaged amplitude spectrum of the FFFF rectangular plate
in air measured with the laser scanning vibrometer system.
Mode 1-1 Mode 2-0 Mode 2-1 Mode 0-2 Mode 3-0 Mode 1-2
2i 20.13 21.60 46.65 50.29 58.20 67.49
Table 3.1: Dimensionless frequency parameter 2i used by Blevins [7] (Table 11-4, page
253) for an aspect ration of the longer edge to the short edge of the plate of 1.5. The
values are accurate for the Poissons ratio = 0.3.
The analytical values are in good agreement with the experimental values, as well as
the values of the simulation. The median of the deviation percentage from the experi-
ment and the simulation is 1.2 % with a standard deviation of 0.57 %. The median of
the deviation percentage could be reduced further with a finer tuning of the material
parameters within the simulation.
3.1.2 Eigenfrequency in Water
After measuring the plate in air, the following test has been conducted in water. The
measurements in water have to be done to validate the acoustic finite element model. The
36
Experiment Analytical Simulation fSim fExp

fExp [Hz] fAna [Hz] fSim [Hz] fExp
Mode 1-1 271.1 271.9 275.21 0.0152
Mode 2-0 289.8 291.7 291.48 0.0058
Mode 2-1 618.0 630.0 622.71 0.0076
Mode 0-2 663.3 679.2 673.98 0.0161
Mode 3-0 775.0 786.1 775.05 0.0001
Mode 1-2 884.4 911.5 902.53 0.0205
Mode 3-1 1135 - 1148.2 0.0116
Mode 2-2 1302 - 1317.8 0.0121
Mode 4-0 1594 - 1611.5 0.0110
Table 3.2: Experimental, analytical and simulated eigenfrequencies in air of the first
nine eigenmodes of the FFFF rectangular plate. For the analytical eigenfrequencies, only
dimensionless frequency parameters for the first six eigenmodes have been published by
Blevins [7].
water height below the plate was kept constant at about 50 mm, whereas the water-level
above the plate was varied between 0 mm and 100 mm. A finite element simulation has
been performed using ANSYS. The model and the used mesh are shown in figure 3.1(b)
and 3.4, respectively. The mesh is the same for all water-levels.
Figure 3.4: The mesh of the FFFF rectangular plate. The plate thickness was resolved
by three elements and the fluid on each side by five elements. Therefore the model has
11 424 elements with a total of 50 725 nodes.
Figure 3.5 shows the experimental averaged amplitude spectrum of the FFFF rectan-
gular plate in water with a water-level of 50 mm below the plate and 40 mm above the
37
plate measured with the laser scanning vibrometer system. The corresponding eigen-
frequencies in water are shown in table 3.3 together with the values of the simulation.
Comparing the simulation with the experiments, one can see that for the 2-0 mode the
discrepancy between the experiment and the simulation is slightly larger. No explanation
has been found for this discrepancy.
105
4 Mode 0-2
Mode 3-0
Amplitude [m/s]
Mode 1-1
Mode 2-1
Mode 3-1
2 Mode 1-2 Mode 2-2
Mode 2-0
Mode 4-0
0
100 200 300 400 500 600 700 800 900 1000 1100 1200
Frequency [Hz]
Figure 3.5: Experimental averaged amplitude spectrum of the FFFF rectangular plate
in water with a water-level of 50 mm below the plate and 40 mm above the plate measured
with the laser scanning vibrometer system.
Experiment Simulation fSim fExp

fExp [Hz] fSim [Hz] fExp
Mode 1-1 190.6 187.77 0.0148
Mode 2-0 200.0 192.52 0.0374
Mode 2-1 447.7 443.13 0.0102
Mode 0-2 478.1 475.65 0.0051
Mode 3-0 560.2 551.90 0.0148
Mode 1-2 635.2 638.29 0.0049
Mode 3-1 831.3 845.20 0.0167
Mode 2-2 973.4 981.05 0.0079
Mode 4-0 1153 1156.5 0.0030
Table 3.3: Experimental and simulated eigenfrequencies in water with a water-level of

50 mm on both sides of the first nine eigenmodes of the FFFF rectangular plate.
In figure 3.6 one can see both the results of the finite element simulation for a water-
level above the plate of 0 mm up to 100 mm in steps of 10 mm and the measurements
for a water-level above the plate of 0 mm up to 100 mm in steps of 20 mm. In this case
the distance between the plate and the bottom was kept constant at 50 mm. To test
38
the repeatability of the measurements, for some water-levels the measurements were
repeated, which is depicted by multiple for one water-level in the figure.
As expected, due to the added mass effect, the eigenfrequencies of the respective modes
in water are lower than the ones in air and have a downward trend with increasing water-
level above the plate (see coupling effect 3b). The graph shows that there is no linear
relation between the water-level and eigenfrequency. It can be seen that the measured
values follow a logarithmic trendline of the form y = a ln(x) + b, where x is the water-
level between 1 mm and 100 mm. This behavior corresponds to the coupling effect 4
(page 32). Furthermore it can be seen that the affected water region ends for this plate
at a distance of about 100 mm depending on the mode shapes. The ratio fwater /fair has
a downward trend for an increasing mode number which corresponds to the coupling
effect 2.
39

0.85
0.8
fwater /fair
0.75
0.7
0.65
0.6
0 0.02 0.04 0.06 0.08 0.1 0 0.02 0.04 0.06 0.08 0.1 0 0.02 0.04 0.06 0.08 0.1

0.85
0.8
fwater /fair
0.75
0.7
0.65
0.6
0 0.02 0.04 0.06 0.08 0.1 0 0.02 0.04 0.06 0.08 0.1 0 0.02 0.04 0.06 0.08 0.1

0.85
0.8
fwater /fair
0.75
0.7
0.65
0.6
0 0.02 0.04 0.06 0.08 0.1 0 0.02 0.04 0.06 0.08 0.1 0 0.02 0.04 0.06 0.08 0.1
Water-level above [m] Water-level above [m] Water-level above [m]
Simulation Experiment Trendline
Figure 3.6: Simulated and experimental eigenfrequencies of the first nine eigenmodes
of a FFFF rectangular plate for different water-levels above the plate and a constant
distance to the bottom of about 50 mm. The eigenfrequency is set in relation with
the eigenfrequency in air. The trendline - - - - of the experimental data has the form
y = a ln(x) + b, where x is the water-level between 1 mm and 100 mm.
40
3.2. One- and Two-Sided Clamped Plate
3.2 One- and Two-Sided Clamped Plate
Similar to Vu et al. [47], a further preliminary study has been done with rectangular
plates to investigate the influence of a nearby wall on the eigenfrequency of the plates.
Two plates have been used: a CFFF rectangular plate with a length of 377 mm and a
CFCF rectangular plate with a length of 500 mm. Both plates have a width of 200 mm
and a thickness of 2 mm which is shown in figure 3.7
Figure 3.7: CFFF and CFCF rectangular plate and its dimensions for the vibration
measurement. The hatched area is the area where the plates are clamped between the
two blocks.
For the experimental setup a box has been used with the dimensions 760 mm, 550 mm
and 300 mm. For the excitation and the measurement, the same systems have been used
as in section 3.1 with an electromagnetic coil which is driven by a frequency genera-
tor, and for the vibration measurement the laser scanning vibrometer system. For the
experiments each plate has been placed inside the box. The plate is clamped between
two blocks of stainless steel for each side of the plate which has to be clamped. The
dimensions of the upper block are 250 mm, 40 mm and 60 mm and the dimensions of the
lower block are 250 mm, 40 mm and 8 mm. As the length of the clamping is 40 mm, each
plate is at the clamped ends 40 mm longer. The blocks were pressed together with six
M6 screws.
3.2.1 Eigenfrequency in Air
The eigenfrequencies of the plates are listed in table 3.4. The corresponding mode shapes
are shown in figure 3.8.
41

CFFF plate 10.47 52.66 80.78 163.0
CFCF plate 31.3 86.6 109.9 181.1
Table 3.4: Experimental eigenfrequencies in air of the CFFF and CFCF rectangular
plate.
CFFF plate
CFCF plate
Figure 3.8: The mode shapes of the first four eigenmodes of the CFFF and CFCF
rectangular plate. These mode shapes are the result of the finite element modal analysis
in air. The color represents the displacement of the mode shape perpendicular to the
plate. The color map is from negative displacement to zero displacement to positive
displacement from blue to green to red.
3.2.2 Eigenfrequency in Water
All measurements in water have been done with a constant water-level above the plate
of 150 mm. The water-level below the plate is investigated for the distances of 8 mm,
18 mm, 28 mm, 38 mm and 48 mm. The eigenfrequencies of the plates with a water-level
below of 48 mm are listed in table 3.5. Because of the limited frequency range of the
used audio amplifier starting at 10 Hz, eigenfrequencies below 10 Hz are not measured.

CFFF plate - 22.5 - 78.0
CFCF plate 11.13 35.75 24.25 79.5
Table 3.5: Eigenfrequencies in water of the CFFF and CFCF rectangular plate. The
water-level above the plate is 150 mm and below 48 mm. Compared to the eigenfrequen-
cies in air, in water the 0-1 mode has a lower eigenfrequency than the 1-0 mode.
In figure 3.9 one can see the normalized eigenfrequency of the CFFF and CFCF plate
for different values of the water-level below the plate for the first four eigenmodes. The
42
3.2. One- and Two-Sided Clamped Plate
eigenfrequencies for different water-levels below show an asymptotic behavior as already

seen before. The difference in this case is that the water-level now is varied between the
plate and the bottom of the box which can be approximated as a rigid wall. As the value
for fwater /fair is asymptotically increasing one can see the coupling effect 3a (page 32).
CFFF plate CFCF plate

0.5
0.4
fwater /fair
0.3
0.2
0.1
0
0.1 0.2 0.3 0.4 0.5 0.1 0.2 0.3 0.4 0.5
Water-level below [m] Water-level below [m]
Figure 3.9: Experimental eigenfrequencies of the CFFF and CFCF rectangular plate
in water. The eigenfrequency is set in relation to the eigenfrequency in air. Due to the
lower frequency limitation of the used measurement setup, not all eigenfrequencies could
be measured. Therefore, only the modes of table 3.5 are shown.
43
44
4 Beam
Beside simulating the eigenfrequencies of a structure in water correctly, this thesis aims
also at simulating the damping of a structure in water. In order to allow for a comparison
between the simulation and the experiment, it was important that the structure is easy to
model and easy to handle for the experiments. Therefore a test case has been performed
with a cantilever beam submerged in water. The experimental part described in this
chapter has been done in collaboration with Philipp Rust and Thomas Wattinger from
the Center of Mechanics of ETH Zurich.
To calculate the eigenfrequencies of a beam analytically, Van Eysden and Sader [46]
published in 2006 an enhanced model. Previously, Elmer and Dreier [15] noted already
that the added mass on a submerged cantilever beam is frequency dependent, with the
highest influence at low frequencies.
4.1 Design
The aluminum beam has the dimensions 190 mm, 10 mm and 5 mm which is shown in
figure 4.1. On the top and bottom it was clamped over a length of 40 mm. Consequently
the free length of the beam is 150 mm.
Figure 4.1: Beam and its dimensions for the vibration measurement. The hatched area
is the area where the beam is clamped between two blocks.
45
Chapter 4. Beam
4.2 Material Properties
As the literature values of the material data for aluminum are not precise enough for
the finite element simulation, an analysis was made first to experimentally determine
the material properties. Two beams were tested. The determined material properties
for them are listed in table 4.1. To calculate the density, the beams were weighed
with a balance (Mettler Toledo) with an accuracy of about 0.01 g. The dimensions
were measured with a caliper gauge. For the determination of the Youngs modulus
the eigenfrequency of the first longitudinal mode was measured with a phase-locked
loop [6]. For this purpose the beams were laid on two tensed strings. The vibration
velocity was measured with a Polytec laser interferometer. The Youngs modulus E was
calculated with E = (2lfres )2 where is the density and fres the eigenfrequency of the
first longitudinal mode.
Length Density Frequency Youngs modulus

[mm] [kg/m3 ] [Hz] [MPa]
Beam 1 190 2670.53 13 001.1 65.181
Beam 2 190 2670.53 12 995.6 65.126
Table 4.1: Measured material properties of the beams. The length was measured with
a caliper gauge, the density was calculated with the measured three dimensions of the
beam and the weight which was measured by a balance. The frequency correspond to
the first longitudinal mode. This eigenfrequency was measured with a phase-locked loop
[6]. The Youngs modulus was calculated with this first longitudinal eigenfrequency.
4.3 Eigenfrequency and Damping in Water
4.3.1 Experimental setup
In the experiments, beam 1 was excited with an electromagnetic acoustic transducer

as described in [45]. Beam 2 was excited with an electromagnetic coil and a magnet
glued on the lower side at the tip of the beam to transfer the force to the beam. The
cylindrical magnet of NdFeB with a diameter of 2 mm and a height of 1 mm has a weight
of 0.024 g. The eigenfrequency and damping were measured with a phase-locked loop
using a Polytec laser interferometer. The experiments have been carried out in a box
with the dimensions 760 mm, 550 mm and 300 mm. The box was filled with water up
46
4.3. Eigenfrequency and Damping in Water
to 105 mm so that below and above the beam there is 50 mm of water. All beams were
clamped for the experiments between two blocks of stainless steel with the dimensions
250 mm, 40 mm and 60 mm. The blocks were pressed together with six M6 screws. These
are the same blocks used in the experiments described in section 3.2.
For the simulation, the space besides and in front of the beam was reduced to a box
with the dimensions 240 mm, 70 mm and 105 mm. It was checked that the box size is
large enough to have a negligible influence on the vibration behavior of the beam.
4.3.2 Two-Way Coupling Simulation
The simulation of the eigenfrequencies has been done first with a modal analysis includ-
ing acoustic elements for air and water. To have the possibility to simulate not only the
eigenfrequency and mode shape of a structure submerged in a fluid but also the added
damping to the structure from the viscosity of the surrounding fluid, a transient two-way
coupling has been implemented. The used model is shown in figure 4.2.
Figure 4.2: The model of the cantilever beam with fluid around is clamped at one end
with the three translational degrees of freedom.
For the clamping of the beam, all degrees of freedom are fixed at the top and bottom
along 40 mm from one side. To excite the beam in the time domain, a force is applied
perpendicular to the length of the beam at the free tip. The excitation force is applied
as a symmetric triangle function for 10 time steps with a total length of 0.2 ms. The
time step for the simulation is 0.02 ms. This means a resolution for the third bending
47
Chapter 4. Beam
mode of about 20 points per period. 2000 time steps were simulated which is about 5.5
periods of the first bending mode.
During the transient two-way coupling, first the displacement of the beam is simulated.
The values for each node of the beam at the interface is input for the neighbor nodes of
the fluid domain. For the same time steps, the fluid domain is simulated. The forces at
the interface are the input for the beam for the next time step.
To evaluate the eigenfrequency and damping, the displacement at one node of the beam
where the third bending mode has its first antinode is taken and fitted with the following
P
function 3i=1 ai ei t cos(i t + i ) where a is amplitude, damping, angular frequency
and the phase of mode i. The displacement and the fitted function is plotted in
figure 4.4. In the upper plot the decay of the first and second bending mode is visible.
The lower plot shows an enlarged version of the time between 0 ms and 2 ms. Here, one
can see also the third bending mode.
Figure 4.3: The mesh of the beam is in the thickness direction resolved by three
elements. The mesh of the fluid is refined into the direction orthogonal to the surfaces
of the beam to have at least one element for the boundary layer. For the solid and fluid
555 820 elements were used with a total of 2 289 476 nodes.
48
106
3
Simulation
P3 i t
2 Fit i=1 ai e cos(i t + i )
1
Amplitude
0 0.5 1 1.5 2 2.5 3 3.5 4

Time [s] 102
106
2
Amplitude
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2

Time [s] 10 3
Figure 4.4: Transient simulation of the beams displacement decay process in time steps
of 0.02 ms for a duration of 40 ms. The amplitude is
Pplotted against the time. The used
3 i t
fit for the first three bending modes has the form i=1 ai e cos(i t + i ). The range
up to 2 ms has been scaled up to see the third bending mode more clearly. There is a
good agreement between the simulation and the fitted function.
The agreement between the simulation and the fitted function is very good. The squared
difference between the fitted function and the values from the simulation is plotted in
figure 4.5. The upper plot shows the squared residual over the time range up to 40 ms.
The squared residual has a correlation with the frequency of the first bending mode
which is negligible. In the scaled up plot between 0 ms and 1 ms one can see in addition
a slight correlation with the fourth bending mode. Due to the marginal content of the
oscillation of the fourth bending mode in comparison to the content of the first three
49
Chapter 4. Beam
bending modes it is not possible to include also the fourth bending mode into the fitting
function.
Squared Residuals
1015 Bending Mode 1
Squared Residuals
1017
1019
1021
1023
0 0.5 1 1.5 2 2.5 3 3.5 4
Time [s] 10 2
1013
Squared Residuals
1014 Bending Mode 4
Squared Residuals
1015
1016
1017
1018
1019
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

Time [s] 10 3
Figure 4.5: Squared residuals of the fitted function and the transient simulation for
each time step. In the upper plot the residual is shown up to 40 ms and in the lower
plot up to 1 ms. The oscillation of the squared residuals up to 1 ms correlates with the
oscillation of the fourth bending mode of the beam For the total range it correlates with
the first bending mode.
4.3.3 Comparison of Experimental and Simulated Results
The comparison of the eigenfrequencies from the experiment and the simulation for the
first three bending modes can be seen in table 4.2. The comparison of the damping values
are shown in table 4.3. Beam 1 is the experimental result for the beam excited with
50
an electromagnetic acoustic transducer, beam 2 is the experimental result for the beam
excited with an electromagnetic coil and a magnet, simulation A is a modal analysis with
acoustic elements and simulation B is a two-way coupling between a transient structural
analysis of the beam and a transient fluid dynamic analysis of the water. The simulations
are done without including structural damping.
Air Water fwater

fair [Hz] fwater [Hz] fair
1st Bending Mode
Beam 1 169.12 128.46 0.760
Beam 2 170.11 130.33 0.766
Simulation A 177.85 135.32 0.761
Simulation B - 137.3 -
Theory1 - 137 -
nd
2 Bending Mode
Beam 1 1060.1 809.3 0.763
Beam 2 1058.0 854.46 0.808
Simulation A 1108.8 849.26 0.766
Theory1 - 832 -
3rd Bending Mode
Beam 1 2947.5 2201.4 0.747
Beam 2 2942.6 2192.8 0.745
Simulation A 3079.7 2382.2 0.774
Theory1 - 2330 -
Table 4.2: Eigenfrequency in air and water of the beams. Beam 1 is the experimental
result for the beam excited with an electromagnetic acoustic transducer, beam 2 is the
experimental result for the beam excited with an electromagnetic coil and a magnet,
simulation A is a modal analysis with acoustic elements around and simulation B is a
two-way coupling between a transient structural analysis of the beam and a transient
fluid dynamic analysis of the water.
The experimental eigenfrequencies for air as well as for water are below the values of
the simulations. For air, the difference between the simulation and the experiment is
for all modes between 4.5 % and 5.2 %. The difference between the simulation and the
experiment for water is for all modes between 5.3 % and 9.6 %. The main reason for
this approximately constant deviation of about 5 % between the simulation and the
experiment is that in the simulation the clamping is modeled by fixing the degree of
1
According to Sader [39], calculated by Philipp Rust [38].
51
Chapter 4. Beam
Air Water
Qtotal Qtotal Qad
1st Bending Mode
Beam 1 234.1 62.18 84.67
Beam 2 312.3 38.78 44.28
Simulation B - - 83.30
Theory1 - - 165.0
nd
2 Bending Mode
Beam 1 3985 97.36 99.80
Beam 2 8201 75.27 75.97
Theory1 - - 472.73
rd
3 Bending Mode
Beam 1 6105 56.18 56.70
Beam 2 6153 44.97 45.30
Theory1 - - 1008.66
Table 4.3: Damping as Q-factor in air and water of the beams. Beam 1 is the exper-
imental result for the beam excited with an electromagnetic acoustic transducer, beam
2 is the experimental result for the beam excited with an electromagnetic coil and a
magnet and simulation B is a two-way coupling between a transient structural analysis
of the beam and a transient fluid dynamic analysis of the water. Qtotal (air, structure) is
the damping of the beam in air, Qtotal (water, structure) is the damping of the beam in
water and Qad (water) is the added damping caused by the water.
freedoms of the nodes of the upper and lower side of the beam in the hatched area, see
figure 4.1. Because of the clamping between the beam and the blocks, in the experiment
the eigenfrequency will be reduced. Comparing the relation between the eigenfrequency
in air and the eigenfrequency in water (fwater /fair ), the deviation between experiment
and simulation is below 4 %.
A mesh convergence study has been done for both simulations. Because of the com-
putational cost of a two-way coupling, a not fully converged mesh had to be used (see
figure 4.3). That is also the reason for the about 1 % higher eigenfrequencies comparing
simulation A with simulation B. Furthermore, the decision to approximate the clamping
by fixing the degree of freedom instead of modeling the clamping blocks is motivated by
the computational costs.
1
According to Sader [39], calculated by Philipp Rust [38].
52
The damping influence of the water is significant. This can be expressed by the added
damping which is the part of the damping due to the water. The added damping Qad
as a function of the water, can be calculated by the assumption that
1 1 1
= + (4.1)
Qtotal (water, structure) Qtotal (air, structure) Qad (water)
where Qtotal (water, structure) is the Q-factor in water and Qtotal (air, structure) is the
Q-factor in air.
Comparing the Q-factors in water for the experiment and the simulation, a discrep-
ancy can be seen. As the Q-factors of beam 1 and beam 2 themselves have no good
agreement, the discrepancy in the simulation is even larger but at least of the same
order of magnitude. One reason for the difference is the mesh being rather coarse due
to the computational expense. Furthermore, the simulated Q-factor shows a frequency
dependency which is visible in the experimental data as well.
From the present analysis, as already noted previously, it can be concluded that the
simulation of the eigenfrequency is state of the art. The simulation of the damping of a
surrounding fluid, however, is not satisfactory. Also with the method presented here it
is not possible to get really a good result. For a first guess it is usable. However, it can
be expected that a refined mesh, the inclusion of the clamping blocks into the simulation
and a larger number of experiments will lead to more accurate results.
53
54
5 Circular Plate
In this chapter an overview is given about some special aspects of the eigenmodes of
circular plates. It is important to mention, that only bending modes are considered.
In section 5.3 a literature review is given for the eigenfrequencies of submerged circular
plates. The experimental study of a submerged plate is presented in the subsequent
chapter 6.
5.1 Basics
5.1.1 Imperfections
Each eigenfrequency of an ideal, homogeneous, rotationally symmetric circular plate

with n nodal diameters has two mode shapes (mode pair) which are congruent but
rotated with respect to each other by /2n. The position of the nodal lines are free in
space around the circumference. For an externally excited oscillation of the plate, the
anti-nodes will align with the excitation.
A real plate has several reasons for being imperfect. There is the geometrical imperfec-
tion if for example the radius or the thickness is not constant or the plate is not plane.
There is the material imperfection if for example the material itself is not homogeneous.
Through production-related techniques the plate can be anisotropic for example in the
stiffness due to the direction of rolling.
As a result of these imperfections the mode pair divides into two single modes with
slightly different eigenfrequencies depending on how strong the imperfection is. In this
case the mode shapes are fixed in space depending on the imperfections but indepen-
dently of the location of the excitation. Already in 1898 Zenneck [49] reported about
this problem.
55
Chapter 5. Circular Plate
5.1.2 Mode Shapes
As a finite element simulation leads to imperfections, due to the mesh and the numerics
itself, the nodal diameters have a fixed position. In figure 5.1 the first mode shapes
of a FC circular plate made of stainless steel with a thickness of 2 mm and a radius of
190 mm are shown. For different conditions like size, material and clamping, the order
of the mode shapes can be different.
5.2 Rotating Mode Shapes
On a non rotating circular plate one can let rotate the mode shapes themselves. The
following topic was presented by the author in [21]. For a FC circular plate, the mode
shapes of modes with nodal lines can be rotated by using two excitation signals at
different locations at the outer edge with a certain phase and space difference. The
superposition of the waves which are generated by the two excitation locations can be
expressed by
w = ei(kx+t) + ei(kxt) + ei(k(xx0 )+t+) + ei(k(xx0 )t) (5.1)
where w is the normal displacement, k is the wave number, x is the location along
the circumference, is the circular frequency, t is the time and the phase difference
between the two excitation signals. For = n x0 (x0 in angular dimension and n as
an integer value), the superposition of the two standing waves result in a traveling wave
and can be observed in a rotating mode shape. The difference between a standing mode
shape and a rotating mode shape is shown in figure 5.2 for half a period of oscillation
from 0 to .
For the experiments, an aluminum circular plate was used with a radius of 220 mm and
a thickness of 2 mm. The plate is clamped in the middle. At each excitation point
a magnet was used to transfer the force from the coil which is driven by a frequency
generator to the plate. To measure the vibration the laser scanning vibrometer system
was used.
In addition a finite element simulation has been performed using ANSYS. In a fully
harmonic response analysis, forces at the two excitations points were applied with a
certain phase difference.
56
5.2. Rotating Mode Shapes
57
Figure 5.1: First 18 mode shapes of a FC circular plate. The dimensions and the
material are the same as for the plate described in chapter 6. The plate is made of
stainless steel, has a thickness of 2 mm and a radius of 190 mm. The color represents
the displacement of the mode shape perpendicular to the plate. The color map is from
negative displacement to zero displacement to positive displacement from blue to green
to red. The modes are addressed by counting the number of nodal lines (n) and nodal
circles (m). The apostrophe behind the mode number refers to the two existing mode
shapes of each mode with nodal diameters which are rotated with respect to each other
by /2n.
For an ideal and perfect circular plate, the frequency to get a rotating mode shape would
be the eigenfrequency of the mode pair. Because of the slightly different eigenfrequen-
cies of the mode pair through imperfections, the frequency in the middle of the two
corresponding eigenfrequencies has to be taken. The direction of the rotation depends
on the sign of the phase difference.
For the experiments, the phase difference between the two points has been fixed at
= 90 . Therefore, the angular distance between the two excitations point has to be
selected for the 2-0 mode as x x0 = 45 , for the 3-0 mode as x x0 = 30 and 4-0 mode
as x x0 = 22.5 . For the 5-0 mode = 45 and x x0 = 45 were used. The rotation
58
5.3. Submerged Circular Plate
of the mode shape is visualized by the software of the laser scanning vibrometer. All
listed mode shapes rotate as predicted.
5.3 Submerged Circular Plate
Not only rectangular plates submerged in water were studied in literature but also
circular plates. Bauer [5] investigated a clamped circular plate which fully covers a fluid
filled container, both theoretically and numerically. Later Amabili [1] made comparisons
to the work of Bauer and studied also sloshing modes in the case where the fluid filled
container is larger than the circular plate. Kwak and Han [29] studied the same problem
and made some experiments to validate their results. Amabili et al. [2] and Kwak and
Amabili [28] studied the eigenfrequencies of a fully submerged circular plate neglecting
the influence of the near bottom of their container and the free water surface theoretically
and experimentally. Jeong and Kim [23] investigated a fully submerged circular plate in
a closed container completely filled with fluid theoretically and numerically. Due to the
fixed size of the container the variation of the water depth below changes also the water
depth above the plate. For this reason they did not study the water depth on each side
independently. Esmailzadeh et al. [16] investigated a circular plate which is in contact
with a fluid on one side. The fluid has either a free surface or is covered with a rigid
plate. They computed the eigenfrequencies of the first three modes for different water
depths but did not validate their results with experiments.
As mentioned earlier, Askari et al. [4] published just recently a study about the vibration
of a circular plate submerged in a water-filled container. The experimental part is similar
to [20]. They studied analytically, experimentally and by simulation a FF circular plate
hold by thin wires. The container was filled with a constant water-level which makes a
dependency of the water-level above and the water-level below instead of varying only
one parameter. Furthermore the excitation has been done with a shaker which needs a
contact to the circular plate instead of using a contactless excitation method to minimize
undesired influences.
As the modes without nodal diameters are not much influenced by a clamping in the
middle of a circular plate, one can compare the results from Askari et al. with the ones
presented later by nondimensionalizing them also for the dimensions by dividing the
water-levels through the diameter of the plates. Comparing the nondimensionalized
eigenfrequencies one will find out that the values of Askari et al. are up to 2.1 % higher
59
than the ones presented here. This is due to the smaller submergence depth to diam-
eter ratio which leads to a higher eigenfrequency in water. When investigating effects
from nearby walls or free water surfaces one has to be careful to have also reference
configurations where these boundary conditions are negligible. One can also compare
the damping values and have to find out that the damping values presented by Askari
et al. are much higher (between 150 % and 1000 %) than the one in table 6.2. This can
be due to the influence of the shaker which was used for the excitation.
60
5.3. Submerged Circular Plate
Standing mode shape Rotating mode shape
1
4

1
2

3
4

Figure 5.2: Standing and rotating mode shape of the 2-0 mode of the circular plate
which is fixed in space. The color map is from negative displacement to zero displacement
to positive displacement from blue to green to red. The mode shapes are plotted for half
a period of oscillation from 0 to .
61
62
6 Experimental Rig
In this chapter the circular plate, introduced in chapter 5, is studied experimentally.

The results are compared with numerical simulations and analytical calculations. This
research examines the eigenfrequency and damping of submerged circular plates first at
different water-levels and second, with water rotating around the plate.
6.1 Design
The experimental rig, used here and shown in figure 6.1, is 1390 mm long and 1595 mm
high. It was previously used amongst others to test different labyrinth seals and has
been adapted to meet the experimental needs here. Significant changes were made to
the top part (figure 6.2). A new ring was designed to be able to separate the rotatable
ground plate from the vibrating plate and the cover by a distance of each 100 mm. The
separation is needed to investigate the water-levels influence of both sides of the vibrating
plate independently. The cover, made from acrylic glass (PMMA), is required for optical
vibrometer measurements of the vibrating plate from above. A rod is fixed to the top
side of the vibrating plate to accurately adjust the water depth. This is accomplished
by adjusting the clamping position on the rod at the PMMA cover. For the excitation,
a hole is drilled in the cover where a thin rod with a coil at its end rests. Finally, the
rig is filled from the side with water.
To have no temperature difference between the rig and the water, the used water was
stored in the same room. To avoid bubbles on the circular plate in the rig, water which
was stored for at least four weeks was used. The ground plate is driven over a belt with
an electric motor.
The circular vibrating plate has a diameter of 380 mm and a thickness of 2 mm (fig-
ure 6.3(a)). The hole in the middle of the vibrating plate has a diameter of 8.5 mm. The
plate is made of stainless steel and has a mass of 1.7967 kg. Two magnets were glued to
63
Chapter 6. Experimental Rig
Figure 6.1: Experimental rig - technical drawing in 2D
the vibrating plate near the outer edge with an angular distance of 45 coaxial to the
coils for the excitation. The cylindrical magnets of NdFeB with a diameter of 5 mm and
a height of 5 mm have a weight of 0.75 g each. As the thickness of the ground plate with
15 mm is much thicker than the vibrating plate and the stiffness of a plate is influenced
by the height with a power of three, the ground plate can be seen as a rigid wall in
comparison to the vibrating plate.
6.2 Tests with Partially Filled Rig
First, the experimental rig is used to investigate the influence of a nearby wall and a free
water surface on the eigenfrequency of a vibrating plate. For this test, the ground plate
64
6.2. Tests with Partially Filled Rig
Figure 6.2: Schematic detailed view of the top part of the experimental rig.
was not rotating. To simulate the experimental setup, a finite element modal analysis
was performed using ANSYS. The used model can be seen in figure 6.3(b).
6.2.1 Eigenfrequency and Damping in Air and Water
The measurements in this section were done with the Lock-In amplifier in order to
achieve a higher frequency resolution than with the laser scanning vibrometer system.
The results of the measurements in air and in water of the vibrating plate at a distance
to the ground plate of 100 mm and a water-level above the vibrating plate of 100 mm are
listed in table 6.1. The difference between the simulation and the experiment in air is
below 1 % except for some modes with nodal circles and some modes of a mode pair. In
water, not all modes can be detected from the measurements due to the high damping
65
(a) Dimensions of the FC vibrating plate (b) Model of the FC vibrating plate
Figure 6.3: The used FC vibrating plate inside the experimental rig. Subfigure (a)
dimensions for the vibration measurement and subfigure (b) model of the FC vibrating
plate. The fluid around the plate is supposed to have a cylindrical shape with a constant
radius for the full height instead of the more complicated shape in reality. The central
rod for the mounting is modeled as rigid. Below the vibrating plate a rigid central rod
is also modeled to achieve a well structured mesh.
of the water resulting in a small amplitude of some modes. The difference between the
simulation and the experiment in water is greater than for air but still below 5 %. Again,
the differences between simulation and experiment for the modes with nodal circles tends
to be higher than for the modes with only nodal diameters.
For the simulation, a mesh convergence study was performed while looking at the eigen-
frequency of the highest mode of interest, while refining the mesh of the solid as well as
of the fluid in radial, angular and longitudinal direction. The mesh is shown in figure 6.4.
The frequency reduction expressed by fwater /fair has an upward trend for an increasing
mode number, but different for modes with and without nodal circles (see coupling
effect 4 on page 32). For the modes with nodal circles the frequency reduction in the
simulation seems to be overestimated. That is also the reason why in the simulation
with water, the order of the modes changes a bit. As for radial runners the difference
of the frequency reduction between modes with and without nodal circles is even more
significant, the lowest modes have no nodal circles. That is the reason why modes with
only nodal diameters are investigated.
66
Air Water fwater

fair [Hz] fwater [Hz] fair
Exp. Sim. Exp. Sim. Exp. Sim.
Mode 1-0 32.03 32.46 11.95 11.96 0.373 0.368
Mode 1-0 32.34 32.46 - 11.96 - 0.368
Mode 0-0 48.44 51.20 - 13.03 - 0.254
Mode 2-0 72.97 72.64 31.68 31.49 0.434 0.434
Mode 2-0 74.69 72.64 32.45 31.49 0.434 0.434
Mode 3-0 169.4 168.2 81.20 80.97 0.479 0.481
Mode 3-0 169.7 168.2 84.44 80.97 0.498 0.481
Mode 0-1 - 294.0 - 119.7 - 0.407
Mode 4-0 296.3 295.3 152.6 153.0 0.515 0.518
Mode 4-0 301.6 295.3 154.3 153.0 0.512 0.518
Mode 1-1 316.9 318.3 - 148.4 - 0.466
Mode 1-1 320.3 318.3 - 148.4 - 0.466
Mode 5-0 453.4 452.9 246.8 248.4 0.544 0.548
Mode 5-0 454.1 452.9 247.5 248.4 0.545 0.548
Mode 2-1 473.4 477.9 257.0 240.2 0.543 0.503
Mode 2-1 475.9 477.9 - 240.2 - 0.503
Mode 6-0 640.5 640.4 377.6 367.7 0.590 0.574
Mode 6-0 640.6 640.4 377.9 367.7 0.590 0.574
Mode 3-1 711.1 716.6 397.3 381.5 0.559 0.532
Mode 3-1 711.7 716.6 - 381.5 - 0.532
Table 6.1: Experimental and simulated eigenfrequencies in air (fair ) and water (fwater )
of the vibrating plate with a distance to the ground plate of 100 mm and a water-
level (distance to the free water surface) above of 100 mm. The nondimensionalized
eigenfrequencies make the eigenfrequencies of the modes comparable.
Air Water
Qtotal Qtotal Qad
Mode 2-0 2206 185 202
Mode 2-0 2097 188 207
Mode 3-0 5279 302 320
Mode 4-0 8158 521 557
Mode 5-0 11536 632 669
Mode 6-0 10285 761 822
Table 6.2: Experimental damping as Q-factor in air and water of the vibrating plate
computed from the same measurements as the values from table 6.1.
67
Figure 6.4: The mesh of the FC vibrating plate from the experimental rig. The thick-
ness of the plate is resolved with three elements. As the hexagonal fluid elements are
not allowed to merge nodes, a hole in the middle of the plate is necessary for using a
spiderweb mesh. Therefore also on the lower side the same hole as on the upper side is
used. For the mesh 38 304 elements were used with a total of 164 928 nodes.
While looking at the damping of the vibrating plate in table 6.2, one can see a similar
damping for the mode pair and decreasing damping for higher modes in general. The
Q-factor for water is measured like before with a distance to the ground plate of 100 mm
and a water-level (distance to the free water surface) above of 100 mm. The significant
damping influence of the water Qad can be calculated with equation (4.1). One can see
that the added damping also increases for an increasing mode number.
6.2.2 Influence of the Water-level
There are two experimental studies that use the vertical adjustability of the experimental
rig. The first is an adjustment of the water-level (between 5 mm and 100 mm) above the
vibrating plate while maintaining constant water-level (100 mm) below. The second is
the opposite, where the water-level is instead adjusted below the plate and held constant
above at 100 mm.
The nondimensionalized eigenfrequency in water is calculated by dividing by the eigen-
frequency of the vibrating plate in air. This is shown in figure 6.5 for different water-levels
68
above and below the vibrating plate. The experimental () and simulation () results
are simultaneously plotted. Only modes with no nodal circles were selected to have a
better comparison and because of the industrial relevance which was highlighted before.
For the eigenmode with two nodal diameters, both mode shapes of the mode pair are
plotted to show their similar eigenfrequencies.
The agreement of the simulation with the experiment is very good. The simulation of
the frequency reduction is within 3 % except for the 6-0 mode. There, the difference is
up to 5 %.
In figure 6.5, one can see all the coupling effects presented earlier. The coupling between
the vibrating plate and boundaries is the weakest at a water-level of 100 mm on both sides
of the vibrating plate (the maximum for the experimental rig). Even at this distance the
boundaries still influence the first modes eigenfrequency. However, the higher frequency
modes are more localized and therefore the 100 mm of separation from the boundaries
is sufficient to consider no coupling.
The experimental measured damping values are plotted in figure 6.6. Unfortunately, the
reproducibility of the results is not good. The damping values measured at water-levels
of 100 mm are up to 12 % apart. However, the trend clearly indicates the damping is
increasing for decreasing distance to a wall. The trend for the water-level above the
vibrating plate changes from decreasing damping for decreasing water-level above for
the lower modes to increasing damping for decreasing water-level above for the higher
modes.
69
Mode 2-0 Mode 2-0
0.7 0.7
0.6 0.6
fwater /fair
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0 0.02 0.04 0.06 0.08 0.1 0 0.02 0.04 0.06 0.08 0.1
Mode 3-0 Mode 4-0
0.7 0.7
0.6 0.6
fwater /fair
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0 0.02 0.04 0.06 0.08 0.1 0 0.02 0.04 0.06 0.08 0.1
Mode 5-0 Mode 6-0
0.7 0.7
0.6 0.6
fwater /fair
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0 0.02 0.04 0.06 0.08 0.1 0 0.02 0.04 0.06 0.08 0.1
Water-level above/below [m] Water-level above/below [m]
Experimental variation of the water-level above

Experimental variation of the water-level below
Simulated variation of the water-level above
Simulated variation of the water-level below
Figure 6.5: Nondimensionalized eigenfrequency in water of the vibrating plate for dif-
ferent water-levels. A constant distance below and the variation of the water-level above
as well as a constant water-level above and the variation of the distance to the ground
plate below is shown for experiment and simulation.
70
Mode 2-0 Mode 2-0
200 150
Q-factor
150
100
100
50
0 0.02 0.04 0.06 0.08 0.1 0 0.02 0.04 0.06 0.08 0.1
Mode 3-0 Mode 4-0
400
600
300
Q-factor
400
200
200
100
0 0.02 0.04 0.06 0.08 0.1 0 0.02 0.04 0.06 0.08 0.1
Mode 5-0 Mode 6-0
1200 1000
1000
800
800
Q-factor
600
600
400 400
200 200
0
0 0.02 0.04 0.06 0.08 0.1 0 0.02 0.04 0.06 0.08 0.1
Water-level above/below [m] Water-level above/below [m]
Experimental variation of the water-level above

Experimental variation of the water-level below
Figure 6.6: Experimentally determined damping in water of the vibrating plate for
different water-levels. A constant distance below and the variation of the water-level
above as well as a constant water-level above and the variation of the distance to the
ground plate below is shown for experiment.
71
6.3 Tests with Brimful Rig
In this section the focus is on the vibration behavior of the vibrating plate with the
water rotating around it. The rotating free water surface is not smooth anymore. But
a smooth water surface is necessary to measure the vibration of the vibrating plate
with the laser scanning vibrometer system. Therefore, the experimental rig was filled to
capacity, leaving no air between the acrylic glass cover and the water surface.
6.3.1 Influence of the Water-level
As opposed to the section before, now there is no free water surface above the vibrating
plate. There is a wall that can be modeled just as the boundary below the plate.
Experiments have shown that reducing the distance of the vibrating plate to the acrylic
glass cover on top has the same influence on the eigenfrequencies of the vibrating plate
as shown before while reducing the distance of the vibrating plate to the ground plate.
For a circular water flow around the vibrating plate, the eigenfrequencies also decrease
for reducing the distance of the vibrating plate to a wall, no matter if the wall is above
or below.
6.3.2 Influence of a Circular Water Flow
In this subsection the circular water flows influence on the eigenfrequencies of a vibrating
plate is investigated. Therefore, the ground plate of the experimental rig is set into
rotation with a controllable speed from 0 rpm up to 500 rpm. The rotation of the ground
plate causes water inside the top part of the experimental rig to rotate. Only the
rotational speed of the ground plate is known accurately. Between the vibrating plate
and the acrylic glass cover the water also rotates but with a smaller rotational speed. For
the experimental investigation, the vibrating plate has a distance to the ground plate
and the acrylic glass cover of each 100 mm.
Figures 6.7, 6.8 and 6.9 show the Bode plots for the 2-0, 3-0, 4-0, 5-0 and 6-0 modes.
The Bode plot for the 2-0 mode differs from the others because also for no rotation
both eigenfrequencies of the mode pair can be seen with a slightly different frequency
and nearly the same amplitude. The resonances of low frequency modes could not be
detected because of significant low frequency noise at higher rotational speeds. For the
other modes without rotation, there is only one peak in the amplitude plot and a phase
72
6.3. Tests with Brimful Rig
102 Mode 2-0

8 0 rpm
80 rpm
Amplitude [m/s]
6 160 rpm
240 rpm
320 rpm
4
0
180
90
Phase [ ]
90
180
30 30.5 31 31.5 32 32.5 33
Frequency [Hz]
Figure 6.7: Bode plot for the 2-0 mode of the vibrating plate with rotating water caused
by the ground plates rotation ranging between 0 rpm and 320 rpm.
shift of about 180 . An exception is the 5-0 mode, where there is a phase shift of 360
even though only one peak is visible in the amplitude plot. When water rotates around
the vibrating plate, the single peak in the amplitude plot splits into two peaks. The
difference between these two peaks has an upward trend with increasing rotational speed.
In the phase plot, the single 180 shift splits into two 180 shifts at different frequencies.
This frequency difference also has an upward trend with increasing rotational speed.
For the 3-0, 4-0, 5-0 and 6-0 mode, the two peak frequencies of the split mode are
plotted in figure 6.10 at different rotational speeds. The plotted peak frequencies are
nondimensionalized by dividing by the eigenfrequency when there is no water rotation.
The Q-factor of these modes are also plotted in figure 6.10. The plotted Q-factor at
different rotational speeds is nondimensionalized by dividing by the Q-factor at no water
rotation.
In figure 6.10, one can see that the difference between the lower and upper frequency of
the split mode has an upward trend for increasing rotational speeds. It is also impor-
tant to note that both the lower and the upper frequency concave down for increasing
rotational speed. It can be observed that the relative frequency difference between the
73
Mode 3-0
0.12 0 rpm
80 rpm
0.1
Amplitude [m/s]
160 rpm
0.08 240 rpm
320 rpm
0.06 400 rpm
0.04 480 rpm
0.02
0
180
90
Phase [ ]
90
180
77 78 79 80 81 82 83 84
Frequency [Hz]
102 Mode 4-0

0 rpm
4 80 rpm
Amplitude [m/s]
160 rpm
3 240 rpm
320 rpm
2 400 rpm
480 rpm
1
0
180
90
Phase [ ]
90
180
148 149 150 151 152 153 154 155 156
Frequency [Hz]
Figure 6.8: Bode plot for the 3-0 and 4-0 mode of the vibrating plate with rotating
water caused by the ground plates rotation ranging between 0 rpm and 480 rpm.
74
Mode 5-0
0.1 0 rpm
80 rpm
Amplitude [m/s]
0.08 160 rpm

240 rpm
0.06 320 rpm
400 rpm
0.04 480 rpm
0.02
0
180
90
Phase [ ]
90
180
242 243 244 245 246 247 248 249 250 251
Frequency [Hz]
102 Mode 6-0

4 0 rpm
80 rpm
Amplitude [m/s]
3 160 rpm
240 rpm
320 rpm
2 400 rpm
480 rpm
1
180
90
Phase [ ]
0
90
180
359 360 361 362 363 364 365 366 367 368 369
Frequency [Hz]
Figure 6.9: Bode plot for the 5-0 and 6-0 mode of the vibrating plate with rotating
water caused by the ground plates rotation ranging between 0 rpm and 480 rpm.
75
Frequency (rel.) 1.02 1.6

1.4
Q-factor (rel.)
Mode 3-0
1 1.2
1
0.98
0.8
0.6
0.96
0 160 320 480 0 160 320 480
1.8
1.02
1.6
Frequency (rel.)
Q-factor (rel.)
1.4
Mode 4-0
1
1.2
1
0.98
0.8
0.6
0.96
0 160 320 480 0 160 320 480
1.02 1.1
Frequency (rel.)
1
Q-factor (rel.)
Mode 5-0
1
0.9
0.8
0.98
0.7
0.96
0 160 320 480 0 160 320 480
1.02
1.6
Frequency (rel.)
Q-factor (rel.)
Mode 6-0
1 1.4
1.2
0.98
1
0.96
0 160 320 480 0 160 320 480
Rotation [rpm] Rotation [rpm]
Lower Frequency Upper Frequency Average Frequency
Figure 6.10: Resonance frequency and damping of the vibrating plate with water
around at different speed which has a circular coaxial flow. The resonance frequency as
well as the damping is relative to the values without rotation of the water.
76
lower and upper frequency has a downward trend for an increasing mode number. The
Q-factor for the lower frequency as well as for the upper frequency has a downward trend
for increasing rotational speed.
Damping of structures placed in unidirectional flow has been investigated by Seeley et
al. [40]. They found that the flow affects not only the eigenfrequencies but also the
damping, which increases linearly with respect to the flow velocity.
It is important to note that the mode shape of the split mode is not fixed in space. The
lower frequency mode shape rotates in the direction opposite the waters rotation. The
upper frequency mode shape rotates in the same direction as the water. The speed of
the rotation is fres /n with n as the number of nodal diameters (compare section 5.2).
A physical explanation of the mode split could relate to the wave speed of the wave
inside the vibrating plate. The vibrating plate is excited at the outer edge. From there,
waves propagate concentrically inside the plate and are reflected at the clamping and at
the plates edge. At steady state, the shape will look like waves traveling in the clockwise
and counterclockwise direction. With no water rotation, there will be standing waves
corresponding to each of the vibrating plates eigenfrequencies.
With water rotation, parts of these waves which have a velocity component in the di-
rection of the waters velocity speed up and other parts of these waves which have a
velocity component in the opposite direction of the waters velocity speed down. As a
consequence, a wave moving in the counterclockwise direction, with a water rotation
in clockwise direction has a smaller wave speed than a wave in the clockwise direction
and vice versa. Because the vibrating plate is a periodic structure waves in the same
direction can constructively interfere with themselves. The interference leads to a peak
in the amplitude plot.
Because the wavelength is the same for both the waves but the wave speed is not,
the frequencies have to be different. The one which propagates in the opposite direction
of the water rotation has the lower frequency while the one propagating in the same
direction as the water rotation has the higher frequency. From this assumption, one
can also explain the increasing difference between the two resonance frequencies of the
split mode for an increasing water rotation speed. The increasing water rotation further
speeds up and down the waves in the corresponding directions. The relatively higher
wave speeds of the higher frequency modes, as compared to the water speed, reduces
this effect.
77
The difference in the wave speed of the two corresponding resonance frequencies should
be around two times the rotational speed. Since the water influence on the vibrating
plates eigenfrequencies is frequency dependent, the difference between the wave speeds
has a downward trend for higher frequency modes. A mathematical explanation is given
in the next paragraph.
Fluid-Conveying Pipe
In the book of Blevins [8] in chapter 10 the vibrations of a pipe containing a fluid flow
are described. He investigates a fluid-conveying pipe with pinned ends.
Figure 6.11: A deflected fluid-conveying pipe with pinned ends and length L [8].
He derives the following partial differential equation for a single fluid and pipe element:
4Y 2
2 Y 2Y 2Y
EI + Av + 2Av + M =0 (6.1)
x4 x2 xt t2
where Y (x, t) is the transverse deflection, E the Youngs modulus of the pipe and I its
area moment of inertia. The fluid density is , A is the internal area and v the velocity
of the fluid. M is the combined mass per unit length of the fluid and the pipe.
The first and last term in equation (6.1) are the pipes stiffness and inertia and are
independent of the fluid velocity. The second term is the centrifugal force resulting
from the acceleration of the fluid due to the curvature of the deformed pipe. This
terms form is identical to that of an axial compression which leads to a reduction in the
eigenfrequency. The third term, with the mixed derivative, results from a Coriolis force.
This term causes instabilities, under a pinned boundary condition and causes the mode
78
split for periodic boundary conditions. The pinned boundary conditions are expressed
by
Y (0, t) = Y (L, t) =0 (6.2)

2 2
Y Y
2
(0, t) = (L, t) = 0 (6.3)
x x2
where L is the length of the pipe in the pipe direction x with the general solution of the
governing equation
X4
Y = An ei(t+kn x) (6.4)
n=1
where kn are the solutions of the characteristic function
EIk 4 v 2 k 2 2vk m 2 = 0 (6.5)
including the frequency .

The periodic boundary conditions are expressed by
Y (0, t) = Y (L, t) (6.6)

Y Y
(0, t) = (L, t) (6.7)
x x
2Y 2Y
(0, t) = (L, t) (6.8)
x2 x2
3Y 3Y
(0, t) = (L, t) (6.9)
x3 x3
Using Wolfram Mathematica 8.0, the eigenfrequencies of the fluid-conveying pipe are
calculated for pinned (see figure 6.12) and periodic (see figure 6.13) boundary conditions
at velocities v ranging between 0 m/s and 65 m/s in steps of 1 m/s. The following
parameters describe the pipe and fluid.
Inner Diameter Outer Diameter Length Youngs modulus Density

[mm] [mm] [mm] [MPa] [kg/m3 ]
Pipe 10 10.2 1000 210 000 7600
Fluid - - - - 1000
Table 6.3: Material properties of the pipe and the fluid.
79
Solving equation (6.1) for the pinned boundary conditions will lead to eigenfrequencies
that have a downward trend with increasing v. The second term (the centrifugal term)
in equation (6.1) causes this downward trend. But the relative frequency decrease with
increasing v has a downward trend for an increasing mode number.
1 Mode 1
Mode 2
Frequency (rel.)
0.8 Mode 3
Mode 4
0.6 Mode 5
Mode 6
0.4
0.2
0 5 10 15 20 25 30 35 40 45 50 55 60 65
Speed [m/s]
Figure 6.12: Analytical solution of a fluid-conveying pipe for pinned boundary con-
ditions. For the speed v = 0, the eigenfrequencies of the first six modes are 89.80 Hz,
359.18 Hz, 808.16 Hz, 1436.7 Hz, 2244.9 Hz and 3232.6 Hz
Only the odd solutions of the pinned boundary conditions for v = 0 are solutions of
the periodic boundary conditions. Most notably, the eigenfrequencies split into a pair
for v 6= 0. This split has an upward trend for increasing v. But the relative frequency
difference has a downward trend for an increasing mode number.
The term with the mixed derivatives is responsible for the split. This splitting is ex-
plained physically by the Coriolis effect and mathematically by the coordinate transform
[48]. The average frequency of the split eigenfrequencies has a downward trend for in-
creasing v. The second term of equation (6.1) is responsible for the decrease of the
average frequency of the split eigenfrequencies. This is also frequency dependent, mean-
ing that for higher modes the decrease is slower. This is illustrated by the derivative of
the relative frequency.
Simulation
It is possible to simulate the eigenfrequencies of a rotating vibrating plate immersed in

still water. Krey et al. [26] describe the coupling of stationary and rotating reference
frames in the time domain with acoustic elements which are partly in the stationary
reference frame and partly in the rotating reference frame without using CFD. This is
80
102
2
Frequency (rel.) 1
Derivative (rel.)
1.5
Mode 1
0
1
0.5 1
0 2
0 20 40 60 0 20 40 60
102
2
1
Frequency (rel.)
Derivative (rel.)
1.5
Mode 2
0
1
0.5 1
0 2
0 20 40 60 0 20 40 60
102
2
1
Frequency (rel.)
Derivative (rel.)
1.5
Mode 3
0
1
0.5 1
0 2
0 20 40 60 0 20 40 60
Speed [m/s] Speed [m/s]
Lower Frequency Upper Frequency Average Frequency
Figure 6.13: Analytical solution of a fluid-conveying pipe for periodic boundary condi-
tions. For the speed v = 0, the eigenfrequencies of the first three modes are 359.18 Hz,
1436.7 Hz and 3232.6 Hz
not possible using only ANSYS. A specialized code is needed for this type of analysis,
which is beyond the scope of this thesis.
81
82
7 Conclusions and Outlook
7.1 Conclusions
The industrial relevance of the eigenmode analysis of structures surrounded by a fluid is

very high. The eigenmode analysis of different structures has been covered within this
thesis.
The study started with the theoretical background for modeling the eigenfrequency of a
submerged structure as well as the necessary techniques for the experimental determina-
tion of them. Coupling effects were elaborated which occur if two identical CF circular
plates are coupled with fluid.
In the following for different submerged rectangular plates and different clamping meth-
ods the eigenfrequencies and damping parameters were experimentally obtained and
compared with analytical values as well as values from the literature. These values have
been used to validate a finite element simulation using ANSYS.
For a submerged beam, the damping influence caused by the viscosity of the fluid has
been simulated. Within a two-way coupled transient simulation between CFD and
the structural analysis, damping values which are in the same order of magnitude are
achieved in comparison to the experimentally determined values.
As water turbines are quite complex structures, for the next step circular plates were
used because of their similarity to the turbine with respect to its periodicity. The
consequences of imperfections for circular plates were highlighted.
With the experimental rig, the vibration behavior due to the variation of the distance to a
free water surface as well as for the distance to a wall of the circular plate was tested. The
results are in good agreement with the values obtained from the finite element simulation
using ANSYS. It became apparent that using the acoustic elements for modeling the fluid
around a structure is a reliable tool for obtaining the eigenfrequencies of a submerged
structure. Also nearby walls, a free water surface, and other structures can be considered.
83
Chapter 7. Conclusions and Outlook
For a circular fluid flow around a stationary circular plate a mode split could be observed.
The frequency difference of the splitted mode is dependent on the rotation speed of the
fluid around.
During the time working at this thesis the eigenfrequency analysis of water turbines
using the finite element simulation with the acoustic elements for modeling the fluid
has become an essential part in the design process of new water turbines due to better
hardware, improved software, and the validation of the simulation with experimental
data.
7.2 Outlook
To accurately predict the eigenfrequencies of a water turbine it will be important in the

future to consider for a simulation the water turbine as a system. Therefore the rotation
of the runner has to be considered together with the movement of the water. Also the
covers and the shaft with the bearings have to be included in the simulation.
As a first step the eigenfrequency and the damping factor of a circular plate with a
rotating fluid around should be simulated. Current software is not able to predict Q-
factors with sufficient accuracy, even without rotation. Mode splitting and Q-factors
with moving fluids need special attention in the future.
High Q-factors for higher modes need special attention, as they might be relevant for
the propagation of fatigue cracks.
In a second step more complex coupling effects between structure and fluid should be
investigated. In this work only the axial gap of a circular plate was addressed. However
the influence of a radial gap should be considered as well.
84
A Photos of Experimental Setups
Utilized for the Present Work
A.1 Experimental Setup for the Rectangular Plate:

Photos
Figure A.1: From the top view into the yellow box one can see the rectangular plate
which is held by four rubber bands each at one edge of the plate. The coil from the top
is used to excite the plate. The power for the coil is coming from the amplifier.
85
Appendix A. Photos of Experimental Setups Utilized for the Present Work
Figure A.2: Experimental setup for the vibration measurement of the FFFF plate.
Inside the yellow box the plate is placed. The laser scanning head is mounted on a
tripod. Besides the box there is the silver colored amplifier to amplify the signal from the
signal generator within the laser scanning vibrometer system. The PC, the demodulator
and the junction box of the system is arranged in the rack.
86
A.2. Experimental Setup for the Beam: Photos
A.2 Experimental Setup for the Beam: Photos
Figure A.3: The beam is clamped between the two solid blocks which are pressed
together with screws. At the tip of the beam the electromagnetic coil excites the beam
with a magnet which is glued at the lower side of the beam.
Figure A.4: The same clamping as used for the excitation with the electromagnetic
coil is used also for the excitation with an EMAT. The EMAT is positioned at the tip
of the beam such that the Lorentz force is acting up and down.
87
A.3 Experimental Setup for the Circular Plate: Photos
Figure A.5: Front view of the FC circular plate. The plate is mounted on a frame of
aluminum profiles. The coils are also mounted on this.
Figure A.6: Top view of the FC circular plate. Magnets on both sides were used to
avoid gluing the magnet to the plate. Coils at several positions at the outer edge were
placed to investigate different configurations and modes.
88
A.4. Experimental Rig: Photos
A.4 Experimental Rig: Photos
Figure A.7: Detailed view of the glued magnet and one of the coils. This particular coil
was used for sensing the vibration. Therefore a thin wire was used with many windings.
89
Figure A.8: View of the main part of the experimental rig during the measurements
with the partially filled rig. Therefore the level indicator was mounted on the side. An
ultrasonic sensor (A) by BAUMER was used to measure the water-level.
90
A.4. Experimental Rig: Photos
Figure A.9: Side view of the complete experimental rig with the engine in the front
and the measurement cylinder behind. The rig is mounted on three vibration-cushioned
mounting feet.
91
Figure A.10: Detailed view of the top of the experimental rig. For the experiments
with the brimful rig only one coil (A) was used, also to have less disturbance of the
water while it is rotating. The water-level indicator (B) is mounted on top to ensure a
brimful rig.
92
References
[1] Amabili, M. Vibration of circular plates resting on a sloshing liquid: solution of

the fully coupled problem. Journal of Sound and Vibration 245, 2 (2001), 261283.
[2] Amabili, M., Dalpiaz, G., and Santolini, C. Free-edge circular plates vi-
brating in water. Modal Analysis: the International Journal of Analytical and
Experimental Modal Analysis 10, 3 (1995), 187202.
[3] Ansys. Ansys 14.0 Help.
[4] Askari, E., Jeong, K.-H., and Amabili, M. Hydroelastic vibration of circular
plates immersed in a liquid-filled container with free surface. Journal of Sound and
Vibration 332, 12 (2013), 3064 3085.
[5] Bauer, H. Coupled frequencies of a liquid in a circular cylindrical container with
elastic liquid surface cover. Journal of Sound and Vibration 180, 5 (1995), 689704.
[6] Best, R. E. Phase-locked loops: theory, design, and applications, vol. 1. McGraw-
Hill New York, 1984.
[7] Blevins, R. D. Formulas for natural frequency and mode shape, vol. 14. Van
Nostrand Reinhold NewYork, 1979.
[8] Blevins, R. D. Flow-Induced Vibration, 2nd ed. Van Nostrand Reinhold, 1990.
[9] Boyd, S. Multitone signals with low crest factor. Circuits and Systems, IEEE
Transactions on 33, 10 (1986), 10181022.
[10] Brennen, C. A review of added mass and fluid inertial forces. Tech. rep., Depart-
ment of the Navy , Port Hueneme, CA, USA, 1982.
[11] den Hartog, J. P. Mechanical vibrations. DoverPublications. com, 1956.
[12] Dobbs, E. R. Electromagnetic generation of ultrasonic waves. In Physical Acous-
tics, W. P. MASON and R. THURSTON, Eds., vol. 10 of Physical Acoustics. Aca-
demic Press, 1973, pp. 127 191.
93
References
[13] Du Buat, P. L. G. Principes DHydraulique: Verifies par un grand nombre

dExperiences faites par ordre du Gouvernement: Ouvrage dans lequel on traite du
mouvement uniforme & varie de leau dans les Rivieres, les Canaux, & les Tayaux
de conduite... Imprimerie de Monsieur, 1786.
[14] Dubas, M. Uber die Erregung infolge der Periodizitat von Turbomaschinen. In-
genieur Archiv 54, 6 (1984), 413426.
[15] Elmer, F.-J., and Dreier, M. Eigenfrequencies of a rectangular atomic force
microscope cantilever in a medium. Journal of Applied Physics 81, 12 (1997), 7709
7714.
[16] Esmailzadeh, M., Lakis, A., Thomas, M., and Marcouiller, L. Three-
dimensional modeling of curved structures containing and/or submerged in fluid.
Finite Elements in Analysis and Design 44, 67 (2008), 334345.
[17] Ewins, D. J. Modal testing: theory, practice and application, vol. 2. Research
studies press Baldock, 2000.
[18] Fu, Y., and Price, W. Interactions between a partially or totally immersed
vibrating cantilever plate and the surrounding fluid. Journal of Sound and Vibration
118, 3 (1987), 495 513.
[19] Haddara, M., and Cao, S. A study of the dynamic response of submerged
rectangular flat plates. Marine structures 9, 10 (1996), 913933.
[20] Hengstler, J., and Dual, J. Fluid structure interaction of a vibrating circular
plate in a bounded fluid volume: Simulation and experiment. In Fluid Structure
Interaction VI (2011), WIT Press.
[21] Hengstler, J., and Dual, J. Rotating eigenmodes on an imperfect circular
plate. In Proceedings of the 23rd International Congress on Theoretical and Applied
Mechanics, Beijing, China (2012).
[22] Iversen, H. W., and Balent, R. A correlating modulus for fluid resistance in
accelerated motion. Journal of Applied Physics 22, 3 (1951), 324328.
[23] Jeong, K.-H., and Kim, K.-J. Hydroelastic vibration of a circular plate sub-
merged in a bounded compressible fluid. Journal of Sound and Vibration 283, 1-2
(2005), 153172.
[24] Jhung, M., Choi, Y., and Jeong, K. Fluid bounding effect on natural frequen-
cies of fluid-coupled circular plates. Journal of Mechanical Science and Technology
17, 9 (2003), 12971315.
94
References
[25] Keck, H., and Sick, M. Thirty years of numerical flow simulation in hydraulic
turbomachines. Acta mechanica 201, 1-4 (2008), 211229.
[26] Krey, F., Hubner, B., and Seidel, U. Transient simulation of a pump-turbine
using coupled vibro-acoustic finite element analyses to investigate rotor-stator in-
teraction phenomena. In 4th International Meeting on Cavitation and Dynamics
Problems in Hydraulic Machinery and Systems (2011).
[27] Kundu, P., and Cohen, I. Fluid Mechanics. Elsevier Science, 2010.
[28] Kwak, M., and Amabili, M. Hydroelastic vibration of free-edge annular plates.
Journal of vibration and acoustics (1999).
[29] Kwak, M., and Han, S. Effect of fluid depth on the hydroelastic vibration of
free-edge circular plate. Journal of Sound and Vibration 230, 1 (2000), 171185.
[30] Leissa, A. W. Vibration of plates. Tech. rep., DTIC Document, 1969.
[31] Liang, C.-C. C., Liao, C.-C. C., Tai, Y.-S. S., and Lai, W.-H. H. The free
vibration analysis of submerged cantilever plates. Ocean Engineering 28, 9 (2001),
12251245.
[32] Liang, Q. W., Rodrguez, C., Egusquiza, E., Escaler, X., Farhat, M.,
Avellan, F., and Rodriguez, C. G. Numerical simulation of fluid added mass
effect on a francis turbine runner. Computers & Fluids 36, 6 (2007), 11061118.
[33] Lindholm, U. S., Kana, D. D., Chu, W.-H., and Abramson, H. N. Elastic
vibration characteristics of cantilever plates in water. Journal of Ship Research 9
(1965), 1122.
[34] Marburg, S. Discretization requirements: How many elements per wavelength
are necessary? In Computational Acoustics of Noise Propagation in Fluids - Finite
and Boundary Element Methods. Springer Berlin Heidelberg, 2008, pp. 309332.
[35] Marcus, M. S. A finite-element method applied to the vibration of submerged
plates. Journal of Ship Research 22, 2 (1978), 9499.
[36] Ohashi, H. Case Study of Pump Failure Due to Rotor-Stator Interaction. Inter-
national Journal of Rotating Machinery 1, 1 (1994), 5360.
[37] Rodriguez, C. G., Egusquiza, E., Escaler, X., Liang, Q. W., and Avel-
lan, F. Experimental investigation of added mass effects on a Francis turbine
runner in still water. Journal Of Fluids And Structures 22, 5 (2006), 699712.
95
References
[38] Rust, P. Micromachined Viscosity Sensors for the Characterization of DNA So-
lutions. PhD thesis, Diss., Eidgenossische Technische Hochschule ETH Zurich, Nr.
21299, 2013, 2013.
[39] Sader, J. Frequency response of cantilever beams immersed in viscous fluids with
applications to the atomic force microscope. Journal of Applied Physics 84, 1 (7
1998), 6476.
[40] Seeley, C., Coutu, A., Monette, C., Nennemann, B., and Marmont,
H. Characterization of hydrofoil damping due to fluidstructure interaction using
piezocomposite actuators. Smart Materials and Structures 21, 3 (2012), 35027
35035.
[41] Sigrist, J.-F. Symmetric and non-symmetric formulations for fluid-structure in-
teraction problems: Reference test cases for numerical developments in a commer-
cial finite element code. In ASME Pressure Vessel and Piping, Vancouver (2006).
[42] Sinha, J. K., Singh, S., Rao, A. R., and Rama Rao, a. Added mass and
damping of submerged perforated plates. Journal Of Sound And Vibration 260, 3
(2003), 549564.
[43] Stutt, C. A. Low-frequency spectrum of lock-in amplifiers. Tech. rep., Mas-
sachusetts Institute of Technology. Research Laboratory of Electronics, 1949.
[44] Tanaka, H. Vibration Behaviour and Dynamic Stress of Runners of very high
Head Reversible Pump-Turbines. In IAHR Symposium Belgrad Yugoslavia (1990).
[45] Thompson, R. B. Physical principles of measurements with EMAT transducers.
Physical acoustics 19 (1990), 157200.
[46] Van Eysden, C. a., and Sader, J. E. Resonant frequencies of a rectangular
cantilever beam immersed in a fluid. Journal Of Applied Physics 100, 11 (2006),
1149161149168.
[47] Vu, V., Thomas, M., Lakis, A., and Marcouiller, L. Effect of added mass
on submerged vibrated plates. In Proceedings of the 25 th Seminar on machinery
vibration, Canadian Machinery Vibration Association CMVA (2007), vol. 7.
[48] Wauer, J. Kontinuumsschwingungen: vom einfachen Strukturmodell zum kom-
plexen Mehrfeldsystem. Vieweg + Teubner Verlag, Wiesbaden (2008).
[49] Zenneck, J. Ueber die freien Schwingungen nur annahernd vollkommener
kreisformiger Platten. Annalen der Physik und Chemie 303, 1 (1899), 165184.
96

Eth 7837 02 PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Eth 7837 02 PDF

Uploaded by

Copyright:

Available Formats

Diss. ETH No.

Influence of the Fluid-Structure Interaction

accepted on the recommendation of

2 System Description, Theory and Working Principle 7

7 Conclusions and Outlook 83

Appendix A Photos of Experimental Setups Utilized for the Present Work 85

2.1 Detailed view of an electromagnetic coil and an EMAT . . . . . . . . . . 13

3.1 Dimensions and model of the FFFF rectangular plate . . . . . . . . . . . 34

4.1 Beam and its dimensions . . . . . . . . . . . . . . . . . . . . . . . . . . . 45

6.1 Experimental rig - technical drawing in 2D . . . . . . . . . . . . . . . . . 64

A.1 Detailed view of the experimental setup of the FFFF plate . . . . . . . . 85

3.1 Blevins table . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36

4.1 Material properties of the beams . . . . . . . . . . . . . . . . . . . . . . 46

6.1 Eigenfrequencies in air and water of the vibrating plate . . . . . . . . . . 67

1.1 Background and Motivation

1.2 Fluid-Structure-Interaction of Water Turbines:

1.3 Scope and Goal of the Present Work

2.1 Fluid-Structure Interaction

2.1.1 Compressible Fluid

A set of assumptions, used for example for ideal gases, are:

" #( ) " #( ) " #( ) ( )

Added Mass Effect

2.1.2 Incompressible Fluid

2.2 Modal Testing

Modal testing is the characterization of the dynamical behavior of a vibratory structure.

Electromagnetic Acoustic Transducer (EMAT) If the structure is not ferromagnetic

(a) electromagnetic coil (b) EMAT

Laser One of these non-contacting measurement methods is an optical method using a

Figure 2.2: Schematic diagram of a heterodyne interferometer with a Bragg cell as an

Figure 2.3: Laser scanning vibrometer setup

Figure 2.6: Vibration measurement setup with a Lock-In amplifier

2.2.2 Finite Element Simulation

Source Code Comment

Modal Analysis To calculate the eigenfrequencies, a modal analysis is performed. This

Harmonic Analysis To get an amplitude and phase spectrum, a harmonic analysis is

To assign the eigenfrequencies from a modal analysis to a certain mode shape it is

2.3 Coupling Effects caused by the Fluid-Structure

(a) out-of-phase mode (b) in-phase mode

(1) in-phase mode (2) free water surface

(3) out-of-phase mode (4) wall

Mode 1-0 Mode 2-0 Mode 3-0 Mode 4-0

In-phase mode Free water surface Wall Out-of-phase mode

Mode 1-0 Mode 2-0 Mode 3-0

In-phase mode Free water surface Wall Out-of-phase mode

Mode 1-0 out-of-phase

Mode 1-0 in-phase

Mode 2-0 out-of-phase

Mode 2-0 in-phase

Mode 5-0 out-of-phase

Mode 5-0 in-phase

Mode 6-0 out-of-phase

Mode 6-0 in-phase

Extended Coupling Effects1

3.1 Plate without clamping

As mentioned in subsection 2.2.1, the experimental realization of a plate without clamp-

Mode 1-1 Mode 2-0 Mode 2-1

Mode 0-2 Mode 3-0 Mode 1-2

Mode 3-1 Mode 2-2 Mode 4-0

3.1.1 Eigenfrequency in Air

3.1.2 Eigenfrequency in Water

Experiment Analytical Simulation fSim fExp

Experiment Simulation fSim fExp

Table 3.3: Experimental and simulated eigenfrequencies in water with a water-level of