Académique Documents
Professionnel Documents
Culture Documents
A Dissertation
Doctor of Philosophy
by
Wenlong Zhang,
Abstract
by
Wenlong Zhang
This dissertation studied the interaction of nonuniform flows with propeller blades
in a submerged elastic duct. The acoustic radiation from the duct is calculated and correlated to the flow nonuniformities and the propeller and duct characteristics. First,
a benchmark problem is studied wherein the sound radiation from an infinite plate
with or without ribs is examined for different excitations sources: normal single force,
monopole, dipole, and vortex excitations. The investigation of the sound radiation from
a plate gives us a fundamental understanding of flexure waves.
Second, the case of a cylindrical duct with or without ribs is considered and the dispersion relation of the rib-stiffened duct modes is compared with that of a un-stiffened
duct. The dispersion relation of the stiffened duct has a periodic structure similar to
that of connected oscillators with large number of independent modes. Because of our
interest in the acoustic radiation from such a system, we focus our attention on the
flexure modes. The sound radiation is first tested with simple internal forces such as
monopoles and dipoles. The results for un-stiffened ducts show strong directivity as the
dipole radial location moves closer to the duct wall. For stiffened ducts, the magnitude
of the acoustic response as well as the directivity vary strongly and show large peaks
near the stiffened duct free modes.
Wenlong Zhang
Third, the scattering phenomena in a rigid duct and an elastic duct is investigated.
The effect of the impedance on the acoustic sources is also examined. The results show
that the impedance begin to have the significant effect on the unsteady lift when the
magnitude of the non dimensional impedance is the order of one. The main effect of
the elastic wall comes from the location of the blade and the upstream of the blade.
Finally, a model for flow-propeller interactions in a submerged elastic duct is developed. This model examines and quantifies the mechanism of flow-propeller interaction
in a flexible duct. The model couples the fluid motion with the elastic duct vibration
and yields the duct flexural displacement. This leads to the evaluation of the radiated
sound. The coupling between the elastic duct and the flow-propeller system is studied
by changing the Euler code which accounts for the rotor/stator interaction problem. The
results suggests that for different combination of rotor/stator blade counts, it is possible to have low circumferential mode number, which is an efficient radiator of acoustic
energy.
CONTENTS
FIGURES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
TABLES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xv
ACKNOWLEDGMENTS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xvi
CHAPTER 1: INTRODUCTION . .
1.1 Structural Acoustics . . . .
1.1.1 Thin Plate . . . . .
1.1.2 Thin Shell . . . . .
1.2 Hydrodynamics . . . . . .
1.3 Fluid-Structure Interaction
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
. 1
. 4
. 4
. 6
. 12
. 15
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
18
18
19
21
23
27
29
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
32
34
35
38
41
43
45
45
47
47
ii
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
3.6
3.7
Vortex . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
iii
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
. 57
. 58
. 59
. 60
. 68
. 70
.
.
.
.
.
.
.
.
71
81
84
87
90
92
93
99
. 99
. 102
. 103
. 107
. 110
. 115
. 115
. 117
. 118
.
.
.
.
.
.
.
.
.
.
122
127
129
132
135
135
136
140
142
143
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
145
147
149
150
151
153
159
166
169
180
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
188
188
190
190
191
193
iv
FIGURES
1.1
Sound radiation from rotor blades interaction with turbulent and swirling
motion. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.2
. . . . . 16
2.1
3.1
3.2
3.3
Far field sound radiation from 1 cm steel plate with water loading on
one side only. The excitation is a unit point force at the origin. . . . . . 39
3.4
Far field sound radiation from 1 cm steel plate with water loading on
one side only. The excitation is a monopole at x0 = 0, y0 = 0 and
z0 = 0.05m. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
3.5
Far field sound radiation from 1 cm steel plate with water loading on
one side only. The excitation is a dipole at x0 = 0, y0 = 0 and z0 =
0.05m. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
3.6
Flexure waves of an infinite steel plate with ribs excited by a unit force.
The color scale represents the log10 displacement value in meters. . . . 46
3.7
Broadside far field sound level of steel plate stiffened by ribs. The
excitation is a unit point force between ribs. . . . . . . . . . . . . . . . 47
3.8
Broadside far field sound level of steel plate stiffened by ribs. The
excitation is a unit point force on a rib. . . . . . . . . . . . . . . . . . 48
3.9
Broadside far field sound level of steel plate stiffened by ribs. The
excitation is a unit point force between ribs. The mass of the ribs is five
times that of Figure 3.7. . . . . . . . . . . . . . . . . . . . . . . . . . 48
3.10 Broadside far field sound level of steel plate stiffened by ribs. The
excitation is a unit point force on a rib. The mass of the ribs is five
times that of Figure 3.8. . . . . . . . . . . . . . . . . . . . . . . . . . 49
3.11 Broadside far field sound level of steel plate stiffened by ribs. The
excitation is a monopole between ribs. . . . . . . . . . . . . . . . . . . 49
3.12 Broadside far field sound level of steel plate stiffened by ribs. The
excitation is a monopole on a rib. . . . . . . . . . . . . . . . . . . . . 50
3.13 Broadside far field sound level of steel plate stiffened by ribs. The
excitation is a dipole between ribs. . . . . . . . . . . . . . . . . . . . . 51
3.14 Broadside far field sound level of steel plate stiffened by ribs. The
excitation is a dipole on a rib. . . . . . . . . . . . . . . . . . . . . . . 51
3.15 Broadside far field sound level of steel plate stiffened by ribs. The
excitation is a vortex above the plate (z0 = 0.05m) . . . . . . . . . . . . 54
3.16 Far field ( = /4) sound level of steel plate stiffened by ribs. The
excitation is a vortex above the plate (z0 = 0.05m). . . . . . . . . . . . 55
4.1
4.2
4.3
4.4
4.5
4.6
4.7
4.8
4.9
vi
vii
viii
4.63 Far field sound pressure levels of the aluminum cylindrical shell caused
by radial dipoles at = /4 with different azimuthal angles. . . . . . . 110
4.64 Pressure directivity of a radial dipole located at axis of cylinder for
reduced frequency = 0.1. . . . . . . . . . . . . . . . . . . . . . . . 111
4.65 Pressure directivity of a radial dipole located at axis of cylinder for
reduced frequency = 1. . . . . . . . . . . . . . . . . . . . . . . . . 112
ix
5.2
5.3
Flexure waves of a water-filled Aluminum cylinder surrounded by water for circumferential harmonic m=0, which is excited by a unit force. . 128
5.4
5.5
5.6
5.7
5.8
5.9
6.1
6.2
6.3
Pressure magnitude distribution along the x-axis at the duct wall for
reduced frequency = 0.5. . . . . . . . . . . . . . . . . . . . . . . . 152
6.4
Unsteady sectional lift coefficient along the span for reduced frequency
= 0.5. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153
6.5
Pressure magnitude distribution along the x-axis at the duct wall for
reduced frequency = 1. . . . . . . . . . . . . . . . . . . . . . . . . 154
6.6
Unsteady sectional lift coefficient along the span for reduced frequency
= 1. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155
6.7
Pressure magnitude distribution along the x-axis at the duct wall for
reduced frequency = 3. . . . . . . . . . . . . . . . . . . . . . . . . 156
6.8
Unsteady sectional lift coefficient along the span for reduced frequency
= 3. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157
6.9
6.20 Pressure magnitude distribution along the x-axis at the duct wall for
reduced frequency = 3 (impedance i = 1.0 1.0i). . . . . . . . . . 168
6.21 Unsteady sectional lift coefficient along the span for reduced frequency
= 3 (impedance i = 1.0 1.0i). . . . . . . . . . . . . . . . . . . . 169
6.22 Pressure magnitude distribution along the x-axis at the duct wall for
different impedances ( = 0) . . . . . . . . . . . . . . . . . . . . . . . 170
6.23 A close up for the peak part of the pressure magnitude distribution along
the x-axis at the duct wall for different impedances ( = 0) . . . . . . . 170
6.24 A close up for the middle part of the pressure magnitude distribution
along the x-axis at the duct wall for different impedances ( = 0) . . . . 171
6.25 Pressure magnitude distribution along the x-axis at the duct wall for
different impedances ( = /10) . . . . . . . . . . . . . . . . . . . . . 171
6.26 Comparison of the magnitudes of the unsteady sectional lift coefficient
along the span for different acoustic impedances . . . . . . . . . . . . . 172
6.27 Far filed sound pressure directivity for an aluminum duct. The noise
sources are obtained from the rotor/stator duct system . . . . . . . . . . 173
6.28 Schematic of computational domain of the duct with impedance discontinuities. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 174
6.29 Pressure magnitude distribution along the x-axis at the duct wall of
impedance discontinuities ( = 0) . . . . . . . . . . . . . . . . . . . . 175
6.30 A close up for the peak part of the pressure magnitude distribution along
the x-axis at the duct wall of impedance discontinuities ( = 0) . . . . . 175
6.31 A close up for the middle part of the pressure magnitude distribution
along the x-axis at the duct wall of impedance discontinuities ( = 0) .
176
6.32 Pressure magnitude distribution along the x-axis at the duct wall of
impedance discontinuities( = /10) . . . . . . . . . . . . . . . . . . 176
6.33 Comparison of the magnitudes of the unsteady sectional lift coefficient
along the span of impedance discontinuities. . . . . . . . . . . . . . . 177
6.34 Imaginary part of exterior water loading on an infinite cylinder of radium 1 m and axial wavenumber = 0.5. The negative sign indicates
that it acts as a resistance. . . . . . . . . . . . . . . . . . . . . . . . . . 178
6.35 Imaginary part of exterior water loading on an infinite cylinder of radium 1 m and axial wavenumber = 5.0. The negative sign indicates
that it acts as a resistance. . . . . . . . . . . . . . . . . . . . . . . . . . 178
6.36 Real part of duct wall pressure in response to rotor stator interaction for
rigid duct. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181
xiii
6.37 Imaginary part of duct wall pressure in response to rotor stator interaction for rigid duct. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181
6.38 Real part of duct wall pressure in response to rotor stator interaction for
elastic duct. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182
6.39 Imaginary part of duct wall pressure in response to rotor stator interaction for elastic duct. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182
6.40 Comparison of unsteady lift coefficient between rigid duct and elastic
duct. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183
6.41 Comparison of sound pressure directivity for an aluminum duct between the rigid wall excitation and the elastic wall excitation. . . . . . . 183
xiv
TABLES
2.1
SCALING OF FORCES. . . . . . . . . . . . . . . . . . . . . . . . . . 27
4.1
6.1
xv
ACKNOWLEDGMENTS
xvi
NOMENCLATURE
Latin Symbols
A
Monopole strength
Shell radius
Youngs modulus
Excitation
Spectral excitation
Direction vector
Field quantity
Hankel function
Shell thickness
Bessel function
Wave number
Mach number
Normal direction
Number of points
Pressure (Pa)
Power (Watts)
Pressure eigenfunction
Entropy (kJ/kg/K)
Stiffness matrix
Time (s)
u, v, w
Velocity (m/s)
Spectral displacement
Angle in degrees (o )
Circumferential position
xviii
Angular spacing
Convected eigenvalue
Poissons ratio
Density (kg/m3 )
Potential (m/s2 )
Stagger angle (o )
Vortical velocity
Perturbation
Dimensionless value
Subscripts
Radial counter
Normal direction
Radial direction
Axial direction
Circumferential direction
xix
CHAPTER 1
INTRODUCTION
The interactions between fluid and flexible structures give rise to numerous physical problems and phenomena. In these cases, the flexible structure is surrounded by
the fluid, and fluid flow exerts pressure on the solid structure causing it to deform. For
nonuniform turbulent flow, the structure vibrates in response to the unsteady fluid loading and radiates sound in the surrounding fluid. When the interaction between nonuniform flow and a propeller in an elastic duct is studied, there are two main questions that
need to be answered: 1) How does the elastic duct change the fluctuating hydrodynamic
pressure along the propeller blades? 2) How does it affect the far field radiated sound?
To answer these questions and obtain a better understanding of fluid-structure interaction phenomena, we need to bridge the gab between hydrodynamics which assumes
rigid ducts and blades and structural acoustics which assumes flexible structures. The
coupling between the fluid motion and the duct vibration need to be modeled, which
takes place at the duct boundary.
In this work, the coupled nonuniform flow interaction with a propeller in an elastic duct is studied. Figure 1.1 schematically shows sound radiation from rotor blades
interaction with turbulent and swirling motion. As can be seen, the elastic duct with a
propeller is submerged and filled with water. The propeller is modeled as a rotor/stator
stage. For simplicity, the duct is modeled as an infinite thin cylindrical shell. The
propeller usually works in a nonuniform flow, and the non-uniformities are caused by
1
Figure 1.1. Sound radiation from rotor blades interaction with turbulent and
swirling motion.
ingested turbulence. To control vibration and noise, while at the same time enhancing
structural integrity, the cylindrical shell is stiffened with ribs.
The ingested turbulence interacts with the blades, and the blade forces can be regarded as dipole sources. Due to the swirling flow, the centrifugal force will act on the
duct too. The forces excite flexure waves in the duct. The radiated sound is produced
by duct vibration and hydro-acoustic scattering. The present research is directed toward
investigating the coupling effects between the fluid motion and the duct vibration and
to control these undesirable effects.
In the traditional approach of studying the flow-propeller-duct interaction, the duct
is assumed to be rigid and the hydrodynamic forces give the equivalent dipole strength.
Moreover, a compact blocked dipole model is used to calculate the radiated sound in the
far field. This approach does not account for the coupling effect between the flow and
the elastic duct which affects the strength of the hydrodynamic forces and the modeling
of the radiated sound. It also ignores the spatially distributed nature of the flow dipoles.
Our study of the sound pressure level radiated from a dipole inside a duct shows strong
dependence of the sound pressure level on the radial position of the dipole and its
orientation. The difference of sound pressure level between two radial positions of
a dipole can reach more than 10 dB. This fact demonstrates the inadequate assumption
often used in structural acoustics wherein distributed dipole sources are treated as a
single blocked dipole when the acoustic transfer function is calculated.
The main source of noise radiated from a submerged flow-propeller-duct system is
the interaction of irregular inflow disturbances with the propeller/vanes blades and the
duct wall. These irregular inflow patterns arise from various flow phenomena, such
as inlet distortion and turbulence, guide vanes-rotor interaction and second flows, etc.
The interaction of inflow disturbances with the blades produces unsteady hydrodynamic
forces (dipole sources) which, coupled with the duct flexural modes, results in unwanted
vibration and noise. For ducts with axial discontinuities, such as control ribs and finite
ducts, the dipole sources and flow vortices may couple with duct modes to produce
additional sources of noise.
The main objectives of this dissertation are to identify and quantify the mechanisms of the fluid-structure interaction relevant to noise radiation of submerged flowpropeller-duct, and develop predictive methods and tools to quantify and control the
interaction processes. Studying this interdisciplinary problem requires using different
fields including structural acoustics, hydrodynamics, fluid-structure interaction, vibration and control, etc.
quencies. DiPerna [23, 24] developed a new method to calculate the Greens function
for a fluid-loaded elastic plate, which is accomplished by introducing a rational function
approximation of the acoustic impedance. This method is numerically efficient and the
results agree very well with the calculations which utilized numerical integration procedures.
Structures, such as aircraft and submarines, are often stiffened with ribs (beams) for
structural integrity. There have been many studies of such structures, including that of
Evseev [27] who obtained a solution in closed form by using integral transformation.
Leppington [46] and Stepanishen [64] studied the sound radiation from thin plates on
which a plane sound wave is incident by using Fourier transforms. Crighton [21] gave
a comprehensive analytical description of acoustic and vibration phenomena associated
with the interaction between plane-wave and single rib on a fluid-loaded plate. Mace
[47, 48] developed a solution for the sound radiation from a point-excited infinite elastic plate stiffened by two sets of parallel stiffeners, bulkhead and frame. It was found
that the effects of stiffeners on the far field radiated pressure depend primarily on the
number of frames and the point at which the excitation is applied. Eatwell [25] developed expressions for the response of a fluid-loaded plate with ribs to a general force
distribution. It was shown that a plate stiffened with equally spaced ribs tends to produce the pass-bands and stop-bands, dependent on frequency and angle. But for the
plate stiffened with unequally spaced ribs, the structure of pass- and stop-bands may
not exist.
The vibration of thin plates is a two-dimensional problem. The study of the sound
radiation from a thin plate gives us a fundamental understanding of structural acoustics. More complicated acoustic and vibration phenomena will be observed from the
vibration of thin shells, which is a complex three-dimensional problem.
to zero. This algebraic equation is called the dispersion equation. By solving this dispersion equation, it can be found that there are three important propagating waves in
the shell which are corresponding to flexure wave, compression wave and torsion wave.
Dispersion curves are obtained by plotting wavenumber versus frequency, which can be
used to explain the acoustic radiation from the shell. In the paper of Brazier-Smith and
Scott [9], the authors used the winding number integrals to calculate the roots of the
dispersion equation. The results show that this method gives a high degree of accuracy,
and it is applicable to solution for the zeros of any analytic function.
In the present work, we have developed a general numerical approach to find the
roots of the dispersion relation. This is because the analytical method can not be used to
find the roots of the dispersion equation of the structure with complex geometry. In our
new approach, the displacements of the shell will be solved by applying a point force
to the system. The dispersion curves are obtained by plotting displacements versus
frequency and wavenumber.
Much research has been devoted to the understanding of wave propagation in shells
for two cases, in vacuo and in fluid filled. Fuller & Fahy [30] investigated wave propagation in cylindrical elastic shells filled with fluid. Dispersion curves for waves of
circumferential number m = 0 and m = 1 were derived and branches in the real, imaginary and complex planes were found. The behavior of free modes was found to depend
strongly on the thickness of the shell wall, and on the ratio of the density of the shell materials to the density of the contained fluid. Further analysis of free propagating waves
in an infinite elastic duct was presented by Scott [62]. In his paper, the free modes of
propagation of an infinite fluid-loaded thin cylindrical shell has been studied and the
general properties of the roots of the dispersion relation has been derived. The roots
were identified and described as the compressional root, shear root(torsional root), and
Hodges et al. [37, 38] developed a theory of vibration of a cylindrical shell stiffened
by circular T-section ribs inside the shell for low frequency. The governing equations
of the shell with ribs were derived by considering the potential and kinetic energies
of the shell and ribs. The total potential energy was represented by the contribution
from the shell and the ribs. And the rib potential energy was decomposed into the antisymmetric and symmetric contributions. The kinetic energy was similarly calculated
as the potential energy. The vibratory behavior of the ribbed shell was studied. For
the frequency range from zero to three times the ring frequency, good agreement were
obtained between the theoretical modeling and the experiments.
For high frequency, Vasudevan[66] developed a general formulation for a shell stiffened by T-section ribs inside the shell. The equations of motion of the shell were developed using the Love-Timoshenko strain relations. And the equations of motion of
the ribs were obtained by representing the properties of the ribs at the centroid of its
cross-sectional area. The numerical results of sound radiation showed that the dispersion curves depend on the properties of the shell/rib system and the spacing between
the ribs. Damping can also be introduced to control sound radiation which is produced
by the action of the flexural waves.
Cuschieri [22] developed a hybrid method to obtain the spatial domain solution
of the response of a ribbed cylindrical shell. The results show this method is very
efficient computationally, especially in the case of evaluation of the response of Greens
function. Choi et al. [16] used a normal mode approach to analyze the vibration of a
submerged periodic cylindrical shell with ribs. It was found that for the equally spacing
ribs, the frequencies near the natural frequencies of the ribs can be interpreted as a pass
band, and the remaining frequency range are essentially stop bands. If the irregularities
exist in the ribs, some stop bands may overlap with pass bands. Similar results were
10
11
motion and the upper band is associated with in-plane flexural rib motion. The small
frequency gap arises from the rib transmission resonance, essentially phase matching
to the flexural wave in the rib. The upper passband involves primarily reverberant flexural motion of the shell and, apart from the very significant band gaps, the Bloch wave
number in this and can be predicted with reasonable accuracy by the fluid loaded flexural wave number of the unribbed shell scattering into the Brillioun zone by the near
periodic array.
Photiadis .el. [59] also studied the resonant response of complex shell structures.
It was found that as the system becomes less uniform, the resonances of the system
becomes more and more local, associated with particular locations along the structure,
rather than with particular wave numbers. Generally, the effects of the increasing complexity are to increase the spatial localization of vibrational energy and to increase the
number of resonant modes which are well coupled to acoustic waves in the surrounding
fluid.
1.2
Hydrodynamics
Structural acoustics focuses on assumed sources of noise, such as, single force,
monopole excitation, and dipole excitation. When sound radiation from a propeller is
investigated, the source of noise is obtained from the interaction of the flow-propellerduct system. The propeller, which works in a nonuniform flow, is modeled as a rotor/stator stage. The shed wakes from rotor blades convect and interact with stator
blades. The common approach of computing such interaction processes is to introduce
an incident gust generated by the rotor blade, then subject the stator blades to the gust.
Although turbomachinery blades have complicated geometries, some simplified
modes were developed to investigate the rotor/stator interaction. Atassi and Hamad [36]
12
13
vorticity is very strong. The vorticity content in the acoustic modes increases as the
Mach number increases.
In the numerical simulation for the interaction problem of the rotor/stator system,
calculations are performed on truncated domains, so non-reflecting boundary conditions are necessary to avoid non-physical reflections at the far field boundaries. Accurate computations of rotor/stator interaction noise depend on the accuracy of the nonreflecting boundary conditions.
For uniform flow, Fang and Atassi [28] used a normal mode analysis to derive a
non-reflecting boundary condition. Golubev and Atassi [34, 35], and Montgomery and
Verdon [53] developed non-reflecting boundary conditions for swirling flow in an annular duct. The boundary conditions developed by Ali et al. [3] require the knowledge of
the eigenmodes, since it uses a numerical filter. The boundary conditions developed by
Montgomery and Verdon need both left and right eigenfunctions from the normal mode
analysis for general swirling flows. Elhadidi [26] developed a new boundary condition
for high frequency calculations. He used Gram-Schmidt procedure and the inner product to calculate the acoustic pressure coefficients, thus avoided matrix inversion which
is usually ill-conditioned.
Atassi et al. [3] has derived non-reflecting boundary conditions for acoustic and
vortical waves propagating in a duct. Ali [1] implemented this non-reflecting boundary condition in a three-dimensional linearized Euler code to study the propagation of
waves in a rigid duct with swirling flows. He examined the effects of swirling flow on
the generation, propagation, and stability of unsteady disturbances. He also investigated
the problem of a rotor-wake interacting with unloaded stator vanes in swirling flows.
The numerical results of scattering vortical waves in an annular duct for a uniform
flow were validated by comparison with the lifting surface theories of Namba [54] and
14
Schulten [55]. The three-dimensional calculations were compared to the strip theory
approximation. The comparison showed that strip theory is not a good approximation
for predicting the acoustics of the three-dimensional scattering problem.
Elhadidi [26] investigated the interaction of high frequency, unsteady, three dimensional incident disturbances with an annular cascade of loaded blades in a rigid duct
with swirling flows. An efficient numerical model was developed and the numerical scheme was made efficient by splitting the velocity field into nearly-acoustic and
nearly-convected vortical components. This leads to a coupled set of equations, which
can be solved iteratively. Numerical results showed that steady blade loading increases
the acoustic pressure compared to the unloaded blades in swirling flows. The results
also indicated that spanwise blade loading and blade twist excite higher order acoustic
modes and may contribute significantly to the sound level. He also investigated passive noise reduction techniques by increasing rotor/stator gap, applying blade lean and
sweep and mean flow acceleration. Numerical Results indicated that blade lean and
sweep are effective means for noise reduction. It was found that significant reduction
in unsteady lift and sound pressure is obtained by increasing the gap.
In the present work, the interaction of rotor-wakes with unloaded stator vanes in
an elastic duct will be studied. The new boundary conditions at the duct wall will be
derived using the approximation of a thin shell and implemented in a three-dimensional
linearized Euler code. The coupling between fluid motion and duct vibration will be
investigated.
15
Fluid forces:
Blade dipoles,
Ingested turbulence,
Swirling centrifugal
Coupling
Acoustic radiation
Figure 1.2. Model development scheme for elastic duct and fluid motion.
generated flow is implemented using Eulers equations and the shell equations, which
are coupled by the boundary condition between the fluid and the duct.
Figure 1.2 shows the model development scheme for the elastic duct vibration and
fluid motion. Several types of fluid forces act on the duct: the blade generated dipole
sources, ingested turbulence and swirling centrifugal force. The elastic duct deforms
in response to the forces. The fluid motion is modeled by the Euler equations and the
duct vibration is modeled by the shell modal equations. And the radiated sound is
investigated by coupling the Euler equations and shell equations at duct boundaries.
The dissertation is organized as follows. In chapter 2, the governing equations
of the shell and rib motions and linearized Euler equations of fluid motion will be
presented. The shell equation is described by Goldenveizer & Novozhilov [63]. The
Euler equations are linearized about the steady mean flow. Then the equations are
reorganized by splitting the disturbance velocity to vortical and potential parts. At
16
last, the boundary condition at the duct radius will be derived, which couples the shell
equations and the Euler equations. In chapter 3, the sound radiation from thin plates will
be investigated. Both the regular thin plate and the thin plate with ribs are considered.
The sound radiations from regular thin shells and rib stiffened shells will be studied in
chapter 4 and chapter 5, respectively. Different excitations sources are considered, such
as single radial point force, monopole excitation, and dipole excitations. In chapter
6, the scattering phenomena in a rigid duct will first be examined, then the scattering
phenomena in an elastic duct will be studied. The sound radiation from a coupled
propeller-elastic-duct system will then be investigated. Finally, conclusions will be
given in chapter 7.
17
CHAPTER 2
MATHEMATICAL FORMULATIONS
The aim of this research is to investigate the interaction of nonuniform flows with
propeller blades in a submerged elastic duct. The fluid motion is governed by Euler
equations, and the duct vibration is modeled by shell equations. The fluid motion and
the duct vibration are coupled at the duct boundaries.
18
Fluid
Force
r
Fluid
a
x
h
(2) Strains and displacements are sufficiently small so that the quantities of second
order and higher order magnitude in the strain-displacement relations may be neglected
in comparison with the first order terms.
(3) The transverse normal stress is small compared with the other normal stress
components and it can be neglected.
(4) Normals to the undeformed middle surface remain straight and normal to deformed middle surface suffer no extension. This means that the strains are zero in
normal direction.
19
transverse shear effects are omitted, the equations for the shell motion are presented
here as,
L
21 L22 L23
Zx Ex
Z = E ,
Zr
Er
where,
(
L11 = E1
2 1 2
+
x2
2a2 2
L12 = E1
2
+ s h 2 ,
t
1+ 2
,
2a x
L13 = E1
,
a x
L21 = L12 ,
[
L22 = E1
]
1 2
2 2 2
2
1 2
2
+
2
)
+
+
(1
+
h
,
s
2 x 2 a2 2
x2 a2 2
t2
[
L23 = E1
]
3
2 3
1
2
(2 )
,
a2
x 2 a2 3
L31 = L13 ,
20
(2.1)
L32 = L23 ,
[
L33 = E1
]
4
1
2 4
4
2
2 2
2
+
+
+
2
h
.
a
s
a2
x 4 a2 4
x2 2
t2
In these equations, Zx , Z and Zr are the axial, tangential and radial displacements at
the cylinders mid-surface, being positive when acting in the positive directions of the
coordinates axes. Ex , E and Er are the axial, tangential and radial excitations, which
are the mechanical traction(forces per unit area). E1 is defined as,
E1 =
Eh
,
1 2
where, E is the Youngs Modulus, is the Poissons ratio. And s is shells density, h
is the thickness of the cylindrical shell, a is shells mean radius, and 2 =
h2
.
12a2
im
f (r, m, )ei x d ,
(2.2)
(2.3)
where m is the circumferential mode number and is the axial wave number. F(r, , x)
is the field quantity and f (r, m, ) is the spectral field quantity. Thus the differential
equations of the shells motions, Eq. (2.1), are reduced to the spectral equations of shell
21
motions,
S
21 S22 S23
x (m, )
(m, )
r (m, )
ex (m, )
= e (m, )
er (m, )
(2.4)
where, the amplitudes x (m, ), (m, ), r (m, ) are the spectral displacements; the
amplitudes ex (m, ), e (m, ), er (m, ) are the spectral excitations. The elements in
the stiffness matrix S are given by,
(
S11 = E1
1
+ m2
2a2
S12 = E1 (1 + )
)
2 s h,
m
,
2a
S13 = iE1 ,
a
S21 = S12 ,
[
S22 = E1
]
(1 ) 2 m2
2 m2
2 2
+ 2 + 2 (1 ) + 2 2 s h,
2
a
a
[
S23 = iE1
]
m
2 m3
2
2
+ (2 ) m + 2 ,
a2
a
22
S31 = S13 ,
S32 = S23 ,
[
S33 = E1
]
1 2 m4
4 2 2
2 2 2
+ 2 + a + 2 m 2 s h.
a2
a
Ex ( , x)
E ( , x)
Er ( , x)
Fx (x x0 ) ( 0 )/a
F (x x0 ) ( 0 )/a
Fr (x x0 ) ( 0 )/a pe (a, , x) + pi (a, , x)
(2.5)
where pe (a, , x) is the exterior fluid pressure and pi (a, , x) is the interior fluid pressure. Its spectral form is,
ex (m, )
e (m, )
er (m, )
fx
ei(m0 + x0 ) /2 a
f ei(m0 + x0 ) /2 a
fr ei(m0 + x0 ) /2 a pe (a, m, ) + pi (a, m, )
23
(2.6)
)
1 D20
2
p = 0,
c20 Dt 2
(2.7)
D0
Dt
+ U(x) .
Dt
t
(2.8)
Generally, the mean swirling flow U(x) is vortical and can be assumed to be axisymmetric, of the form,
U(x) = Ux (r)ex +U e ,
(2.9)
where Ux and U are the mean velocity components in the axial and circumferential
directions, respectively; ex and e represent unit vectors in the axial and circumferential
directions, respectively.
Using Fourier transform Eqs.(2.2) and (2.3),
D0
p = i( +Ux +U m/r) p,
Dt
(2.10)
mU
,
r
(2.11)
)
2m
2
+ p = 0.
c20
24
(2.12)
2m
=
c20
mM
+ Mx +
c0
r
)2
.
(2.13)
mM
r
can be neglected
mM
r
pressure must be a solution of the reduced wave equation which is finite at the origin
[63]. Thus,
pi (r, m, ) = Am ( )J|m| (1 r),
pi (r, m, )
= 1 2 r (m, ),
r
at
r = a,
where, 1 is interior fluid density, thus, the interior spectral pressure is obtained as,
pi (r, m, ) = 1 2 r (m, )
J|m| (1 r)
( a) .
1 J|m|
1
(2.14)
Due to the cylinder motion, the exterior spectral pressure must be a solution of the
reduced wave equation which satisfies the radiation condition at infinity. [63]. Thus,
pe (r, m, ) = Bm ( )H|m| (2 r),
25
pe (r, m, )
= 2 2 r (m, ),
r
r = a,
at
where, 2 is exterior fluid density, thus, the exterior spectral pressure is obtained as,
pe (r, m, ) = 2 2 r (m, )
H|m| (2 r)
( a) .
2 H|m|
2
(2.15)
Plug Eq. (2.6), Eq. (2.14) and Eq. (2.15) into Eq. (2.4), the spectral equations of
motion of the cylindrical shell are obtained as,
S13
S11 S12
S
S23
21 S22
x (m, )
(m, )
r (m, )
fx
= f ei(m0 + x0 ) /2 a
fr ei(m0 + x0 ) /2 a
ei(m0 + x0 ) /2 a
(2.16)
H|m| (1 a)
2 H|m| (2 a)
1 2
J|m| (1 a)
1 J|m| (1 a)
(2.17)
From the spectral equations of motion of the cylindrical shell, it can be seen that
there are three force terms: (1) elastic term; (2) duct inertia term; (3) fluid loading term.
The scales of these force terms are shown in Table 2.1. To compare these force terms,
a steel cylindrical shell is considered. The shell is submerged in water with a radius
a = 1m, and a thickness h = 0.01m. In table 2.1, frequency is normalized, = a/c0 ,
where c0 is the sound speed in water. s is the density of the shell, and w of water. The
duct inertia term and fluid loading term are normalized by the elastic term. From table
2.1, it can be seen that the elastic term is the dominant term when the reduced frequency
is very small ( 1). When the reduced frequency is larger than 1, the duct inertia
term and the fluid loading term become the dominant ones. This table shows that the
fluid loading term is very important for lager frequencies ( > 1).
26
TABLE 2.1
SCALING OF FORCES.
Reduced frequency
Elastic
Duct inertia
Fluid loading
= a/c0
Eh/a2
s 2 h
w c0
(s c20 /E) 2
(w c20 /E)(a/h)
1 (dominant term)
0.1 2
0.1
2(dominant term)
0.1 2
2 (dominant term)
Normalization
D0 U
= p,
Dt
(2.18)
(2.19)
where , U, and p are the density, velocity, and pressure of the fluid, respectively.
The governing equations are linearized about the steady mean flow quantities,
(2.20)
(2.21)
(2.22)
27
where x stands for the position vector, t for time, and 0 , U0 , and p0 are the steady
density, velocity, and pressure of the fluid, respectively, and , u, and p are the corresponding unsteady perturbation quantities such that
|u(x,t)| |U0 (x)|,
(2.23)
(2.24)
| (x,t)| |0 (x)|.
(2.25)
Thus, the first-order continuity and momentum equations resulting from the linearization are given,
D0
+ (u )0 + 0 u + U0 = 0,
Dt
(
)
D0 u
+ (u )U0 + (U0 U0 ) = p .
Dt
(2.26)
(2.27)
We used a velocity splitting to describe the basic modes of the disturbance [5, 32].
The disturbance velocity is decomposed into vortical and potential parts,
u = u(R) + .
(2.28)
1 D0
c20 Dt
)
( ) = (u(R) ),
D0 u(R)
+ (u(R) )U0 = ( U0 ) .
Dt
28
(2.29)
(2.30)
(2.31)
In the normal mode analysis, the flow disturbances are represented as the sum of linearly independent modes. This can be represented as a Fourier expansion,
m= n=1
(2.32)
where m and n are integer modal numbers characterizing the circumferential and radial
eigenmodes. is the frequency, and k is the wavenumber. Thus, Eq. (2.31) reduces to,
umn (r)
= 0,
Lr
p
(r)
mn
mn (r)
(2.33)
ur
r
)
= 0.
rh
29
(2.34)
D0 r
.
Dt
(2.35)
By applying Eq. (2.32) to Eq. (2.35), the boundary condition at the duct tip is obtained as,
urmn = imn r mn ,
where, mn is an eigenvalue of the convected operator
mn = + Ux +
(2.36)
D0
Dt ,
defined by,
mU
.
r
(2.37)
From the shell equation, the displacements of the shell can be calculated by,
mn = RFmn ,
where matrix R = {Ri j } is the inverse matrix of stiffness matrix S, and Fmn is the excitation which is given by,
Fmn = {0, 0, pmn }.
30
r mn = R33 pmn ,
31
(2.38)
CHAPTER 3
SOUND RADIATION FROM THIN PLATES
As stated in the previous chapters, the main objective of this research is to study the
interaction between the vibrating structures and the fluids in which the structures are
submerged. The knowledge of both the structural acoustics and the fluid mechanics is
required for this purpose. Before the vibrations of thin cylindrical shells are studied,
the vibrations of thin plates are first examined, which is a two-dimensional problem,
rather than the complex three-dimensional one as the former cases.
The sound radiation from an infinite thin plate has been studied by many researches.
The geometry of an infinite thin plate of uniform thickness h can be shown in Figure
3.1. In this chapter, the governing equation of the transverse displacement of the thin
plate, without moment loading, is given by [63],
(
)
4W (x, y)
4W (x, y) 4W (x, y)
2W (x, y)
h
D
+
2
+
+
=
s
x4
x 2 y2
y4
t2
Sz pu (x, y, 0) + pl (x, y, 0),
(3.1)
F(x, y, z) =
1
4 2
f ( , , z) =
+ +
+ +
f ( , , z)ei( x+ y) d d ,
(3.2)
(3.3)
where F(x, y, z) is the field quantity and f ( , , z) is the spectral field quantity. and
33
(3.4)
3.1
(3.5)
Dispersion Equation
The dispersion relation is the relation between the frequency and wavenumber, at
which the resonance occurs. For an infinite plate in vacuum, the dispersion equation is,
D( 2 + 2 )2 2 s h = 0.
(3.6)
By solving this equation, we can always find real values of wavenumbers( and ) for
a given frequency( ), which means that there are always propagating modes.
When the plate is submerged in fluid, the dispersion equation becomes,
D( 2 + 2 )2 2 s h
iu 2 il 2
= 0,
u
l
(3.7)
34
(3.8)
Sz ( , ) = F0 ei( x0 + y0 ) ,
(3.9)
where F0 is the amplitude of the point force. And the spectral equation of motion of the
plate is obtained as,
D f ( , )W ( , ) = Sz ( , ),
(3.10)
where
D f ( , ) = D( 2 + 2 )2 2 s h
iu 2 il 2
u
l
(3.11)
is the spectral stiffness of the plate with the fluid loading. And the solution of this
spectral equation is,
W ( , ) =
Sz ( , )
.
D f ( , )
(3.12)
Due to the plate motion, the acoustic pressure in the upper halfspace must be an
outgoing wave solution of the reduced wave equation. Thus, the acoustic pressure is
given by,
1
pte (x, y, z) = 2
4
+ +
P( , )ei( x+ y+u z) d d .
(3.13)
By applying the boundary condition, reflecting equality of acoustic and plate transverse
displacements at z = 0,
pte
= u 2W (x, y),
z
35
(3.14)
the unknown function P( , ) can be eliminated from Eq. (3.13), then we obtain,
pte (x, y, z) =
u 2
4 2
+ +
W ( , )
iu
ei( x+ y+u z) d d .
(3.15)
In the upper halfspace, the far field pressure is obtained by the method of stationary
phase [63] as,
pte f (R, , ) = u 2W (0 , 0 )
eiku R
,
2 R
(3.16)
(3.17)
iu 2
= 0.
u
(3.18)
Crighton [20] has studied this relation and showed that at all frequencies there is
only one subsonic free mode (non-propagating) for which > ku . And there are no
purely real roots for which < ku . But there is a root which has a small imaginary
part and the real part is less than ku when the frequency is larger than the coincidence
frequency. This corresponds to the leaky waves. The coincidence frequency of the
plate in vacuo is obtained by setting the acoustic wavenumber ku = c /cu equals the
36
c = c2u (s h/D)1/2 .
In the following numerical examples, sound radiation from a steel plate of thickness 1 cm with damping factor 0.01 is considered. The upper halfspace contains water
and the lower halfspace is a vacuum. The excitation is a unit transverse point force located at the origin. The problem has cylindrical symmetry because the field quantities
are independent of the azimuthal angle . We take the angle = 0 in the numerical
examples, thus 0 = ku sin , 0 = 0, and u = ku cos .
Figure 3.2 shows the magnitudes of the displacements of an infinite steel plate. The
color scale represents the log10 displacement value in meters. The yellow curve is the
dispersion curve which corresponds to the larger values of the radial displacements. It
can be seen from the figure that leaky wave exists when the frequency is larger than
coincidence frequency. The leaky wave has strong effect on far field sound only for
a very small frequency range, which occurs at certain angle c ,
c = sin1 ( c / ).
37
Leaky wave
Flexture wave
Figure 3.2. Flexure waves of an infinite steel plate excited by a unit force.
The color scale represents the log10 displacement value in meters.
pi (x, y, z) = pi (R, R0 ) =
m eiku |RR0 |
m eiku r
=
,
t 4 |R R0 |
t 4 r
(3.19)
where m is a mass flow rate per unit volume, R = (x, y, z) is the location of a field point,
R0 = (x0 , y0 , z0 ) is the location of the source, and
r = |R R0 | =
(x x0 )2 + (y y0 )2 + (z z0 )2 .
38
150
140
130
dB
120
110
400Hz
2000Hz
20000Hz
30000Hz
40000Hz
100
90
80
70
10
20
30
40
50
60
70
80
90
Figure 3.3. Far field sound radiation from 1 cm steel plate with water loading
on one side only. The excitation is a unit point force at the origin.
(3.20)
m
= i0 Q0 ei t .
t
(3.21)
Thus, we have,
pi (x, y, z) = i0 Q0 ei t
eiku |RR0 |
eiku |RR0 | i t
=A
e
,
4 |R R0 |
|R R0 |
39
(3.22)
pi (x, y, z) =
f or
z > z0 ,
f or
z < z0
(3.23)
When a monopole is located at (x0 , y0 , z0 ) in the upper halfspace, where z0 is distance above the plate which is located at z = 0, a image monopole need to be placed
at (x0 , y0 , z0 ) to obtain the pressure at the plate. The transform of the pressure at the
plate is given by,
+
1
i4 A i( x0 + y0 u z0 ) i( x+ y)
e
= 2
d d ,
e
4
u
(3.24)
4 iA i( x0 + y0 u z0 )
e
.
u
(3.25)
Sz ( , )
.
D f ( , )
(3.26)
The far field pressures in the upper halfspace, at observation angles ( , ), are,
pi f (R, , ) = A
eiku R
eiku (x0 sin cos +y0 sin sin +z0 cos ) ,
R
(3.27)
pr f (R, , ) = A
eiku R
eiku (x0 sin cos +y0 sin sin z0 cos ) ,
R
(3.28)
40
pte f (R, , ) = u 2W (0 , 0 )
eiku R
,
2 R
(3.29)
where pi f is the incoming wave pressure and pr f is the reflected wave pressure and pte f
is the far field pressure due to plate displacements which is the same as before.
In the following numerical example, sound radiation from a steel plate of thickness
1 cm and with damping factor 0.01 is considered. The upper halfspace contains water
and the lower halfspace is a vacuum. The excitation is a monopole located at x0 = 0,
y0 = 0 and z0 = 0.05m. Take the monopole strength A = 1. Figure 3.4 shows the
angular distribution of far field sound pressure level due to steel plate displacements.
The coincidence lobe occurs at at c = 50o for frequency 40000 Hz; and the coincidence
lobe occurs at at c = 62o for frequency 30000 Hz. They are the same as the results of
the single force excitation. Again, when the frequency is smaller than the coincidence
frequency, there is no coincidence lobe in Figure 3.4.
(3.30)
where d is distance between two monopoles. In the far field, the equation of dipole is,
pd (R, R0 ) = iku d cos pi (R, R0 ) = iku d cos A
where is the angle between dipole direction and r direction.
41
eiku r
,
r
(3.31)
130
125
dB
120
115
110
400Hz
2000Hz
20000Hz
30000Hz
40000Hz
105
100
10
20
30
40
50
60
70
80
90
Figure 3.4. Far field sound radiation from 1 cm steel plate with water loading
on one side only. The excitation is a monopole at x0 = 0, y0 = 0 and
z0 = 0.05m.
42
4 ku cos dA i( x0 + y0 u z0 )
.
e
u
(3.32)
As we mentioned in the section 3.2, the field quantities are independent of the azimuthal angle , and we can take the angle = 0 in the numerical examples, thus
(3.33)
We notice that the expression of the dipole spectral excitation is very similar to that of
the single force. And the numerical result of dipole excitation will be similar to that of
single force too.
In the following numerical example, sound radiation from a steel plate of thickness
1 cm and with damping factor 0.01 is considered. The upper halfspace contains water
and the lower halfspace is a vacuum. The excitation is a dipole located at x0 = 0, y0 = 0
and z0 = 0.05m, and the dipole strength is taken to be 4 A d = 1, which is the same
as the unit point fore used in section 3.2. Figure 3.5 shows the angular distribution of
far field sound pressure level. As we expected, we obtain the exact same result as single
force excitation shown in Figure 3.3. The interspersion of the results are the same as
that of the single force excitation.
43
150
140
130
dB
120
110
400Hz
2000Hz
20000Hz
30000Hz
40000Hz
100
90
80
70
10
20
30
40
50
60
70
80
90
Figure 3.5. Far field sound radiation from 1 cm steel plate with water loading
on one side only. The excitation is a dipole at x0 = 0, y0 = 0 and z0 = 0.05m.
W ( , ) =
1
D f ( , ) D ( , ) 1 + G ( ) n=+
f
f
n= D ( 2 n/d, )
(3.34)
(3.35)
where for ribs, E f is Youngs modulus, I f is the cross-section area moment of inertia,
44
The far field pressure due to plate displacements is the same as before,
pte f (R, , ) = u 2W (0 , 0 )
eiku R
,
2 R
(3.36)
where, it is recalled, 0 = ku sin cos and 0 = ku sin sin are the stationary phase
wavenumbers.
In a numerical example, a 1 cm stiffened steel plate loaded with water on one side
only is considered. The rectangular cross-section of the steel ribs have width a = 1cm
and depth b = 2cm, and the rib spacing is d = 20cm. The area moment of inertia of the
rib cross-section is taken about the rib contact line with the plate, i.e. I f = ab3 /12 +
ab3 /4. Three types of excitations are chosen: unit force, monopole and dipole. And the
excitation is either on a rib or between ribs. The damping factors of the plate and ribs
are chosen as 0.02.
Figure 3.6. Flexure waves of an infinite steel plate with ribs excited by a unit
force. The color scale represents the log10 displacement value in meters.
The peaks in the figures are associated with the flexure waves in the dispersion curves
in the Figure 3.6. We can see that the peaks in the spectra are more than 10 dB above
the levels of unstiffened plate. When the single force is applied on the ribs, the far
field sound level of stiffened plate is less than that of unstiffened plate at the frequency
range 2000Hz to 6000Hz, because the excitation is applied at a higher impedance. On
the other hand, when the force is applied between two ribs, the far field sound pressure
level is a little bit higher than that of the unstiffened plate, because the excitation is
applied at a lower impedance. This pattern can be observed more clearly when the
mass of the ribs is increased.
In the following examples, the depth of the rectangular cross-section of the steel
ribs is increase from b = 2cm to b = 10cm. Thus the mass of the ribs is five times that
of the previous example. The unit point force is apply at x0 = d/2 and x0 = 0, and the
46
140
135
130
125
dB
120
115
110
105
No ribs
With ribs
100
95
90
2000
4000
6000
Frequency (Hz)
8000
10000
Figure 3.7. Broadside far field sound level of steel plate stiffened by ribs. The
excitation is a unit point force between ribs.
results are shown in Figures 3.9 and 3.10, respectively. Compare these two figures with
figures 3.7 and 3.8, it is very clear that the ribs have more effect on the far field sound
level when the ribs are more heavy.
3.5.3 Monopole
In this part, the excitation is a monopole located between ribs and on ribs, respectively. The results are shown in figures 3.11 and 3.12. We can see that the peaks in the
spectra are about 5-8 dB above the levels of unstiffened plate.
3.5.4 Dipole
In this part, the excitation is a dipole located between ribs and on ribs, respectively.
The results are shown in figures 3.13 and 3.14. It is noticed that the peaks in the spectra
47
140
135
130
125
dB
120
115
110
105
No ribs
With ribs
100
95
90
2000
4000
6000
Frequency (Hz)
8000
10000
Figure 3.8. Broadside far field sound level of steel plate stiffened by ribs. The
excitation is a unit point force on a rib.
140
135
130
125
dB
120
115
110
105
100
No ribs
With ribs
95
90
2000
4000
6000
Frequency (Hz)
8000
10000
Figure 3.9. Broadside far field sound level of steel plate stiffened by ribs. The
excitation is a unit point force between ribs. The mass of the ribs is five times
that of Figure 3.7.
48
140
135
130
125
dB
120
115
110
105
100
No ribs
With ribs
95
90
2000
4000
6000
Frequency (Hz)
8000
10000
Figure 3.10. Broadside far field sound level of steel plate stiffened by ribs.
The excitation is a unit point force on a rib. The mass of the ribs is five times
that of Figure 3.8.
130
No ribs
With ribs
125
120
dB
115
110
105
100
95
2000
4000
6000
Frequency (Hz)
8000
10000
Figure 3.11. Broadside far field sound level of steel plate stiffened by ribs.
The excitation is a monopole between ribs.
49
130
No ribs
With ribs
125
dB
120
115
110
105
100
2000
4000
6000
Frequency (Hz)
8000
10000
Figure 3.12. Broadside far field sound level of steel plate stiffened by ribs.
The excitation is a monopole on a rib.
are about 5 dB above the levels of unstiffened plate. It is also noticed that the ribs do
not have much effect on the far field sound level except near the flexural modes.
3.6 Vortex
First, no mean flow case is considered. Assume a vortex locates at (x0 , z0 ) and the
image vortex locates at (x0 , z0 ), then the velocity field is given by,
u iv =
2 i
)
1
1
,
(x + iz) (x0 + iz0 ) (x + iz) (x0 iz0 )
50
(3.37)
140
135
No ribs
With ribs
130
125
dB
120
115
110
105
100
95
90
2000
4000
6000
Frequency (Hz)
8000
10000
Figure 3.13. Broadside far field sound level of steel plate stiffened by ribs.
The excitation is a dipole between ribs.
140
No ribs
With ribs
135
130
125
dB
120
115
110
105
100
95
90
2000
4000
6000
Frequency (Hz)
8000
10000
Figure 3.14. Broadside far field sound level of steel plate stiffened by ribs.
The excitation is a dipole on a rib.
51
where is the circulation, u and v are the velocities in x and y direction, respectively.
At plate surface, z = 0, so the above equation becomes,
u iv =
2 i
1
x x0 iz0
1
x x0 + iz0
(3.38)
u=
z0
.
(x x0 )2 + z20
(3.39)
v = 0.
(3.40)
z0
(x x0 )2 + z20
)2
.
(3.41)
,
2 i (x0 + iz0 ) (x0 iz0 )
4 z0 .
And x0 =
(3.42)
t
4 z0 .
4 z0 .
z0
1 2
p+ V +
=
p
+
V .
2
(x x0 )2 + z20
2
p = p
V z0
.
(x x0 )2 + z20
p = p p =
52
V z0
.
(x x0 )2 + z20
(3.43)
(3.44)
(3.45)
i t , and we know x0 = V t,
Assume = e
p =
Kei t
,
(x V t)2 + z20
(3.46)
0V / .
where, K = z
Fourier expansion:
f (x) =
1
,
(x V t)2 + z20
(3.47)
z0 | |
e
.
z0
(3.48)
F( ) =
1
p(x) =
2
1
p(x) =
2
+
K i t z0 | | i (xV t)
e
e
e
,
(3.49)
+
K z0 | | i x i( +V )t
e
e e
,
(3.50)
z0
z0
K z0 | |
e
,
z0
(3.51)
1 = + V.
(3.52)
The spectral stiffness of the plate with fluid loading is given by,
D f ( , ) = D( 2 + 2 )2 12 s h
iu 12 il 12
,
u
l
(3.53)
where u and l are given in section 1. The displacement of the plate with ribs is given
by equation(17), and again, the far field pressure due to plate displacements is the same
53
175
170
165
160
dB
155
150
145
140
135
130
No ribs
With ribs
125
120
2000
4000
6000
Frequency (Hz)
8000
10000
Figure 3.15. Broadside far field sound level of steel plate stiffened by ribs.
The excitation is a vortex above the plate (z0 = 0.05m) .
as before,
eiku R
2
pte f (R, , ) = u 1 W (0 , 0 )
,
2 R
(3.54)
where, it is recalled, 0 = ku sin cos and 0 = ku sin sin are the stationary phase
wavenumbers.
In the following numerical example, a 0.01m stiffened steel plate loaded with water
on one side only is considered. The rectangular cross-section of the steel ribs have width
a = 0.01m and depth b = 0.02m, and the rib spacing is d = 0.2m. The damping factors
of the plate and ribs are chosen as 0.02. Other parameters are chosen as followed:
K = 1, V = 10m/s, z0 = 0.05m, and = 0 and /4. The results are shown in figures
3.15 and 3.16. We can see that the ribs have strong effects on the far field sound pressure
levels. The peaks in the figures can be associated with the dispersion curves in Figure
3.6.
54
175
170
165
160
dB
155
150
145
No ribs
With ribs
140
135
130
125
120
2000
4000
6000
Frequency (Hz)
8000
10000
Figure 3.16. Far field ( = /4) sound level of steel plate stiffened by ribs.
The excitation is a vortex above the plate (z0 = 0.05m).
3.7 Conclusions
In this chapter, the sound radiation from an infinite plate with or without ribs is studied for different excitations sources: radial single force, monopole, dipole and vortex.
For the unstiffened plate, there is a single propagating flexure wave when the frequency is lower than the coincidence frequency. This flexure wave can not propagate to
the far field. When the frequency is higher than the coincidence frequency, there is no
purely propagating flexure wave. But the leaky wave is found for the frequency higher
than the coincidence frequency. The leaky wave is very important over a narrow range
of angle c , and it is related to the lobes in the spectra of the far field sound pressure
level of different excitation sources.
When the plate is stiffened with ribs, the dispersion curves are strongly changed.
And there are many propagating flexure waves found for a single frequency. Some of
55
these flexure waves can propagate to the far field. Both the pass and stop band structures
are found in the dispersion curves. They are corresponding to the peaks in the spectra
of different excitation sources.
56
CHAPTER 4
SOUND RADIATION FROM THIN SHELLS
In last chapter, a two-dimensional case of the sound radiation from the plates has
been studied. In this chapter, we will carry on studying the three-dimensional problem
of the sound radiation from thin infinite cylindrical shells.
Mechanical and acoustic excitation of a cylindrical shell will result in several types
of structural wave motions, for example, some waves propagate away from the source
into the fluid as acoustic waves and some waves travel along the shell associated with
disturbances in the fluid [62]. The source excitation can be decomposed into wave
components with different frequencies, and the waves have the form ei( x+m t) in
complex notation. In the dispersion equation, the wavenumber can be found by
fixing frequency , and the value of can be either real or complex. The complex
values represent the disturbances which decay from the source. The propagating modes
are of interest, so only the real values of are considered.
In this chapter, the dispersion curves for in-vacuo infinite shell and water-filled infinite shell will be studied, since the dispersion curves can be used to explain the strong
acoustic radiation from the shell. The far field acoustic radiations for different shell
materials and for different excitation sources will also be investigated.
57
S13
S11 S12
S
S23
21 S22
x (m, )
(m, )
r (m, )
0
= 0 ,
0
(4.1)
where the matrix [S] is the 3 3 spectral dynamic stiffness matrix with fluid loading
term FL in element S33 . For a nontrivial solution, the determinant of the stiffness
matrix is set equal to zero, |S( , )| = 0, which yields the dispersion equation. The
dispersion relation is the relation between frequency and wavenumber at which the
resonance occurs. The dispersion equation yields either of the following two eigenvalue
problems:
1. For constant mode number m and frequency , there exists one or more values
of called free propagating wavenumbers, at which |S( , )| vanishes.
2. For given mode number m and wavenumber , there exists one or more values
of frequency such that the determinant vanishes.
One of the methods to find the propagating modes is to find the real roots of the
dispersion equation. The mode shapes (or eigenfunctions) of free vibration are found by
returning to the homogeneous set of equations which yielded the dispersion equation.
Any two of the equations are chosen and the third is discarded. The two remaining
equations can be solved for the ratios of displacements, the convenient ratios to choose
being x /r and /r . For example, using the first two of equations Eq. (4.1), which
58
=
.
/r
S21 S22
S23 ,
(4.2)
59
50
45
40
Wavenumber
35
30
25
20
15
10
5
0
5000
10000
15000
Frequency
4.1.2 In-vacuo Wave Numbers versus Frequency for Different Circumferential Mode
Numbers
In this section, a simple case is considered: a vacuum cylindrical shell surrounded
by a vacuum, which means fluid loading term FL = 0. The non-dimensional variables
of wavenumber and frequency are defined as,
= a,
=
a
,
c0
where a is the radius of the shell and c0 = 1482m/s is the speed of sound in water. The
cylindrical shell has a radius of 1.0 m, thickness of 0.01 m, and some properties of steel
are given in Table 4.1.
60
TABLE 4.1
PROPERTIES OF DIFFERENT MATERIALS.
E(Pa )
(kg/m3 )
c p (m/s)
Steel
0.195 1012
0.29
7700
5358
Aluminum
0.700 1011
0.33
2700
5394
Hard rubber
0.200 1010
0.40
1100
1471
Material
The method used to find the real roots of dispersion equation, is to search for sign
changes of the determinant in a given wavenubmer range for a fixed frequency, rejecting
those false roots which arise from an infinity rather than a zero.
First, the dispersion curves and ratio of displacement for circumferential mode number m = 1 are plotted. Figure 4.2 shows the dispersion curves for the steel shell in
vacuum. In this figure, three branches are observed (named by branch 0, branch 1, and
branch 2, respectively). For each branch, the ratio of displacements versus frequency
is plotted, and the results are shown in Figures 4.3, 4.4, and 4.5, respectively. From
Figure 4.3, it can be found that when is small, |x /r | 1 and | /r | < 1, which
means x is much larger than r and , and branch 0 cuts on as a compression wave at
the beginning. But as frequency increases, becomes the dominant term, thus branch
0 becomes a wave of torsion. So it can be concluded that branch 0 cuts on as a compression wave then changes to a wave of torsion. In Figure 4.4, it can be found that
|x /r | < 1 and | /r | < 1, which means r is dominant term. So it can be concluded
that branch 1 mainly is a wave of flexure, and this branch is the most important one
that we are interested in. From Figure 4.5, it can be found that at the beginning, is
larger than r and x , and branch 2 propagates as a torsion wave; then x becomes the
61
10
branch 0
branch 1
branch 2
9
8
7
6
5
4
3
2
1
0
0
10
dominant term for larger frequency, and branch 2 propagates as a compression wave.
So it can be concluded that branch 2 cuts on as a torsion wave then changes to a wave
of compression.
Then, the dispersion curves and ratio of displacement for circumferential mode
number m = 2 are plotted. Figure 4.6 shows the dispersion curves for circumferential mode number m = 2. Compare figure 4.6 with figure 4.2, the similar branches can
be found. For each branch, the ratio of displacements versus frequency are plotted, and
the results are shown in Figures 4.7, 4.8, and 4.9, respectively. The interpretations for
each branch are the same as previous case: branch 0 cuts on as a compression wave
then changes to a wave of torsion; branch 1 propagates as a wave of flexure; branch
2 cuts on as a torsion wave then will change to a wave of compression for even larger
frequencies. It is noticed that as circumferential number increases, branch 0 and branch
62
12
| / |
x
Ratio of displacement
| / r|
10
0
2
6
*
10
1.4
| / |
x
Ratio of displacement
| / r|
1.2
1
0.8
0.6
0.4
0.2
0
0
0.5
1.5
2.5
3.5
63
6
| / |
x
Ratio of displacement
| / |
0
5
10
2 move to right. It means that the branches cut-on for larger frequencies.
The dispersion curves and ratio of displacement for circumferential mode number
m = 3 are plotted in figure 4.10. Similar branches are observed, except branch 2 disappeared. And branch 2 will begin to propagate for larger frequency. The results of
displacement ratio for each branches are shown in figure 4.11, 4.12, respectively. The
interpretations for each branch are the same as previous case: branch 0 cuts on as a
compression wave then changes to a wave of torsion; branch 1 propagates as a wave of
flexure.
The dispersion curve for circumferential mode number m = 10 is plotted in figure
4.13. As we expected, only branch 1 is seen up to reduced frequency = 10, and
branch 0 and branch 2 will begin to propagate for higher frequency. The results of
displacement ratio for branch 1 are shown in figure 4.14. The interpretations for this
64
10
branch 0
branch 1
branch 2
9
8
7
6
5
4
3
2
1
0
0
10
20
| / |
x
18
16
Ratio of displacement
| / r|
14
12
10
8
6
4
2
0
4
7
*
10
65
0.7
| / |
x
Ratio of displacement
| / r|
0.6
0.5
0.4
0.3
0.2
0.1
0
0
0.5
1.5
2.5
3.5
Ratio of displacement
2.5
1.5
|x /r|
| / |
0.5
0
8
8.5
9
*
9.5
10
66
10
branch 0
branch 1
9
8
7
6
5
4
3
2
1
0
0
10
12
| / |
x
10
Ratio of displacement
| / r|
0
6
6.5
7.5
8
*
8.5
9.5
10
67
0.35
| / |
x
Ratio of displacement
| / r|
0.3
0.25
0.2
0.15
0.1
0.05
0
0
0.5
1.5
2.5
3.5
branch are the same as previous cases: branch 1 propagates as a wave of flexure.
68
10
branch 1
9
8
7
6
5
4
3
2
1
0
0
10
0.11
| / |
x
0.1
0.09
Ratio of displacement
| / r|
0.08
0.07
0.06
0.05
0.04
0.03
0.02
0.01
1
1.2
1.4
1.6
1.8
*
2.2
2.4
2.6
69
10
9
8
7
h/a=0.01
h/a=0.03
h/a=0.05
h/a=0.08
6
5
4
3
2
1
0
0
10
frequency, the wavenumber of the shell with thickness h = 0.01a is much larger than
that of the shell with thickness h = 0.08a.
Then, a larger circumferential number is chosen, m = 10, and the results are shown
in figure 4.16. It is clearly shown that for large circumferential numbers, the waves of
the flexure cut-on for larger frequencies as the thickness of the shell increases.
70
10
9
8
7
6
5
4
3
h/a=0.01
h/a=0.03
h/a=0.05
h/a=0.08
2
1
0
0
10
of hard rubber shell is much larger than that of the steel shell and the aluminum shell.
And the flexure curves of the steel shell and the aluminum shell are very close. For
larger circumferential mode number, m = 10, the similar results are observed, which
are shown in Figure 4.18.
4.1.5 Water-filled Wave Numbers versus Frequency for Different Circumferential Mode
Numbers
In this section, the properties of fluid loading will be examined firstly. The fluid
loading is the same as Eq.(2.11),
FL = 2 2
H|m| (2 a)
2 H|m| (2 a)
71
1 2
J|m| (1 a)
1 J|m| (1 a)
8
Steel
Aluminum
Hard Rubber
7
6
5
4
3
2
1
0
0
10
20
18
16
14
12
10
8
Steel
Aluminum
Hard Rubber
6
4
2
0
0
10
72
fluid loading
5
m=0
m=1
m=2
m=3
10
0.5
1.5
2
*
2.5
3.5
From the above equation, it can be seen that there are two fluid loading terms: exterior
and interior water loading.
In the following calculation, the radius of the thin shell is 1 m, and the axial wavenumber has been chosen as /2. The results are shown in Figures 4.19, 4.20 and 4.21.
The negative sign of real part of exterior water loading indicates that it acts as a mass.
The negative sign of imaginary part of exterior water loading means that it acts as a
resistance. The real part of interior water loading, which oscillates between , acts as
a stiffness or a mass depending on the frequency. The imaginary part of interior water
loading is zero so there is no resistance term.
Now, the dispersion curves of the water-filled cylindrical shell are investigated. A
water-filled steel cylinder surrounded by a vacuum is considered, which means fluid
loading FL = 0. Other parameters are as the same as the previous cases. In this part,
73
m=0
m=1
m=2
m=3
fluid loading
10
0.5
1.5
2
*
2.5
3.5
20
15
10
fluid loading
m=0
m=1
m=2
m=3
10
15
20
0.5
1.5
2
*
2.5
3.5
74
only interior water loading term is included, and the values of interior water loading
are real, so the real roots of the dispersion equation can be found by searching for sign
changes of the determinant in a given wavenumber range for a fixed frequency, rejecting
those false roots which arise from an infinity rather than a zero.
The dispersion curves for mode number m = 1 are obtained by solving the dispersion equation, and the result is plotted in Figure 4.22. Five branches are found for this
case. For each branch, the ratios of displacements are calculated, and the results are
plotted in Figures 4.23, 4.24, 4.25, 4.26, and 4.27, respectively. From Figure 4.23, it
can be seen that at the beginning, x is larger than r and , then becomes the larger
term, at last r becomes the dominant term. So it can be concluded that branch 0 cuts
on as a compression wave firstly, then changes to a torsion wave, and becomes a flexure
wave finally. r is the dominant term in Figure 4.23, so it can be concluded that branch
1 is a wave of flexure, which is the most important branch that of our interest. There is a
point of infinity in Figure 4.25, and branch 2 cuts on as a wave of flexure, then changes
to a wave of compression and torsion, at last, changes back to flexure wave. There also
is a point of infinity in Figure 4.26, and branch 3 is a wave of compression and torsion.
In Figure 4.27, x is dominant term, so branch 4 mainly is a wave of compression.
Compare the result of the water-filled shell (Figure 4.22) with that of the in-vacuo
shell (Figure 4.2), it can be found that because of the effect of the inside water loading,
two more branches begin to propagate for the water-filled shell. The shapes of the
flexure curves are similar in these cases, although the water loading changes the values
of wavenumber for the same frequency.
The dispersion curves for mode numbers m = 2 and m = 3 are also obtained, and
the results are shown in Figures 4.28 and 4.29, respectively. The similar branches can
be found for these mode numbers, and the interpretations are similar too. It is still
75
12
branch 0
branch 1
branch 2
branch 3
branch 4
10
0
0
4
*
60
x / r
/
Ratio of displacement
50
40
30
20
10
0
2
5
*
76
1.4
x / r
/ r
Ratio of displacement
1.2
1
0.8
0.6
0.4
0.2
0
0
0.5
1.5
2.5
50
45
Ratio of displacement
40
x / r
/
35
30
25
20
15
10
5
0
3
77
40
x / r
/
35
Ratio of displacement
30
25
20
15
10
5
0
4
4.5
5.5
6
*
6.5
7.5
9
8
/
x
r
/ r
Ratio of displacement
7
6
5
4
3
2
1
0
6.6
6.8
7.2
7.4
7.6
7.8
78
12
branch 0
branch 1
branch 2
branch 3
10
0
0
4
*
79
12
branch 0
branch 1
branch 2
10
0
0
4
*
10
branch 1
9
8
7
6
5
4
3
2
1
0
0
10
80
0.11
| / |
x
0.1
0.09
Ratio of displacement
| / r|
0.08
0.07
0.06
0.05
0.04
0.03
0.02
0.01
0.8
1.2
1.4
1.6
1.8
81
persion curves of the shell with ribs. A water-filled aluminum shell surrounded by a
vacuum is considered. The shell has a radius of 1 m and thickness of 0.01 m.
Figure 4.32 represents the dispersion curves of a water-filled aluminum cylinder surrounded by a vacuum for circumferential harmonic m=0, which is obtained by finding
the roots of the dispersion equation (the determinant of stiffness matrix equals zero).
From the previous sections, it is known that the two outside branches are the flexure
waves. Figure 4.33 shows the flexure waves of a water-filled aluminum cylinder surrounded by a vacuum for circumferential harmonic m=0, which is obtained by calculating the radial displacements excited by a unit force. The color represents the magnitude
of radial displacements. Larger radial displacements, which represent flexural modes,
indicate resonant conditions. It can be observed that the two outside yellow branches
present the larger values of radial displacements, which means these two branches are
the flexure waves. Figure 4.33 shows the flexure waves for a simple shell (no ribs) and
is compared for validation with figure 4.32 calculated analytically. Comparing Figures
4.32 and 4.33, it can be found that the two figures match well, which validates this new
method.
In the following case, the new method is used to obtain the flexure waves of a
thin aluminum shell which is filled and surrounded by water. The shell has the same
radius and thickness as previous cases. The radius is 1 m and the thickness is 0.01
m. The result is shown in Figure 4.34. From this figure, it also can be seen the two
outside yellow branches, which present the larger values of radial displacements. These
two branches are the flexure waves. Compare these two branches with those shown in
Figure 4.33, it can be found that at the same frequency, the wavelength of a water-filled
shell is larger than that of a submerged shell. This is because the outside water loading
acts as a mass and a resistance to the system.
82
8
7
6
5
4
3
2
1
0
30
20
10
0
*
10
20
30
83
2 m=+ im
pe (r, , x) =
e
2 m=
r (m, )
H|m| ( r)
H|m| ( a)
ei x d ,
(4.3)
where radial displacement r (m, ) is the third element of the vector (m, ).
In the acoustic far fieldm the exterior pressure is obtained by direct application of
84
eikR
.
R
(4.4)
(4.5)
y
P
M
r
The code used in the present study is verified by comparing the results with those
of Skelton [63] (page 252). In the following numerical calculations, the same parameters as those in Skeltons book have been used. A water-filled steel cylindrical shell
85
100
95
90
85
80
75
70
65
60
55
50
0
5000
10000
Frequency (Hz)
15000
Figure 4.36. Far field sound radiation from a water filled steel shell excited by
a unit radial point force at = 90o .
surrounded by air is considered. The cylindrical shell has radius 0.1 m, thickness 0.01
m. The damping factor = 0.02. The cylindrical shell is excited by a unit radial point
force.
The far field sound pressure level is defined as 20 log10 |Pe f (R = 1, , )|+120. Two
observation angles are chosen: = 90o (broadside) and = 70o . The far field sound
pressure level at = 90o (broadside) is shown in Figure 4.36 and the result at = 70o
is shown in Figure 4.37. The peaks in the figures are associated with the flexure modes
for different mode number m = 0, 1, 2, etc. By comparing these two figures with the
results in Skeltons book, it can be concluded that the results are same with those of
Skelton proving the validity of the code.
86
100
95
90
85
80
75
70
65
60
55
50
0
5000
10000
Frequency (Hz)
15000
Figure 4.37. Far field sound radiation from a water filled steel shell excited by
a unit radial point force at = 70o .
(i)
(i)
p(i) = p f + pr + pe ,
(i)
(4.6)
(i)
where, p f is the free field spectral pressure of the monopole, pr is the interior pres(i)
sure scattered as thought boundary is rigid, pe is the interior pressure due to the motion
of the elastic cylinders. p(e) is set as the exterior pressure due to the motion of the elastic
cylinder.
(i)
p f is given by,
(i)
r/c0 )
(i) Q(t
p f = 0
4 r
87
(4.7)
(i)
where 0 is the fluid density, Q(t) is the volume flow rate of the source. For a harmonic source, Q(t) = Q0 ei t , thus,
(i)
pf
(i)
ei t + eik r
(i)
= i0 Q0
.
4 r
(4.8)
Let,
(i)
i Q0
A= 0
,
4
(4.9)
eik | x x0 | i t
e
=A
.
|
x
x0 |
(i)
(i)
pf
(4.10)
(4.11)
(k(i) )2 2 .
(i)
The interior spectral pressure pr scattered as though the boundary is hard is ob(i)
tained by solving the total spectral pressure ph inside a cylinder with a hard boundary.
The general solution for this interior spectral pressure is obtained by adding a solution of the homogeneous wave equation, which is finite at the origin, to the monopole
spectral pressure; thus,
(i)
(i)
(i)
(i)
88
(4.12)
ph / r = 0, at r = a, giving,
(i)
pr
J (
i(m0 + x0 ) |m|
= i Ae
(i) r )H ( (i) a)
0
|m|
( (i) a)
J|m|
(4.13)
At r = a, we have,
p f + pr = i Aei(m0 + x0 )
]
[
( (i) a)J ( (i) a)
J|m| ( (i) r0 )H|m|
|m|
.
J|m| ( (i) r0 )H|m| ( (i) a)
( (i) a)
J|m|
(i)
(i)
(4.14)
Note that,
J|m|
(z)H|m| (z) J|m| (z)H|m|
(z) =
2i
.
z
(4.15)
So,
(i)
(i)
p f + pr
= 2Ae
i(m0 + x0 )
J|m| ( (i) r0 )
( (i) a)
(i) aJ|m|
(4.16)
For the elastic cylindrical shell, the spectral excitation vector in the absence of an
external mechanical drive is simply,
Ex (m, )
E (m, )
Er (m, )
0
0
(i)
(i)
(i)
( p(e) + pe ) + ( p f + pr )
(i)
(4.17)
where, p(e) and pe is exterior and interior fluid loading respectively, defined by Eq. (2.17).
89
S13
S11 S12
S
S23
21 S22
x (m, )
(m, )
r (m, )
0
0
J
( (i) r0 )
( (i) a)
|m|
2Aei(m0 + x0 ) (i)|m|
aJ
(4.18)
By solving this equation, the radial displacement r can be obtained, and the far field
pressure can be obtained from Eq. (4.5).
4.3.1
In this section, numerical results are presented for the far field sound pressure level
of the acoustic radiation from an infinite aluminum cylindrical shell which is filled
and submerged with water. The non-dimensional variables of wavenumber and frequency are defined as, = a, and =
a
c2 ,
and c2 = 1482m/s is the speed of sound in water. The following parameters for the
shell were used in the numerical calculations. The aluminum cylindrical shell has a
radius of 1.0m and a thickness of 0.01m. The structural damping, = 0.02, is introduced. The excitation is considered as a monopole located at different radial positions,
r0 = 0.1a, 0.3a, 0.5a and 0.8a.
The sound pressure level is defined as 20 log10 limR |RPe f (R, , )| + 120, corresponding to reference pressure of 1 micro-pascal. The radiated sound pressure is
normalized to the free field pressure of a monopole.
Figures 4.38 and 4.39 show the far field sound pressure levels of the aluminum
cylindrical shells by a monopole at the observation angle = /2 and = /3, respectively. It can be seen that the radial locations of the monopole do not have much
effects on the far field sound radiation.
90
125
120
115
r = 0.1a
0
r0 = 0.3a
110
r0 = 0.5a
r0 = 0.8a
105
100
4
*
122
120
118
r0 = 0.1a
116
r = 0.3a
0
r0 = 0.5a
114
112
r = 0.8a
0
4
*
91
= 0.02.
Case 1: The monopole is put at the axis of cylinder, that is, r0 = 0, 0 = 0,and
x0 = 0. For this case, only one mode number m = 0 contributes to the far field pressure
and the problem is axisymmetric.
Four reduced frequencies are chosen, = 0.1, 1, 2, 5, and is changed from 0
to 2 . The results are shown in Figures 4.40, 4.41, 4.42 and 4.43, respectively. A
monopole is a source which radiates sound equally in all directions. From these results,
it can be observed that for hard rubber cylinder, almost all the pressures are unit for
different frequency and different values of . For steel cylinder and Aluminum cylinder,
the pressure directivity does not change much for = 0.1 and = 1, but it changes
quite a bit for = 2 and = 5. This is because the interior fluid loading greatly
contributes to the system for larger frequencies. The peaks occur at the wavenumber at
which the determinant of matrix S is near zero.
Case 2: The monopole is put at, r0 = 0.5a, 0 = 0, and x0 = 0. Similar results
are obtained as shown in Figures 4.44, 4.45, 4.46 and 4.47, respectively. Compare
these results with those in case 1, it can be found that the figures are not symmetrical
along horizontal line anymore, because the location of the monopole is moved up to
r0 = 0.5a. For steel cylinder and Aluminum cylinder, the pressure directivity does not
change much for = 0.1 and = 1, but it changes a lot for = 5. This is because
the interior fluid loading greatly contributes to the system for larger frequencies. The
92
90
120
Aluminium
Hard Rubber
Steel
60
150
30
1.5
1
0.5
180
330
210
300
240
270
peaks occur at the wavenumbers at which the determinant of matrix S is near zero.
f = f (cos f
e x + sin f
e f ),
(4.19)
e y + sin f
e z.
e f = cos f
(4.20)
93
(4.21)
90
120
Aluminium
Hard Rubber
Steel
60
150
30
1.5
1
0.5
180
330
210
300
240
270
90
120
Aluminium
Hard Rubber
Steel
60
150
30
1.5
1
0.5
180
330
210
300
240
270
94
90
120
Aluminium
Hard Rubber
Steel
60
150
30
1.5
1
0.5
180
330
210
300
240
270
90
120
Aluminium
Hard Rubber
Steel
60
150
30
1.5
1
0.5
180
330
210
300
240
270
95
90
120
Aluminium
Hard Rubber
Steel
60
150
30
1.5
1
0.5
180
330
210
300
240
270
90
120
Aluminium
Hard Rubber
Steel
60
150
30
1.5
1
0.5
180
330
210
300
240
270
96
90
120
Aluminium
Hard Rubber
Steel
60
150
30
1.5
1
0.5
180
330
210
300
240
270
where
x 0 is the location of the dipole excitation, and r is the radial displacement,
which is calculated by,
x (m, )
(m, )
r (m, )
0 ,
Fer
(4.22)
where matrix R is the inverse matrix of stiffness matrix S, = k cos is the stationary
phase wavenumber, and Fer is given by,
Fer = 2Aei(m0 + x0 )
97
J|m| ( r0 )
( a) ,
aJ|m|
(4.23)
0
0
r = R33 Fer .
x 0 r =
f
x 0 r =
(r k sin )
J|m|
0
(4.24)
im
ik cos , k sin
,
J|m| (r0 k sin ) r0
{
ik cos cos f , k sin
(r k sin )
J|m|
0
r Fer .
(4.25)
sin f cos( f 0 )
(4.26)
}
im
sin f sin( f 0 ) r Fer .
r0
98
Px =
ie 2 eikR
(i f k cos f cos )
kR sin
m=+ (i)|m| ei(m0 +x0 k cos ) J (r k sin )
|m| 0
2 Z|m|ak sin J (ak sin )H (ak sin ) eim ,
m=
|m|
|m|
(4.27)
2
2
akR(sin )
m=0 Z|m| J|m| (ak sin )H|m| (ak sin )
Px =
ak2 (sin )2
m= J (r k sin ) cos m( )
0
|m| 0
Z|m|J (ak sin )H (ak sin ) .
m=0
|m|
|m|
(4.29)
125
120
115
r0 = 0
110
r0 = 0.1a
r0 = 0.3a
105
r0 = 0.5a
r0 = 0.8a
100
4
*
Figure 4.49 shows the far field sound pressure levels of the aluminum cylindrical
shell caused by axial dipoles at the observation angle = /3 with azimuthal angle
= 0. It can be seen that the farfield sound radiation depends on the radial location
of axial dipoles. When is less than 3.5, the far field sound level increases as the
location of the dipoles changes from r0 = 0.0 to r0 = 0.8a. But, when is larger
than 3.5, the far field sound levels of axial dipoles located at r0 = 0.5a and r0 = 0.8a,
decrease as the frequency increases.
Figures 4.50 shows the far field sound pressure levels of the aluminum cylindrical
shell caused by axial dipoles at the observation angle = 0 and = /4. The results
are similar to figure 4.49. It is noticed that the far field sound level increases about 9
dB as the location of axial dipole moves from the axis of the shell to r0 = 0.8a.
From the results of sound radiation, it can be observed that the far field sound radi-
100
125
120
115
r0 = 0
110
r0 = 0.1a
r0 = 0.3a
105
r0 = 0.5a
r0 = 0.8a
100
4
*
ation depends on the radial location of the axial dipole. In the next example, the effect
of the azimuthal observation angle will be examined. An axial dipole is considered,
which is located at 0 = 0 with different radial positions, r0 = 0, 0.1a, 0.3a, 0.5a and
0.8a. The frequency is chosen as = 3, and the observation angle is = /4. And
the azimuthal angle is changed from = 0o to = 360o . Figure 4.51 represents the
far field sound pressure levels of an aluminum cylindrical shell. From this figure, it can
be seen the far field sound pressure level depends on the azimuthal angle. Depending
on the azimuthal angle, the far field sound pressure level may increase or decrease as
the location of the axial dipole is moved towards the shell wall. Therefore, the far field
sound pressure level depends on the radial position of axial dipoles and the azimuthal
angle of the observation.
101
125
120
115
110
105
100
95
90
r = 0.1a
85
80
75
70
0
r0 = 0.3a
r = 0.5a
0
r0 = 0.8a
50
100
150
200
(degree)
250
300
350
Figure 4.51. Far field sound pressure levels of the aluminum cylindrical shell
caused by axial dipoles at = /4 with different azimuthal angles.
102
90
1
120
60
Aluminium
Hard Rubber
Steel
0.8
0.6
150
30
0.4
0.2
180
330
210
300
240
270
Pr =
103
(4.30)
90
1
120
60
0.8
Aluminium
Hard Rubber
Steel
0.6
150
30
0.4
0.2
180
330
210
300
240
270
90
1
120
60
Aluminium
Hard Rubber
Steel
0.8
0.6
150
30
0.4
0.2
180
330
210
300
240
270
104
90
1
120
60
Aluminium
Hard Rubber
Steel
0.8
0.6
150
30
0.4
0.2
180
330
210
300
240
270
90
1
120
60
Aluminium
Hard Rubber
Steel
0.8
0.6
150
30
0.4
0.2
180
330
210
300
240
270
105
90
1.5
120
60
Aluminium
Hard Rubber
Steel
1
150
30
0.5
180
330
210
300
240
270
90
1.5
120
60
Aluminium
Hard Rubber
Steel
1
150
30
0.5
180
330
210
300
240
270
106
90
120
60
Aluminium
Hard Rubber
Steel
1.5
150
30
0.5
180
330
210
300
240
270
107
130
120
110
100
r0 = 0
90
r0 = 0.1a
80
r0 = 0.3a
r0 = 0.5a
70
r0 = 0.8a
60
4
*
130
120
110
100
r0 = 0
90
r0 = 0.1a
80
r0 = 0.3a
r = 0.5a
70
r = 0.8a
0
60
4
*
108
0.6
M=0
M=1
M=2
M=3
0.4
DJ
|m| 0
(r k sin )
0.2
0.2
0.4
0.6
0.8
6
r k sin
10
12
(r k sin ).
Figure 4.62. Derivative of Bessel function J|m|
0
difference between the formulation (4.29) for axial dipole and the equation (4.30) for
(r k sin ) for different mode numbers are shown in
radial dipole. The values of J|m|
0
Figure and 4.62. Because the far field sound pressure is the summation of all the mode
numbers, the big drop in Figures 4.60 and 4.61 are related to some specific mode
(r k sin ) is zero. This suggested that when the
number, at which the value of J|m|
0
radial dipole excitation is applied to the duct, the normal velocity of the duct motion
could be zero for some frequency, which means there is no coupling between the duct
and fluid flow. The mode does not radiate the sound to the farfield.
From the results of the axial dipole, it was found that the far field sound radiation depends on the azimuthal angle of observation. In the next example, the effect
of the azimuthal observation angle will also be examined for the radial dipole. A radial dipole is considered, which is located at 0 = 0 with different radial positions,
109
120
110
100
r = 0.1a
0
r0 = 0.3a
90
r = 0.5a
0
r0 = 0.8a
80
70
60
0
50
100
150
200
(degree)
250
300
350
Figure 4.63. Far field sound pressure levels of the aluminum cylindrical shell
caused by radial dipoles at = /4 with different azimuthal angles.
r0 = 0, 0.1a, 0.3a, 0.5a and 0.8a. The frequency is chosen as = 3, and the observation angle is = /4. And the azimuthal angle is changed from = 0o to = 360o .
Figure 4.63 represents the far field sound pressure levels of an aluminum cylindrical
shell. From this figure, it can be seen the far field sound pressure level depends on the
azimuthal angle. The far field sound pressure level depends on the radial position of
radial dipoles and the azimuthal angle of the observation.
4.4.2.2
110
90
1.5
120
60
Aluminium
Hard Rubber
Steel
1
150
30
0.5
180
330
210
300
240
270
Case 1:
which is located at axis of the cylinder (r0 = 0, 0 = 0, x0 = 0). The results are shown in
Figure 4.64, 4.65, 4.66 and 4.67. For low frequency, the pressure directivities of three
materials change a little, but for higher frequencies, they change a lot. This is because
the fluid loading has more effect for larger frequency. The peaks occur at wavenumber
at which determinant of matrix S is zero. The sharp directivity pattern for hard rubber
shell are also observed.
Case 2: A radial dipole is put at, r0 = 0.5, 0 = 0, and x0 = 0. Similar results are
obtained as shown in Figures 4.68, 4.69, 4.70 and 4.71. Compared with case 1, it is
found that the figures are not symmetrical along horizontal line anymore, because the
location of dipole is moved up to r0 = 0.5.
111
90
1.5
120
60
Aluminium
Hard Rubber
Steel
1
150
30
0.5
180
330
210
300
240
270
90
1.5
120
60
Aluminium
Hard Rubber
Steel
1
150
30
0.5
180
330
210
300
240
270
112
90
1.5
120
60
Aluminium
Hard Rubber
Steel
1
150
30
0.5
180
330
210
300
240
270
90
1.5
120
60
Aluminium
Hard Rubber
Steel
1
150
30
0.5
180
330
210
300
240
270
113
90
1.5
120
60
Aluminium
Hard Rubber
Steel
1
150
30
0.5
180
330
210
300
240
270
90
1.5
120
60
Aluminium
Hard Rubber
Steel
1
150
30
0.5
180
330
210
300
240
270
114
90
1
120
60
Aluminium
Hard Rubber
Steel
0.8
0.6
150
30
0.4
0.2
180
330
210
300
240
270
Px =
4.4.3.1
(4.31)
In this section, the far field sound pressure level of the acoustic radiation from an
infinite aluminum cylindrical shell is studied, which is filled and submerged with water.
The excitation is considered as a circumferential dipole which is located at different
115
130
125
120
115
r0 = 0
r0 = 0.1a
110
r0 = 0.3a
r0 = 0.5a
105
r0 = 0.8a
100
4
*
radial positions, r0 = 0, 0.1a, 0.3a, 0.5a and 0.8a. The other computational conditions
are the same as in section 4.3.1.
Figures 4.72 and 4.73 show the far field sound pressure levels of the aluminum and
ribbed cylindrical shells caused by circumferential dipoles at the observation angle =
/2 and = /4, respectively. From these figures, it is found that the far field sound
level increases as the location of the dipole changes from r0 = 0 to r0 = 0.8a, which
means the circumferential dipole has stronger effect on the far field sound radiation
when the location of the circumferential dipole is closer to the cylindrical shell.
In the next example, the effect of the azimuthal observation angle will also be examined for the circumferential dipole. A circumferential dipole is considered, which is
located at 0 = 0 with different radial positions, r0 = 0, 0.1a, 0.3a, 0.5a and 0.8a. The
frequency is chosen as = 3, and the observation angle is = /4. And the az-
116
130
125
120
115
r0 = 0
r0 = 0.1a
110
r0 = 0.3a
r0 = 0.5a
105
r0 = 0.8a
100
4
*
imuthal angle is changed from = 0o to = 360o . Figure 4.74 represents the far field
sound pressure levels of an aluminum cylindrical shell. Again, it can be seen the far
field sound pressure level depends on the azimuthal angle. The far field sound pressure
level depends on the radial position of radial dipoles and the azimuthal angle of the
observation.
117
120
110
100
90
r0 = 0.1a
80
r = 0.3a
0
70
r0 = 0.5a
r = 0.8a
0
60
50
0
50
100
150
200
(degree)
250
300
350
Figure 4.74. Far field sound pressure levels of the aluminum cylindrical shell
caused by circumferential dipoles at = /4 with different azimuthal angles.
results are shown in Figure 4.75, 4.76, 4.77 and 4.78. The sharp directivity at high
frequency suggests that there are non-compact source effects due to interaction with
shell.
4.5 Conclusions
The results of a numerical study of the far field sound radiation from an infinite thin
cylindrical shell is presented in this chapter.
By solving the dispersion relation, it is found that there are three types of propagating waves in the thin shell which correspond to flexure, torsion and compression
waves. The flexure waves are of interest because they are related with the far field
radiated sound, and the curves of the flexure waves can be used to explain the strong
acoustic radiation from the shell.
118
90
120
Aluminium
Hard Rubber
Steel
60
150
30
0.8
0.6
0.4
0.2
180
330
210
300
240
270
90
120
Aluminium
Hard Rubber
Steel
60
150
30
1.5
1
0.5
180
330
210
300
240
270
119
90
120
Aluminium
Hard Rubber
Steel
60
150
30
1
0.8
0.6
0.2
180
0.4
0
330
210
300
240
270
90
120
Aluminium
Hard Rubber
Steel
60
150
30
1
0.8
0.6
0.4
0.2
180
330
210
300
240
270
120
Different excitation sources are examined in this chapter: radial point force, monopole
and dipoles. In the spectra of the far field sound radiation, the peaks are associated with
the flexure waves. The results of the sound pressure level radiated from a dipole inside
a shell clearly show strong dependence of the sound pressure level on the radial position
of the dipole and its orientation. The results show that sound pressure level radiations
can reach 5-10 dB. These results demonstrate the inadequate assumption often used in
structural acoustics wherein distributed dipole sources are treated as a single blocked
dipole when the acoustic transfer function is calculated.
121
CHAPTER 5
SOUND RADIATION FROM THIN SHELLS WITH RIBS
In this chapter, the effect of the ribs on the far field radiated sound from an infinite
cylindrical shell will be studied. The structures, such as submarines, are often stiffened
with ribs for structural integrity. A rib is assumed to run circumferentially around the
shell and is schematically shown in Figure 5.1. In this figure, M is the observer point in
the spherical coordinate system. R is the distance between the source and observer, is
the polar angle, and is the azimuthal angle. The excitation sources could be a single
radial force, monopole, dipole or the forces produced by a propeller inside the shell.
The spectral response of the elastic cylindrical shell without ribs is given by Eq. (2.16).
The solutions of the shell displacements can be obtained by a simple matrix inversion
from Eq. (2.16),
x (m, )
(m, )
r (m, )
fx
f ,
fr
(5.1)
The structural damping can be included in the equations of the motion as a loss factor.
If damping is included, the Youngs Modulus E is replaced by Ec ,
Ec = E(1 i ),
122
M (R , , )
Water
Dipoles
rt
rh
Force
Water
o
Ribs
Propeller
)
(m,
x
(m, )
)
(m,
r
(m, )
123
Fx
,
Fr
(5.2)
where,
R14 = i R13 ,
R24 = i R23 ,
R41 = i R31 ,
R42 = i R32 ,
R34 = i R33 ,
R43 = i R33 ,
R44 = 2 R13 .
s q=
q=
(5.3)
q=
where s is rib spacing, Fe is the excitation, and q is the number of ribs. Matrix B
is a 4 4 dynamics stiffness matrix, which connects the forces and displacements at
cylinder attachment points. The T-rib model and the elements in matrix B are presented
in Appendix I .
In this chapter, we use a simple rib model wherein the ribs are represented by simple
masses. Assuming rib runs circumferentially around the cylindrical shell, the schematic
of the rib stiffened cylindrical shell is shown in Figure 5.2. The radius of the shell is
a and the ribs cross section is rectangular, with a depth of h and a width of b. The
radial displacement and the radial force per unit length are w and qr , respectively. The
circumferential displacement and the circumferential force per unit length are v and q ,
respectively. The cross-sectional area of the rib is A = bh and the bending moment of
inertia is I = bh3 /12.
124
v, q
b
w, qr
a
h
A
= q ( ,t),
a 2
t2
(5.4)
[
]
EA v
2w
EI 4 w
+
+
w
+
A
= qr ( ,t),
a2
a4 4
t2
(5.5)
where, E is the Youngs modulus of the rib, is the density of the rib and a is the radius
of the duct.
Now assuming an expansion in a harmonic series for the direction and ei t time
dependence, consider the nth term.
v( ,t) = Vn ei t ein ,
(5.6)
w( ,t) = Wn ei t ein ,
(5.7)
q ( ,t) = Q n ei t ein ,
(5.8)
125
(5.9)
BV = Q,
(5.10)
Vn
V =
,
Wn
and Q is the excitation:
(5.11)
Q n
Q=
,
Qrn
(5.12)
B11 B12
B=
.
B21 B22
(5.13)
The elements in the stiffness matrix of the rib are given by,
B11 =
EAn2
2 A,
a2
B12 = i
B22 =
EAn
,
a2
(5.14)
(5.15)
B21 = B12 ,
(5.16)
EA EIn4
+ 4 2 A.
a2
a
(5.17)
126
5.1 Dispersion Curves of Waves for the Water-Filled Shell with Ribs
When the cylindrical shell is stiffened by control ribs, the dispersion relation will be
strongly modified by the control ribs, and the dispersion curves can be used to explain
the strong acoustic radiation from the duct.
For the shell with ribs, it is not possible to use the simple analytical method to
obtain dispersion curves. From Eq. (5.3), it can be found that the spectral displacements
of the shell with ribs are composed of two terms: the first term, of axial wavenumber
, is the spectral displacements of the shell without ribs; the second term, of discrete
axial wavenumber + 2 q/s, comes from the interactions between the periodically
spaced ribs. Since the forces, which are applied to the shell and the ribs, depend on
different wavenumbers, the dispersion curves can not be obtained by solving the roots
of dispersion relation, |S( , )| = 0. We have developed a new method to find the
dispersion curves. A unit force is applied to the system and the displacements of the
shell are calculated. The dispersion curves can be obtained by plotting displacements
versus frequency and wavenumber. Since the flexure waves are of our interest, only the
radial displacements are plotted against frequency and wavenumber. The larger values
of radial displacements represent the flexure waves, which are related to the far field
acoustic pressure. This method has been validated in the previous chapter.
The flexure waves in an infinite cylindrical shell will be obtained in this section. A
water-filled aluminum cylindrical shell is considered, which is submerged in water. In
the following calculation, the shell has a radius of 1.0m and a thickness of 0.01m. The
steel rib dimensions are: depth h = 0.06m and width b = 0.06m. The rib spacing is 1
m. One percent structural damping is included in the equations of the motion.
First, the effects of the ribs on the flexure waves are examined. The flexure curve of
an unribbed shell with that of an ribbed shell is compared. Figure 5.3 shows the flexure
127
waves of a water-filled aluminum unribbed shell surrounded by water for the circumferential harmonic m=0, which is excited by a unit force. As examined in the previous
chapter, it is known that the yellow branches are the flexure waves. Figure 5.4 represents the flexure waves of a water-filled aluminum ribbed cylinder surrounded by water
for circumferential harmonic m=0, which is also excited by a unit force. Comparing the
two figures, it is seen that when the cylindrical shell is stiffened by ribs, the ribs strongly
change the dispersion curves and lead to free modes with much smaller wavelength. It
can also be found that for the shell without ribs, there are always free flexure modes for
all the frequencies. While for the shell with ribs, there are no free flexure modes for
some frequency zones indicating that the far field radiated sound is very weak for these
frequencies. Those zones are called silence zones. This result suggests that the control
ribs can be used to reduce noise for some frequencies.
128
The flexure waves of the ribbed shell for mode number m = 1, 5, 8, 10 are obtained
as shown in Figures 5.5, 5.6, 5.7 and 5.8, respectively. For different mode numbers,
the flexure curves are different and the locations of silence zones are different. These
results clearly show the strong influence of the ribs on the radiated sound. The results
suggest that the radiated sound is enhanced at some frequencies, and weakened for
some frequency zones.
5.2 Dispersion Curves of Waves for the Air-Filled Shell with Ribs
In this section, the dispersion curves of an air-filled shell with ribs are studied. The
geometry and material of the shell is provided by Scott Morris at the University of
Notre Dame. An infinite cylindrical shell with ribs is considered. The material of the
shell is Nickel The radius of the duct is 0.10287 meter, and the thickness of the shell
129
is 0.00006858 meter. The material of the ribs is steel. The width of the rib is 0.0127
meter, the depth of the ribs is 0.0381 meter, and the distance between two ribs is 0.2032
meter.
In the following numerical calculation, the circumferential mode number M = 10.
The reduced frequency is defined as = a/cair , where a is the radius of the shell,
and cair = 343m/s is the density of the air. The non-dimensional wavenumber is defined
as = a.
Figure 5.9 presents the magnitude of the radial displacements of the shell surrounded by air for circumferential harmonic m=10, which is excited by a unit force.
It can be observed that the air-filled shell has the similar pattern of dispersion curves
as the water-filled shell, although the fluid loading effect is very small for the air-filled
shell. Again, for the shell with ribs, there are no free flexure modes for some frequency
130
zones indicating that the far field radiated sound is very weak for these frequencies.
Figure 5.10 presents the magnitude of the circumferential displacements of the shell,
and Figure 5.10 shows the magnitude of the axial displacements of the shell. Comparing
figure 5.10 with figure 5.9, it can be seen that the patterns of the two types of dispersion
curves are almost same, but the magnitudes of circumferential displacements is smaller
than those of radial displacements. Similarly, it can be seen that the magnitudes of
axial displacements is much smaller than those of radial displacements. It means that
the radial displacements are the dominant term. Therefore, the dispersion curves in
figure 5.9 mainly are the flexure waves.
131
133
134
for the shell and rib were used in the numerical calculations. The aluminum cylindrical
shell has a radius of 1.0m and a thickness of 0.01m. The steel rib dimensions are depth
0.06m and width 0.06m. The rib spacing is 1.0m. The structural damping, = 0.02, is
introduced.
5.3.1
A radial point force excitation has been used to determine the dispersion relation.
Here, the broadside far field sound pressure levels of the thin aluminum shell is shown
in figure 5.12. Comparing the far field sound pressure level of the shell with ribs and
that of the shell without ribs, it can be found that the control ribs strongly change the
far field sound radiations. The peaks in the spectrum can be associated with the flexure
waves in the shell. At some frequencies, the sound level of ribbed shell is much lower
than that of unribbed shell. The reason could be the cancellation between the waves
due to control ribs.
Figure 5.13 shows 45o off broadside far field sound pressure levels of the thin aluminum shell when the excitation is a unit radial force. Again, it can be found that the
control ribs strongly change the far field sound radiations.
135
150
140
130
120
110
100
90
with ribs
No ribs
80
70
4
*
field sound radiation. To test the effect of the ribs on the far field sound radiation, these
two figures are compared with Figures 4.38 and 4.39 in the previous chapter. It is seen
that the flexure modes cause significant change to the radiated sound because of the
interaction with the control ribs. The peaks in the spectrum can be associated with the
flexure modes in the flexure curves.
136
150
140
130
120
110
100
with ribs
No ribs
90
80
70
4
*
Figure 5.13. Far field sound pressure levels at 45o off broadside
( = /4, = 0) of ribbed and unribbed cylindrical shells caused by a unit
radial point force.
137
125
120
115
r0 = 0.1a
110
r0 = 0.3a
r = 0.5a
0
105
100
r0 = 0.8a
4
*
122
120
118
r0 = 0.1a
116
r = 0.3a
0
r0 = 0.5a
114
112
r0 = 0.8a
1
4
*
138
125
120
115
r0 = 0
110
r0 = 0.1a
r0 = 0.3a
105
r0 = 0.5a
r0 = 0.8a
100
4
*
drical shell caused by axial dipoles at the observation angles = /3 and = /4,
respectively. It is seen that the farfield sound radiation depends on the radial location
of axial dipoles. When is smaller than 3.5, the far field sound level increases as the
location of the dipoles changes from r0 = 0.0 to r0 = 0.8a.
Comparing these two figures with the Figures 4.49 and 4.50 shown in the previous
chapter, it can be seen that the control ribs change the far field sound level. The peaks
in the plot occur near the shell flexure modes. It is noticed that the far field sound level
increases about 9 dB as the location of axial dipole moves from the axis of the shell to
r0 = 0.8a.
139
125
120
115
r0 = 0
110
r0 = 0.1a
r0 = 0.3a
105
r0 = 0.5a
r0 = 0.8a
100
4
*
= /4, respectively. Comparing these two figures with the Figures 4.72 and 4.73
shown in the previous chapter, it can be seen that the control ribs change the far field
sound level. From these figures, it is found that the far field sound level increases as the
location of the dipole changes from r0 = 0 to r0 = 0.8a, which means the circumferential dipole has stronger effect on the far field sound radiation when the location of the
circumferential dipole is closer to the cylindrical shell.
140
130
125
120
115
r0 = 0
r0 = 0.1a
110
r0 = 0.3a
r0 = 0.5a
105
r0 = 0.8a
100
4
*
130
125
120
115
r0 = 0
r0 = 0.1a
110
r0 = 0.3a
r = 0.5a
105
r = 0.8a
0
100
4
*
141
130
120
110
100
r0 = 0
90
r0 = 0.1a
80
r0 = 0.3a
r0 = 0.5a
70
r0 = 0.8a
60
4
*
142
130
120
110
100
r0 = 0
90
r0 = 0.1a
80
r0 = 0.3a
r0 = 0.5a
70
r0 = 0.8a
60
4
*
5.4 Conclusions
The results of a numerical study of the far field sound radiation from an infinite thin
cylindrical shell with control ribs are presented in this chapter.
It is found that when the cylindrical shell is stiffened by the ribs, the control ribs
strongly change the dispersion curves, and the silence zones are found for the shell with
ribs. This suggests that control ribs can be used to reduce noise for some frequencies.
When the source excitation is a radial point force, the control ribs strongly change
the far field sound radiations. The peaks in the spectrum are associated with the flexure
waves in the shell. And for some frequencies the far field sound level can be reduced
by the cancellation between the waves due to the control ribs.
For different radial positions of a monopole in a rib-stiffened duct, the locations of
the peaks in the spectrum are almost the same, although the magnitude are different.
143
The peaks in the spectrum are associated with the flexure modes in the flexure curves.
For dipoles, the radial positions significantly change the sound pressure level for
the axial, radial and circumferential orientations. As mentioned in the previous chapter,
these results demonstrate the assumption ofblocked dipole is inadequate when the
acoustic transfer function is calculated.
144
CHAPTER 6
SOUND RADIATION FROM A PROPELLER
In the previous chapters, the sound radiation from the submerged structures, such
as plate and cylindrical shells, has been studied. Structural acoustics focuses on assumed sources of noise, e.g., single force, monopole, dipoles. When sound radiation
from a propeller is investigated, the source of noise is obtained from the propellerduct system. In this chapter, the effects of an elastic duct on the sound radiation from
a propeller inside the duct will be examined. The coupling of the duct motion and the
propeller generated flow is implemented using Eulers equations and the shell equations
by exchanging information between the fluid and the shell at the boundaries.
The propeller, which works with incoming nonuniform flow, is modeled as a rotor/stator stage. Because of the rotor rotation, it is necessary to represent the propagation of upstream disturbances in a swirling mean flow [6]. For a uniform flow, the
pressure and vorticity are uncoupled. For a swirling mean flow, Atassi et al. [6, 33]
developed a model for the representation of upstream disturbances and their interaction
with a row of blades. The results showed that the swirling flow may strongly affect the
acoustic and vortical spectral composition of propagating modes in the duct.
The governing equations for the fluid motion are the linearized Euler equations
discussed in chapter 2. The mean swirling flow U(x) is vortical and can be assumed
145
(6.1)
where Ux and U are the mean velocity components in the axial and circumferential
directions, respectively; ex and e represent unit vectors in the axial and circumferential
directions, respectively. It is assumed that the flow is isentropic and all the length scales
are normalized to the mean radius rm .
In this chapter, the origin of the x-axis is taken at the blade midchord at the mean
radius. The blades are placed along the mean flow streamlines, and the computations
are taken in the curvilinear coordinates,
= x,
(6.2)
= r,
(6.3)
xU
.
rUx
(6.4)
U = r + ,
r
(6.5)
where, is an angular velocity and is the strength of the free vortex. It is further
assumed that the flow is isentropic with a uniform enthalpy from hub to tip. Thus, the
axial component of the mean velocity is given by,
2
) + 2ln(r/rm )],
Ux2 = Um2 2[2 (r2 rm
146
(6.6)
where Um is the axial velocity at the mean radius of the duct rm . In the code, the velocity
is defined in terms of the Mach number, such as M0 = Um /com , M = (r)/com and
M = /(rcom ), where com is the sound speed at rm .
The scattering phenomena in a rigid duct will first be examined, then the scattering phenomena in an elastic duct will be studied. At last, the sound radiation from a
propeller-elastic-duct system will be investigated.
(6.7)
147
(6.8)
Rigid
rtip
Rigid
rhub
Incoming
disturbance
gth
Le
n
Bla
de
Ch
or d
Stagger
angle
Rigid
(6.9)
where e = 2 /V is the angular spacing and = 2pB /V is the inter blade phase angle.
V and B are the number of stator blades and the number of rotor blades, respectively.
3. At the inlet and outlet of the computational domain, the non-reflecting boundary
conditions are applied.
At the inlet, a gust is introduced, and the initial conditions for this problem are given
by,
where we take,
)
mU(r) xU(r)
(u)
am =
,
, 0 Am ,
r
(
)
1
mUs
x =
x,
Ux
r
U = Ux2 +Us2 ,
148
(6.10)
(6.11)
(6.12)
(u)
Am =
a(u) eikn r ,
(6.13)
n =
kn =
n
,
rt rh
r = r rh .
(6.14)
(6.15)
For our calculations, a(u) is the amplitude of the disturbance velocity, which is equal to
1 m/s.
L
,
0 ca(u)U
(6.16)
where L is the force per unit span and c is the chord length. All the quantities are
calculated at the radial location of the considered section. The pressure is normalized
(u)
(u)
by 0m c0m am , where 0m , c0m and am are the mean density, the mean speed of sound,
and the upwash at the mean radius, respectively.
In this numerical example, the number of rotor blades is taken as B = 16, and the
number of stator blades V = 24. A swirl flow of M0 = 0.3536, M = 0.1, and M = 0.1
is considered. The non-dimensional chord at the mean radius is 0.3491. The reduced
frequency = rm /c0m = 3 . At the inlet of the computational domain, a unit gust
upwash amplitude, a(u) = 1, is imposed. The hub-tip ratio is chosen as rh /rt = 0.6.
The numerical result is shown in figure 6.2, in which nx, nt and nr are the number of
grids in x-axis, -axis and r-axis, respectively. For this case, there are two propagating
modes in the duct. The effect of different grid points on the results is examined. From
149
0.75
nx=181,nt=nr=31
nx=121,nt=nr=31
nx=121,nt=nr=21
nx=91,nt=nr=21
nx=61,nt=nr=21
Atassi, et.al (2004)
0.7
0.65
0.6
|c |
0.55
0.5
0.45
0.4
0.35
0.3
0.75
0.8
0.85
0.9
0.95
1.05
1.1
1.15
1.2
1.25
r/rm
the figure, it can be seen that results are a little bit different for different grid points.
Comparing the case of nx = 181, nt = nr = 31 with the case of nx = 61, nt = nr = 21,
the biggest difference can be found as about 8 percent.
Comparing these results with the result of Atassi et al. (2004), it can be concluded
that our results agree well with the result of Atassi et al. (2004). The better result
could be obtained by increasing the grid points, but that requires more computational
time. From figure 6.2, it can be concluded that the number of grid points of nx = 121,
nt = nr = 31 gives us petty good agreement with the result of Atassi et al. (2004).
inlet of the computational domain, we impose a unit gust upwash amplitude, a(u) = 1.
The hub-tip ratio is chosen as rh /rt = 0.6.
Figure 6.3 shows the distribution of pressure magnitude along the x-axis at the duct
wall for reduced frequency = 0.5. = 0 is the location of the stator blade, and
= 10 is in the middle of two stator blades. From this figure, a peak value of pressure
can be observed at = 0, which means the duct wall bears more pressure when the
flow passes the blade. For this uniform flow, there is no propagating modes in the duct.
Figure 6.4 represents the unsteady sectional lift coefficient along the span for reduced
frequency = 0.5. It can be seen that the magnitude of the unsteady sectional lift does
not change much from the hub to the tip.
For the case of the reduced frequency = 1, figures 6.5 and 6.6 show the distribution of pressure magnitude along the x-axis and the unsteady lift coefficient along
the span, respectively. When the reduced frequency is increased to 3, the results are
presented in figures 6.7 and 6.8, respectively. From these figures, it can be seen that the
magnitudes of the pressure and the unsteady lift are reduced as the frequency increased.
But, for the uniform flow, the magnitude of the unsteady sectional lift does not change
much from the hub to the tip.
151
0.25
=0
=/10
0.2
|p*|
0.15
0.1
0.05
0
1
0.5
0
*
x
0.5
Figure 6.3. Pressure magnitude distribution along the x-axis at the duct wall
for reduced frequency = 0.5.
152
0.8
0.6
real part
imag part
abs value
0.4
|c |
0.2
0
0.2
0.4
0.6
0.8
0.8
0.9
1
r*
1.1
1.2
Figure 6.4. Unsteady sectional lift coefficient along the span for reduced
frequency = 0.5.
153
(6.17)
0.25
=0
=/10
0.2
|p*|
0.15
0.1
0.05
0
1
0.5
0
*
x
0.5
Figure 6.5. Pressure magnitude distribution along the x-axis at the duct wall
for reduced frequency = 1.
and the radial component of the velocity at the duct surface is given by,
Vr =
DF
F
F
F
=
+Vx
+V
,
Dt
t
x
r
(6.18)
where Vx and V are the mean velocity components in the axial and circumferential
directions, respectively.
If the unsteady velocity is small compared to its corresponding mean value, the
above equation can be linearized about the steady mean flow quantities by writing,
V (x,t) = U(
x) +u(x,t),
(6.19)
|u| |U|.
(6.20)
154
0.8
0.6
0.4
real part
imag part
abs value
|c |
0.2
0
0.2
0.4
0.6
0.8
0.8
0.9
1
r*
1.1
1.2
Figure 6.6. Unsteady sectional lift coefficient along the span for reduced
frequency = 1.
Moreover, at the duct surface, the radial component of the fluid mean velocity is
zero, Ur = 0, therefore the radial unsteady velocity of the fluid is obtained by,
(f)
ur =
F
F
F
.
+Ux
+U
t
x
r
(6.21)
ur =
F
.
t
(6.22)
155
(6.23)
0.25
=0
=/10
0.2
|p*|
0.15
0.1
0.05
0
1
0.5
0
*
x
0.5
Figure 6.7. Pressure magnitude distribution along the x-axis at the duct wall
for reduced frequency = 3.
thus, the radial unsteady velocities of the fluid motion and the duct motion become,
(f)
ur = i F +Ux
F
F
+U
,
x
r
(6.24)
and
(l)
ur = i F.
(6.25)
(l)
Now, we can express F in terms of ur ,
i (l)
F = ur .
156
(6.26)
0.8
real part
imag part
abs value
0.6
0.4
|c |
0.2
l
0
0.2
0.4
0.6
0.8
0.8
0.9
1
r*
1.1
1.2
Figure 6.8. Unsteady sectional lift coefficient along the span for reduced
frequency = 3.
0.8
0.7
|cl|
0.6
0.5
0.4
0.3
0.2
0.75
0.8
0.85
0.9
0.95
1.05
1.1
1.15
1.2
1.25
r/rm
157
Elastic
rtip
Elastic
rhub
Bla
Incoming
disturbance
gth
Le
n
de
Ch
or d
Stagger
angle
Elastic
(f)
(f)
ur
(l)
= ur +
or
(f)
ur
(l)
(l)
ur
ur
Ux
+U
x
r
(
)
iUx
iU
(l)
= 1+
+
ur .
x r
(6.27)
(6.28)
Now assume that the relation between the pressure and the velocity at the duct
surface is,
(l)
p = z ur ,
(6.29)
z = R + iX,
158
(6.30)
[
(
)]
1
i
=
1+
Ux +U
p( f ) .
z
x
r
(6.31)
(6.32)
where i is the non dimensional impedance, 0m is the fluid density, and c0m is the
sound speed in fluid. The schematic of the computational domain is shown in figure
6.10.
The boundary conditions and initial conditions are as follow. At the hub and the
blade surfaces, we have rigid surfaces, for which the impermeability condition is applied. At the tip, the surface is elastic, and the boundary condition is given by Eq. (6.31).
At the surfaces of upstream and downstream of blade section, the periodicity conditions
are applied. At the inlet and outlet of the computational domain, the non-reflecting
boundary conditions are applied. At the inlet, a gust is introduced (Eq. (6.10)), and the
magnitude of velocity is 1m/s.
In the following example, an elastic duct is considered. The computational domain
is limited to three times of the projection of the chord length. The projection of the blade
along the x-axis extends from (c/2)cosm to +(c/2)cosm . The inlet of the computational domain is located at x = (3c/2)cosm , and the outlet is at x = +(3c/2)cosm .
A uniform flow with a low Mach number of 0.05 is chosen. The number of rotor blades
159
is taken as B = 2, and the number of stator blades V = 10. The non-dimensional chord
at the mean radius is 2 /V = 0.628. The hub-tip ratio is chosen as rh /rt = 0.6.
Case One:
First the effects of the acoustic resistance, i = 2, and acoustic reactance, i = 2i,
are examined. The reduced frequency is chose as = 1. Figure 6.11 shows the pressure
distribution along the x-axis at the duct wall for = 0. The results of the rigid duct and
the elastic duct with different impedances are compared in this figure. A close up for
the peak part of the pressure distribution is shown in Figure 6.12. It can be observed
that the peak values of the rigid duct and the elastic duct with i = 2 are very close,
which means that the acoustic resistance does not affect the pressure distribution very
much. Figure 6.13 shows a close up for the middlepart the pressure distribution along
the x-axis. It also can be seen that the pressure distributions of the rigid duct and the
elastic duct with i = 2 are very close. But the pressure of the elastic duct with i = i2
is smaller than other cases. This means that the acoustic reactance has more effect on
the pressure distribution than the acoustic resistance.
Figure 6.14 shows the pressure distribution along the x-axis at the duct wall for =
/10. It is very clear that the pressure at the duct wall is reduced by the impedances.
Figure 6.15 presents the magnitudes of the unsteady sectional lift coefficient along the
span. Again it can be observed that the lift coefficient is reduced more by the acoustic
reactance.
Case Two:
In the second case, the value of the impedance is fixed to be i = 1.0 1.0i, and the
reduced frequency are chose as = 0.5, 1 and 3.
Figure 6.16 shows the distribution of pressure magnitude along the x-axis at the
duct wall for reduced frequency = 0.5. = 0 is the location of the stator blade, and
160
0.18
rigid
impedance=i2
impedance=2
0.16
0.14
0.12
|p*|
0.1
0.08
0.06
0.04
0.02
0
1
0.5
0.5
x*
Figure 6.11. Pressure magnitude distribution along the x-axis at the duct wall
for different impedances ( = 0) .
rigid
impedance=2
impedance=i2
0.173
0.172
0.171
0.17
|p*|
0.169
0.168
0.167
0.166
0.165
0.164
0.315
0.3145
0.314
0.3135
0.313
x*
Figure 6.12. A close up for the peak part of the pressure magnitude
distribution along the x-axis at the duct wall for different impedances ( = 0) .
161
0.04
rigid
impedance=i2
impedance=2
0.035
0.03
|p*|
0.025
0.02
0.015
0.01
0.005
0.5
0.4
0.3
0.2
0.1
0.1
0.2
Figure 6.13. A close up for the middle part of the pressure magnitude
distribution along the x-axis at the duct wall for different impedances ( = 0) .
= 10 is in the middle of two stator blades. Just like the case of rigid duct, a peak value
of pressure can be observed at = 0, which means the duct wall bears more pressure
when the flow passes the blade. For this uniform flow, there is no propagating modes in
the duct. Figure 6.17 represents the unsteady sectional lift coefficient along the span for
reduced frequency = 0.5. It can be seen that the magnitude of the unsteady sectional
lift does not change much from the hub to the tip.
For the case of the reduced frequency = 1, figures 6.18 and 6.19 show the distribution of pressure magnitude along the x-axis and the unsteady lift coefficient along
the span, respectively. When the reduced frequency is increased to 3, the results
are presented in figures 6.20 and 6.21, respectively. From these figures, it can be seen
that the magnitudes of the pressure and the unsteady lift are reduced as the frequency
increased. But, for the uniform flow, the magnitude of the unsteady sectional lift does
162
0.015
rigid
impedance=i2
impedance=2
|p*|
0.01
0.005
0
1
0.5
0.5
Figure 6.14. Pressure magnitude distribution along the x-axis at the duct wall
for different impedances ( = /10) .
0.54
rigid
impedance=i2
impedance=2
0.52
|cl|
0.5
0.48
0.46
0.44
0.42
0.75
0.8
0.85
0.9
0.95
1.05
1.1
1.15
1.2
1.25
r*
163
0.25
=0
=/10
0.2
|p*|
0.15
0.1
0.05
0
1
0.5
0
*
x
0.5
Figure 6.16. Pressure magnitude distribution along the x-axis at the duct wall
for reduced frequency = 0.5 (impedance i = 1.0 1.0i).
= 0. The close-ups for the peak part and the middle part of the pressure distribution
at the duct wall are shown in Figures 6.23 and 6.24, respectively. From these figures, it
can be seen that the wall pressure is reduced more as the magnitude of the impedance
decreases. The result of the pressure distribution along the x-axis at the duct wall for
= /10 is shown in Figure 6.25. From this figure, it can be seen more clearly that the
164
0.8
0.6
0.4
real part
imag part
abs value
|c |
0.2
0
0.2
0.4
0.6
0.8
0.8
0.9
1
r*
1.1
1.2
Figure 6.17. Unsteady sectional lift coefficient along the span for reduced
frequency = 0.5 (impedance i = 1.0 1.0i).
magnitude of the impedance has strong effect on the pressure field in the duct.
Figure 6.26 presents the magnitudes of the unsteady sectional lift coefficient along
the span. In this figure, it can be observed that the unsteady lift coefficient of the
impedance i = 10 10i is very close to that of the rigid duct. When the impedance is
reduced to i = 1.01.0i, the magnitude of the unsteady lift coefficient is reduced about
15 percent. Thus it can be concluded that the impedance begin to have the significant
effect on the lift when the magnitude of the impedance is the order of one. When the
impedance is further reduced to i = 0.01 0.01i, the lift coefficient is reduced about
40 percent at the mean radius of the duct.
In the previous chapter, the far field sound radiation from cylindrical shells is calculated by assumed sources of noise, such as single force, monopole and dipole excitations. In this chapter, the source of noise is obtained from the rotor/stator duct system.
165
0.25
=0
=/10
0.2
|p*|
0.15
0.1
0.05
0
1
0.5
0
*
x
0.5
Figure 6.18. Pressure magnitude distribution along the x-axis at the duct wall
for reduced frequency = 1 (impedance i = 1.0 1.0i).
6.2.2
Effect of Discontinuities
In this section, the effects of the impedance discontinuities will be studied. Only
part of the duct wall is considered as elastic, and the other part of the wall is rigid.
166
0.8
0.6
0.4
real part
imag part
abs value
|c |
0.2
0
0.2
0.4
0.6
0.8
0.8
0.9
1
r*
1.1
1.2
Figure 6.19. Unsteady sectional lift coefficient along the span for reduced
frequency = 1 (impedance i = 1.0 1.0i).
The impedance is fixed in this section, i = 1.0 1.0i. The computational domain
is limited to six times of the projection of the chord length. Figure 6.28 shows the
schematic of computational domain of the duct with impedance discontinuities. The
inlet of the computational domain is located at x = (6c/2)cosm , and the outlet
is at x = +(6c/2)cosm . The projection of the blade along the x-axis extends from
(5c/2)cosm to (4c/2)cosm . Other parameters are the same as the previous section.
Several locations of the elastic wall are chosen: (1) The elastic wall is located from
the leading edged (L.E.) of the blade to the trailing edge(T.E.) of the blade; (2)The
elastic wall is located from T.E. to one chord length of T.E.; (3) The elastic wall is
located from L.E. to two chord length of L.E.; (4) The elastic wall is located from the
inlet to L.E.; (5)The elastic wall is located from the inlet to T.E.
167
0.25
=0
=/10
0.2
|p*|
0.15
0.1
0.05
0
1
0.5
0
*
x
0.5
Figure 6.20. Pressure magnitude distribution along the x-axis at the duct wall
for reduced frequency = 3 (impedance i = 1.0 1.0i).
The results of different locations of the elastic wall are compared with those of the
fully rigid wall and the fully elastic wall. Figures 6.29 shows the pressure distribution
along the x-axis at the duct wall for = 0. The close-ups for the peak part and the
middle part of the pressure distribution at the duct wall are shown in Figures 6.30 and
6.31, respectively. Figures 6.32 shows the pressure distribution along the x-axis at the
duct wall for = /10. Figure 6.33 presents the magnitudes of the unsteady sectional
lift coefficient along the span.
From those plots, it can be seen that the pressure and the lift coefficient are not
affected much by the elastic wall located from T.E. to one chord length of T.E. When
the elastic wall is from L.E. to T.E., it has significant effect on the lift coefficient which
is reduced about 11 percent based on the lift coefficient of the fully rigid duct. If the
elastic wall covers from the inlet to T.E., the lift coefficient is close to that of the fully
168
0.8
real part
imag part
abs value
0.6
0.4
|c |
0.2
0
0.2
0.4
0.6
0.8
0.8
0.9
1
r*
1.1
1.2
Figure 6.21. Unsteady sectional lift coefficient along the span for reduced
frequency = 3 (impedance i = 1.0 1.0i).
elastic wall. Thus it can be concluded that the main effect of the elastic wall comes
from the location of the blade and the upstream of the blade. The elastic wall located
downstream of the blade trailing edge does not affect the unsteady lift coefficient much.
0.18
rigid duct
10i10
1i
0.5i0.5
0.1i0.1
0.01i0.01
0.16
0.14
0.12
|p*|
0.1
0.08
0.06
0.04
0.02
0
1
0.8
0.6
0.4
0.2
0.2
0.4
0.6
0.8
x*
Figure 6.22. Pressure magnitude distribution along the x-axis at the duct wall
for different impedances ( = 0) .
rigid duct
10i10
1i
0.5i0.5
0.1i0.1
0.01i0.01
0.17
0.165
|p |
0.16
0.155
0.15
0.145
0.14
0.317
0.316
0.315
0.314
0.313
0.312
0.311
Figure 6.23. A close up for the peak part of the pressure magnitude
distribution along the x-axis at the duct wall for different impedances ( = 0) .
170
0.04
rigid duct
10i10
1i
0.5i0.5
0.1i0.1
0.01i0.01
0.035
0.03
|p |
0.025
0.02
0.015
0.01
0.005
0
0.5
0.4
0.3
0.2
0.1
0.1
Figure 6.24. A close up for the middle part of the pressure magnitude
distribution along the x-axis at the duct wall for different impedances ( = 0) .
0.015
rigid duct
10i10
1i
0.5i0.5
0.1i0.1
0.01i0.01
|p*|
0.01
0.005
0
1
0.8
0.6
0.4
0.2
0.2
0.4
0.6
0.8
Figure 6.25. Pressure magnitude distribution along the x-axis at the duct wall
for different impedances ( = /10) .
171
0.5
|cl|
0.4
0.3
rigid duct
10i10
1i
0.5i0.5
0.1i0.1
0.01i0.01
0.2
0.1
0
0.8
0.9
1
*
r
1.1
1.2
and hydroacoustic scattering. The coupling effects between the fluid motion and the
duct vibration will be investigated.
The following numerical scheme is used. The computational grid for the unsteady
flow uses the mean flow streamlines as body-fitted coordinates. Equations (2.29) and
(2.30) are then integrated using a fourth-order Runge-Kutta integrator. The resulting
algebraic system of equations is solved using a PETSc package based on an iterative
biconjugate gradient stabilized technique and incomplete lower-upper factorization.
The boundary conditions and initial conditions are as follow. At the hub and the
blade surfaces, we have rigid surfaces, for which the impermeability condition is applied. At the surfaces of upstream and downstream of blade section, the periodicity
conditions are applied. At the inlet and outlet of the computational domain, the nonreflecting boundary condition is applied. At the inlet, a gust is introduced, and the initial
172
90
60
120
30
150
0.03
0.04
0.02
0.01
180
210
330
240
300
270
Figure 6.27. Far filed sound pressure directivity for an aluminum duct. The
noise sources are obtained from the rotor/stator duct system .
conditions are given by Eq. (6.10). At the tip, the surfaceis elastic, and the boundary
condition is given by Eq. (2.38). And the impedance for the elastic duct is calculated
by the following equation,
i =
1
imn R33
(6.33)
where mn is an eigenvalue of the convected operator D0 /Dt , and R33 is the third element in the stiffness matrix for the elastic shell. In this equation, mn is a real number,
but R33 could be a real number or a purely imaginary number, which is determined by
the exterior fluid loading. From chapter 4, it is known that the imaginary part of the
interior fluid loading term is zero, which means there is no resistance term for interior
fluid loading. And the imaginary part of the exterior fluid loading could also be zero
only for the case where /c0 < .
In the following example, an aluminum shell is considered. The shell is submerged
173
Elastic
rtip
rhub
Rigid
gth
Le
n
Bla
de
Ch
or d
Stagger
angle
Incoming
disturbance
Elastic
174
0.12
Fully Rigid Wall
Fully Elastic Wall (EW)
EW 1 c from T.E.
EW 2 c from L.E.
EW Blade only
EW 1 c upstream of L.E.
EW 2 c upstream of T.E.
0.1
|p*|
0.08
0.06
0.04
0.02
0
2
1.5
0.5
0.5
1.5
x*
Figure 6.29. Pressure magnitude distribution along the x-axis at the duct wall
of impedance discontinuities ( = 0) .
0.116
0.115
|p |
0.114
0.113
0.112
0.111
0.11
1.2585
1.258
1.2575
1.257
1.2565
1.256
1.2555
1.255
x*
Figure 6.30. A close up for the peak part of the pressure magnitude
distribution along the x-axis at the duct wall of impedance discontinuities
( = 0) .
175
0.035
Fully Rigid Wall
Fully Elastic Wall (EW)
EW 1 c from T.E.
EW 2 c from L.E.
EW Blade only
EW 1 c upstream of L.E.
EW 2 c upstream of T.E.
0.03
0.025
|p |
0.02
0.015
0.01
0.005
1.5
1.4
1.3
1.2
1.1
0.9
0.8
0.7
0.6
0.5
x*
Figure 6.31. A close up for the middle part of the pressure magnitude
distribution along the x-axis at the duct wall of impedance discontinuities
( = 0) .
0.015
Fully Rigid Wall
Fully Elastic Wall (EW)
EW 1 c from T.E.
EW 2 c from L.E.
EW Blade only
EW 1 c upstream of L.E.
EW 2 c upstream of T.E.
|p*|
0.01
0.005
0
2
1.5
0.5
0.5
1.5
x*
Figure 6.32. Pressure magnitude distribution along the x-axis at the duct wall
of impedance discontinuities( = /10) .
176
0.5
0.45
cl
0.4
Fully Rigid Wall
Fully Elastic Wall (EW)
EW 1 c from T.E.
EW 2 c from L.E.
EW Blade only
EW 1 c upstream of L.E.
EW 2 c upstream of T.E.
0.35
0.3
0.25
0.8
0.9
1.1
1.2
in water, with a radius of 1 m, and a thickness of 0.01m. The axial wavenumbers are
chosen as 0.5 and 5.0. The fluid loading term is normalized by 0 c0 , where 0 is the
density of water and c0 is the sound speed in water. Four mode numbers are chosen,
m = 0, 2, 8, 12. The results for wavenumbers = 0.5 and 5.0 are shown in Figures
6.34 and 6.35, respectively. It can be found when frequency /c0 < , the imaginary
part of exterior water loading is zero. Thus, for the lower frequency, the element R33
also is a real number. Therefore, the impedance for the elastic duct could be purely
imaginary for the lower frequency.
In the following example, the material of the shell is chosen as aluminum and hard
rubber. The shell is surrounded by water, with a radius of 1 m, and a thickness of
0.01m. The impedance is calculated for two reduced frequencies = 1, 3, and the
reduced frequency is defined as = a/c0 . Four mode numbers are chosen, m =
177
0
m=12
m=8
m=2
m=0
0.5
1.5
2.5
3.5
0
10
12
14
0.5
1.5
m=12
m=8
m=2
m=0
2.5
3.5
0
10
12
14
178
TABLE 6.1
IMPEDANCES OF THE ALUMINUM SHELL.
Al ( = 1)
Al ( = 3)
m=0
= 0.5
17.8-2.7i
60.0-7.0i
m=0
= 5.0
2.9i
-50.6i
m=2
= 0.5
1.2-13.3i
71.1-28.1i
m=2
= 5.0
0.98i
-40.9i
m=-8
= 0.5
-2.9i
0.0002-30.0i
m=-8
= 5.0
-2.2i
-31.8i
m=12
= 0.5
-1.0i
-19.7i
m=12
= 5.0
-0.34i
-22.3i
0, 2, 8, 12. For each mode number, two wavenumbers are chosen = 0.5, 5.0. The
results are shown in table 6.1. From this table, it can be found that the impedance
strongly depends on the frequency, wave number and mode number. This suggests that
the correct approach to calculate acoustics must be fully coupled between fluid and
elastic duct because there is no single impedance value for the duct.
In the following numerical example, the material of the elastic duct is considered
as aluminum. The elastic duct is submerged and filled with water. The number of rotor
blades is B = 2, and the number of the stator blades is V = 10. The mean-flow total
Mach number is 0.05 and the hub-tip ratio is 0.6. Reduced frequency of 1 is chosen and
a grid of {nx n nr } = {121 41 41} is used. The pressure along the duct wall
is obtained by solving the linearized Euler equations (2.29) and (2.30). The far field
sound pressure level is calculated by equation (4.5).
179
For this aluminum duct, the non-dimensional impedances are i = 3.18i, 7.55i, 3.54i,
which are correspond to the mode number m = 12, 2, 8, respectively. Figures 6.36
and 6.37 represent the real part of duct wall pressure and imaginary part of duct wall
pressure in response to rotor/stator interaction for a rigid duct. Figures 6.38 and 6.39
represent the real part of duct wall pressure and imaginary part of duct wall pressure in
response to rotor/stator interaction for an elastic duct. Comparing these figures, it can
be observed that the wall pressure is changed by the elastic wall.
Figure 6.40 shows the comparison of magnitude of the blade sectional lift coefficient
for the rigid duct and elastic duct. It can be observed that the lift coefficient is reduced
by the impedance.
Sound pressure directivity for rigid wall excitation and elastic wall excitation is
shown in Figure 6.41. The excitations are applied to an aluminum duct with a radius
of 1 meter and a thickness of 0.01 meter. The pressure is normalized by 0U0 ug , where
0 is the fluid density, U0 is the mean fluid velocity and ug is the velocity of gust. It
can be seen that the blade force is mainly a dipole source. And the acoustic radiation is
reduced when the excitation is elastic wall excitation.
6.4 Conclusions
In this chapter, the scattering phenomena in a rigid duct was studied first and the
effect of the swirling on the propagating modes and the unsteady lift has been examined.
The results shows that the new propagating modes exist for the swirling flow, and the
unsteady lift is increased by the swirling flow.
Then the duct was considered as an elastic duct with different impedances. The
effect of the impedance on the acoustic sources was examined. The results show that the
impedance begin to have the significant effect on the unsteady lift when the magnitude
180
0.04
0.03
0.02
p*
0.01
0
0.01
0.02
0.03
0.04
0.8
1
0.6
0.5
0.4
0
0.2
0.5
0
x*
Figure 6.36. Real part of duct wall pressure in response to rotor stator
interaction for rigid duct.
0.04
0.03
0.02
p*
0.01
0
0.01
0.02
0.03
0.04
0.8
1
0.6
0.5
0.4
0
0.2
0.5
0
x*
Figure 6.37. Imaginary part of duct wall pressure in response to rotor stator
interaction for rigid duct.
181
0.04
p*
0.02
0.02
0.04
0.8
0.6
1
0.5
0.4
0.2
0.5
0
x*
Figure 6.38. Real part of duct wall pressure in response to rotor stator
interaction for elastic duct.
0.04
p*
0.02
0.02
0.04
0.8
0.6
1
0.5
0.4
0.2
0.5
0
x*
Figure 6.39. Imaginary part of duct wall pressure in response to rotor stator
interaction for elastic duct.
182
0.5
0.4
Elastic wall
Rigid wall
0.3
0.2
0.1
0.8
0.9
1.1
1.2
r*
Figure 6.40. Comparison of unsteady lift coefficient between rigid duct and
elastic duct.
90
0.04
60
120
0.03
0.02
150
30
0.01
180
210
330
240
300
270
183
of the non dimensional impedance is the order of one. The main effect of the elastic wall
comes from the location of the blade and the upstream of the blade. The elastic wall,
which is located downstream of the blade trailing edge , does not affect the unsteady
lift coefficient much.
At last, a model for flow-propeller interactions in a submerged elastic duct has been
developed. This model examines and quantifies the mechanism of flow-propeller interaction in a flexible duct. The model couples the fluid motion with the elastic duct
vibration and yields the duct flexural displacement. This leads to the evaluation of the
radiated sound.
The present model was applied to the rotor/stator interaction problem in an elastic
duct. It was founded that the elasticity of the shell weakens the acoustic sources. The
results suggest that for different combinations of rotor/stator blade counts, it is possible
to have low circumferential mode number, which is an efficient radiator of acoustic
energy.
184
CHAPTER 7
CONCLUSIONS
In this dissertation, a model was developed for fluid-structure interactions in a submerged elastic duct. The model coupled the fluid motion with the elastic duct vibration
at the duct boundary. The model brings together two classical fields: hydro-acoustics
and structural acoustics.
The sound radiation from an infinite plate with or without ribs was first studied to
understand the characteristics of flexure waves. Different excitations sources were considered: radial single force, monopole, dipole and vortex. For the un-stiffened plate,
there is a single propagating flexure wave when the frequency is lower than the coincidence frequency. This flexure wave can not propagate to the far field. When the
frequency is higher than the coincidence frequency, there is no purely propagating flexure wave. But the leaky wave is found for the frequency higher than the coincidence
frequency. The leaky wave is very important over a narrow range of angle c , and it
is related to the lobes in the spectra of the far field sound pressure level of different
excitation sources.
When the plate is stiffened with ribs, the dispersion curves are strongly changed.
And there are many propagating flexure waves found for a single frequency. Some of
these flexure waves can propagate to the far field. Both the pass and stop band structures
are found in the dispersion curves. They are corresponding to the peaks in the spectra
of different excitation sources.
185
By solving the dispersion relation of a thin cylindrical shell, it is fond that there are
three types of propagating waves in the thin shell which correspond to flexure, torsion
and compression waves. The flexure waves are of interest because they are related with
the far field radiated sound, and the curves of the flexure waves can be used to explain
the strong acoustic radiation from the shell.
The case of a cylindrical duct with ribs is also considered and the dispersion relation
of the stiffened duct modes is compared with that of a un-stiffened duct. It is found that
when the cylindrical shell is stiffened by the ribs, the control ribs strongly change the
dispersion curves. The dispersion relation of the stiffened duct has a periodic structure
similar to that of connected oscillators with large number of independent modes. It is
also found that for the stiffened shell, there are no propagating flexure modes for some
frequency zones, which means that the far field radiated sound is very weak for these
frequencies. Those zones are called silence zones. For different mode number, the
locations of silence zones are different too. This suggests that the control ribs can be
used to reduce noise for some frequencies.
When the source excitation is a radial point force, the control ribs strongly change
the far field sound radiations. The peaks in the spectrum are associated with the flexure
waves in the shell. And for some frequencies the far field sound level can be reduced
by the cancellation between the waves due to the control ribs.
The results of the sound pressure level radiated from a dipole inside the duct clearly
show strong dependence of the sound pressure level on the radial position of the dipole
and its orientation. The results show that sound pressure level radiations can reach 510 dB. These results demonstrate the inadequate assumption often used in structural
acoustics wherein distributed dipole sources are treated as a single blocked dipole
when the acoustic transfer function is calculated.
186
The scattering phenomena in a rigid duct was studied first and the effect of the
swirling on the propagating modes and the unsteady lift has been examined. The results
shows that the new propagating modes exist for the swirling flow, and the unsteady lift
is increased by the swirling flow.
The duct was also considered as an elastic duct with different impedances. The
effect of the impedance on the acoustic sources was examined. The results show that
the impedance begin to have the significant effect on the unsteady lift when the magnitude of the non-dimensional impedance is the order of one. The main effect of the
elastic wall comes from the location of the blade and the upstream of the blade. The
elastic wall, which is located downstream of the blade trailing edge, does not affect the
unsteady lift coefficient much.
A model for flow-propeller interactions in a submerged elastic duct has been developed in this dissertation. This model examines and quantifies the mechanism of
flow-propeller interaction in a flexible duct. The model couples the fluid motion with
the elastic duct vibration and yields the duct flexural displacement. This leads to the
evaluation of the radiated sound.
The present model was applied to the rotor/stator interaction problem in an elastic
duct. It was founded that the elasticity of the shell weakens the acoustic sources. The
results suggest that for different combinations of rotor/stator blade counts, it is possible
to have low circumferential mode number, which is an efficient radiator of acoustic
energy.
187
APPENDIX A
RIB MODEL
In this part, a rather detailed rib model is described. We treat the case of a cylindrical
shell stiffened by a large number of ribs. The treatment includes dynamical modeling
of these ribs to allow for deformations of their cross-section during vibration. In what
follows, all the differentials are represented by subscripts for convenience. For example,
( / r) is denoted as ,r .
A.1
For a circular ring of radius ar and cross sectional area A, the strain energy is given
by [66] [15],
U=
ar
2
r r dAd +
1
(Gr Jr )
2ar
(r, + x, /ar )2 d
(A.1)
r =
r + ,
yx + xy
ar
(A.2)
In Eq. (A.2)., the coordinates x and y are measured from the centroid of the ring
and x , and r are the axial, circumferential and radial displacements of the ring at
its centroid. The deformation r is the out-of-plane rotation of the ring. The ring hoop
188
r = E r r
(A.3)
r, ,
a2r
(A.4)
x, + ar r
a2r
(A.5)
x =
y =
Substitute them into Eq. (A.1), and drop the terms which contains only ring parameters
[15], we obtain the expression for the strain energy as,
U=
ar Er
2
[(
r + ,
ar
)2
A + x2
y2 dA + y2
x2 dA 2x y
1
+
(Gr Jr )
2ar
U=
ar E r
2
[(
r + ,
ar
]
xydA d
(r, + x, /ar )2 d
)2
A + x2 Ixx + y2 Iyy 2x y Ixy
1
+
(Gr Jr )
2ar
(r, + x, /ar )2 d
(A.6)
where Ixx and Iyy are the moments of inertia about x and y axis respectively, Ixy is the
product moment of inertia, and Gr Jr is the torsional rigidity.
189
U=
ar
2
(A.7)
Iyy
Iyy = 4 ,
ar
Ixx
Ixx = 4 ,
ar
Ixy
Ixy = 4 ,
ar
Jr
Jr = 4 .
ar
(A.8)
r ar
T=
2
2
2
[A(x,t
+ 2,t + r,t2 ) + I p r,t
]d
(A.9)
where r is the mass density of the ring and I p is the polar moment of inertia of the ring
section, I p = Ixx + Iyy .
A.3
Applied forces
Assume the forces applied on the shell are line loads. The work done by such a
system of forces is,
W = ar
(Fx x + F + Fr r + M r )d
190
(A.10)
where Fx , F and Fr are the corresponding load intensities of line loads, and M is
the moment excitation. Any concentrated force can be treated as a special case of the
corresponding line load.
(A.11)
L
i
L
=0
i
(A.12)
From Eq. (A.7), Eq. (A.9) and Eq. (A.10) we know that U/ = 0, W / = 0
and T / = 0, thus the Lagranges equation becomes,
d
dt
)
+
U W
=
(A.13)
Using Lagranges equation, we can get the equations of the ring as,
d
dt
T
x,t
r ar
=
2
d
(2Ax,t )d =
dt
W
=
x
ar Fx d
191
r ar Ax,tt d
U
=
x
ar
2
Fx =
(A.14)
Similarly, we can obtain the other three equations of motion of the ring as,
F =
r, + , ) + Er Ixx (r, + , )
Er A(
Er Ixy (x, ar r, ) + r A ,tt
Fr =
r + , ) + Er Ixx (r, , )
Er A(
Er Ixy (x, + ar r, ) + r Ar,tt
M =
(A.15)
(A.16)
192
(A.17)
x = xm ei(m t) ,
= m ei(m t)
r = rm ei(m t) ,
r = rm ei(m t)
Fx = Fxm ei(m t) ,
F = F m ei(m t)
Fr = Frm ei(m t) ,
M = M m ei(m t)
Substitute these equations into Eq. (A.14) to Eq. (A.17), we can obtain following
equations,
Fxm =
F m =
rm m2 m ) + Er Ixx (im3 rm m2 m )
[Er A(im
Er Ixy (im3 xm imar rm )] r A 2 m
Frm =
(A.18)
(A.19)
193
(A.20)
M m =
(A.21)
(A.22)
where ym and Fm are the ring displacements and the forces in the x , , r and r
coordinates with a wave number m. The matrix Bm is used to denote the impedance
matrix of size (4 4) and its elements are defined as,
B11 = m4 Er Iyy m2 Gr Jr r A 2 ,
B13 = m4 Er Ixy ,
B21 = B12 ,
B14 = m2 Er Iyy ar m2 Gr Jr ar
B22 = m2 Er A + m2 Er Ixx r A 2
B32 = B23
B33 = Er A + m4 Er Ixx r A 2 ,
B34 = m2 Er Ixy ar
B41 = B14 ,
B42 = B24
B43 = B34 ,
194
APPENDIX B
THIN SHELL THEORIES
The equations of shell motion have been studied by various researchers. A brief
review of thin shell theories is presented in appendix B.
Numerous thin shell theories are available in the literature. These theories are slight
different because the assumptions are different. In this chapter, only those theories
which use Loves first approximations, are considered.
The deformation of a thin shell can be determined by the displacements of its middle
surface. The total strains at any point can be represented as the sum of two parts [44],
one due to stretching and the other due to bending. The theories of Byrne, Flugge,
Goldenveizer, and Novozhilov et al. are the direct results of application of the Kirchhoff
hypothesis to the stain-displacement relationships.
e =
1
( + z ),
1 + z/R
(B.1)
e =
1
( + z ),
1 + z/R
(B.2)
where e and e are normal stains, and are normal stains in the middle surface
(z = 0). z measures the distance of the point from the corresponding point on the middle
surface along the normal direction and varies over the thickness of the shell. R and
R are the curvature radius in and directions, respectively. and are the
midsurface changes in curvature.
195
1
1+z/R .
(B.3)
where {ui } is displacement and [L ] is a matrix differential operator, which is the sum
of two terms,
[L ] = [LDM ] + [LMOD ],
(B.4)
h2
,
12R2
(B.5)
where h is the thickness of the shell and R is the radius of the shell. provides an
indication of the relative contribution of the strains of the shells surface. The larger ,
the more important is wall flexure in controlling the dispersion relation. Therefore it
affects the cut off frequency of the flexural modes.
196
In the book of Authur Leissa[44], different thin shell theories are presented. It is
shown that the theories of Biezeno-Grammel and Vlasov are the only ones that yield
the correct zero frequency (corresponding to rigid body translation in the transverse
direction) in the case of circumferential mode number of unit. When circumferential
mode is larger than 2, most theories are in close agreement.
197
BIBLIOGRAPHY
198
199
202
This document was prepared & typeset with LATEX 2 , and formatted with
nddiss2 classfile (v3.0[2005/07/27]) provided by Sameer Vijay.
203