Sie sind auf Seite 1von 207

DISPERSE User's

A system for Generating Dispersion Curves


Manual

Michael Lowe
and
Brian Pavlakovic

Non-Destructive Testing Laboratory


Department of Mechanical Engineering
Imperial College London
London, SW7 2AZ
UK
Version
Email: m.lowe@imperial.ac.uk
2.0.20a Web: www.imperial.ac.uk/nde

July 2013 Copyright M Lowe, B Pavlakovic (c) 2013


Contents

1 Introduction 10
1.1 What Disperse Can Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.2 Limitations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.3 What Else Disperse Can Do . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.4 Installing and running Disperse . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
1.5 Display Logic and Mouse Actions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
1.6 New Features for Version 2.0 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
1.7 Warnings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15

2 Getting Started Quickly 16


2.1 Quick Start - Flat Isotropic Free Plate . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.2 Quick Start - Flat Isotropic Plate in Water . . . . . . . . . . . . . . . . . . . . . . 18

3 Defining the Structure 20


3.1 Getting Started . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
3.2 Defining the Geometry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
3.2.1 Some Simple Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
3.3 Defining a Layer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
3.3.1 Liquid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
3.3.2 Isotropic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
3.3.3 Anisotropic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
3.3.4 Variation of properties with frequency . . . . . . . . . . . . . . . . . . . . . 33
3.3.5 Interface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
3.3.6 Boundary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
3.4 Opening an Existing File . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35

4 Tracing the Dispersion Curves 36


4.1 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
4.2 Automatic Tracing of Dispersion Curves . . . . . . . . . . . . . . . . . . . . . . . . 37
4.3 Manually Tracing Dispersion Curves . . . . . . . . . . . . . . . . . . . . . . . . . . 39
4.3.1 Sweeping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
4.3.2 Tracing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
4.3.3 Converging on Roots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
4.4 Special Solution Options . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
4.4.1 Types of Solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
4.4.2 Material Property Variation . . . . . . . . . . . . . . . . . . . . . . . . . . . 48

5 Displaying the Dispersion Curves 51


5.1 The Basics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
5.2 Default two-dimensional Display Types . . . . . . . . . . . . . . . . . . . . . . . . 52
5.2.1 Phase Velocity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
5.2.2 Group Velocity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52

2
5.2.3 Attenuation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
5.2.4 Real Wave Number . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
5.3 Other Forms of Display . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
5.3.1 Second Graph . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
5.3.2 Select Axes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
5.3.3 Hide Curves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
5.3.4 Derived Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
5.3.5 Sweep Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
5.4 Using the Display . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
5.4.1 Zooming and Moving . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
5.4.2 Using the Pointer in the Display Window . . . . . . . . . . . . . . . . . . . 58
5.4.3 Toolbar and Edit Options . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
5.5 Preferences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60

6 Tools for Processing the Results 62


6.1 Mode Shapes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
6.1.1 Lines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
6.1.2 Grid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
6.1.3 Power . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
6.2 Simulated signal . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
6.2.1 Excitation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
6.2.2 Multi-mode . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
6.3 Calculate function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
6.3.1 Input Using Menu Options . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
6.3.2 Input Using General Equation . . . . . . . . . . . . . . . . . . . . . . . . . 76
6.3.3 Sampling Information . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
6.4 Show Bulk Velocities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
6.5 Verify Solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
6.6 Resample Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
6.7 Input and Output of Information . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
6.7.1 Saving . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
6.7.2 Exporting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
6.7.3 Import Text File . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
6.7.4 Copy and Paste . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
6.7.5 Printing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
6.8 Labels . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84

7 Wave Propagation Model 86


7.1 Wave Equations in Bulk Isotropic Media - General Theory . . . . . . . . . . . . . . 87
7.1.1 Properties of the Wave Equation with Attenuation . . . . . . . . . . . . . . 90
7.1.2 Treatment for Fluid Materials . . . . . . . . . . . . . . . . . . . . . . . . . . 91
7.2 Wave Propagation in Isotropic Flat Plate Structures . . . . . . . . . . . . . . . . . 96
7.2.1 Historical Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
7.2.2 Plane Waves in an Infinite Elastic Solid . . . . . . . . . . . . . . . . . . . . 96
7.2.3 Plane waves in a Two-Dimensional Space . . . . . . . . . . . . . . . . . . . 98
7.2.4 The Superposition of Plane Waves in a Layered Plate . . . . . . . . . . . . 99
7.3 Cylindrical Wave Propagation in Isotropic Materials . . . . . . . . . . . . . . . . . 101
7.3.1 Historical Background of Cylindrical Wave Propagation . . . . . . . . . . . 101
7.3.2 Assumptions and Limitations . . . . . . . . . . . . . . . . . . . . . . . . . . 102
7.3.3 Boundary Conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
7.3.4 Waves in Finite Layers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
7.4 Cartesian Wave Propagation in Orthotropic Media . . . . . . . . . . . . . . . . . . 113
7.4.1 Bulk Waves in Anisotropic Media . . . . . . . . . . . . . . . . . . . . . . . . 113
7.4.2 Lamb-Type Modes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114

3
7.4.3 Shear Horizontal Modes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
7.4.4 The Layer Matrix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
7.5 Global Matrix Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
7.6 Spring Interface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
7.7 Analytical Solutions for Leaky Lamb Case . . . . . . . . . . . . . . . . . . . . . . . 123
7.8 Finding a Root . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
7.9 Tracing a Dispersion Curve . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
7.10 Mode Shapes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130

8 Cartesian Examples 133


8.1 Lamb Waves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
8.1.1 Tracing the Curves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
8.1.2 Interpreting the Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
8.2 Leaky Lamb Waves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138
8.3 Surface and Interface Waves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140
8.3.1 True Rayleigh Waves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142
8.3.2 Leaky Rayleigh Waves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142
8.3.3 Stoneley Waves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
8.3.4 Thin Layer on a Half-Space . . . . . . . . . . . . . . . . . . . . . . . . . . . 144
8.4 Visco-Elastic Layers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147
8.5 Multi-Layered Structures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149
8.6 Embedded Layers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150
8.7 Anisotropic Layers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153

9 Cylindrical Examples 162


9.1 Cylinders in Vacuum . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163
9.1.1 Tracing the Dispersion Curves . . . . . . . . . . . . . . . . . . . . . . . . . 163
9.1.2 Projections . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164
9.1.3 Naming . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164
9.1.4 Nature of the Modes in Solid Cylinders . . . . . . . . . . . . . . . . . . . . 165
9.1.5 Nature of the Modes in Hollow Cylinders . . . . . . . . . . . . . . . . . . . 166
9.1.6 Effect of Changing the Radius . . . . . . . . . . . . . . . . . . . . . . . . . 168
9.2 Cylinders Immersed in a Fluid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 169
9.2.1 Modelling the Fluid Inside the Cylinder . . . . . . . . . . . . . . . . . . . . 169
9.2.2 Leakage into the Surrounding Medium . . . . . . . . . . . . . . . . . . . . . 173
9.3 Cylinders Embedded in a Solid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175
9.3.1 Different Types of Leakage . . . . . . . . . . . . . . . . . . . . . . . . . . . 176
9.3.2 Multi-layered, Visco-elastic, Pipes, Embedded in a Solid . . . . . . . . . . . 177

10 Additional Information 179


10.1 File Structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 179
10.2 How to Get in Contact . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 180
10.3 This User Manual . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 180

A Summary of Useful Relations 181


A.1 Relations between Lambda, Rho and Mu . . . . . . . . . . . . . . . . . . . . . . . 181
A.2 Wave Speeds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181
A.3 Constitutive and Compatibility Relations for Elastic Isotropic Material . . . . . . . 182
A.4 Relations between Velocity and Attenuation Constants for Isotropic Materials . . . 183
A.5 Engineering and Cij Constants for Orthotropic Materials . . . . . . . . . . . . . . 184
A.6 Attenuation of waves in materials with complex Cij constants . . . . . . . . . . . . 186
A.7 Bessel Function Recurrence Relations . . . . . . . . . . . . . . . . . . . . . . . . . 187
A.8 Operations in Cylindrical Coordinates . . . . . . . . . . . . . . . . . . . . . . . . . 188

4
B List of Examples 189

C Common Material Properties 191

D Bibliography 194

E Index 200

5
List of Tables

7.1 Substitutions that should be made and criteria that should be used for the selection
of the type of Bessel functions to be used for the cylindrical layer matrix. . . . . . 106
7.2 Criteria for the the choice of phase for the arguments of the Bessel functions de-
pending on the type of Bessel function and the type of wave (homogeneous or
inhomogeneous). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108

8.1 Stacking sequence of example 8-ply composite. . . . . . . . . . . . . . . . . . . . . 158

9.1 Material constants used in the cylindrical wave propagation examples. . . . . . . . 178

C.1 Isotropic Solid Materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 192


C.2 Fluid Materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193
C.3 Anisotropic Materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193

6
List of Figures

3.1 A Bond transformation matrix is used to rotate an anisotropic material. The orien-
tation is specified by a rotation about the x̂ axis, followed by a rotation about the
ŷ axis and a second rotation about the x̂ axis. The angles are specified in degrees. 27
3.2 A fibrous material is typically defined in the literature with the X axis parallel to
the fibres. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
3.3 View of how Disperse sees the example material after the constants have been
entered using “Trans.Isotropic (Cij’s - along x)” but before any rotation angles
have been specified. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
3.4 View of the example material after the constants have been entered and a theta
rotation of 90 degrees has been applied. . . . . . . . . . . . . . . . . . . . . . . . . 32
3.5 View of the defined material after a theta rotation of 90 degrees followed by a psi
rotation of 45 degrees. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33

7.1 Sample geometry of a five layer flat plate system showing the partial waves in each
layer (L+-, SV+-, and SH+-) that combine to produce a guided wave. The SH+-
partial waves are omitted in the case of Lamb waves in an isotropic plate, leaving
just 4 partial waves in each layer. The L+- and SV+- partial waves are omitted in
the case of Love waves in an isotropic plate, leaving just two partial waves in each
layer. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
7.2 Sample geometry of a five layer cylindrical system showing the partial waves in each
layer (L+-, SV+-, and SH+-) that combine to produce a guided wave. . . . . . . . 88
7.3 The structure of the global matrix for a (a) solid, (b) liquid, and (c) vacuum half-
space, where blank spaces are zeros, Dij is the layer matrix, and L+-, SV+-, and
SH+- are the partial waves amplitudes in the various layers. . . . . . . . . . . . . . 119
7.4 The process of finding a root involves a coarse sweep (thick line) to find an initial
minimum and a fine search (dashed lines) to narrow down on a root (solid circle).
The example above shows the absolute value of the determinant of the global matrix
for a 1 mm steel plate immersed in water at a wave number of 4 rad/mm. . . . . . 128
7.5 Using an extrapolating routine to trace modes dramatically improves the program’s
speed and reliability. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129

8.1 (a) Phase velocity and (b) group velocity dispersion curves for a typical Lamb wave
problem, a 1 mm thick steel plate in vacuum . . . . . . . . . . . . . . . . . . . . . 136
8.2 (a) Wave number and (b) angle of incidence dispersion curves for a typical Lamb
wave problem, a 1 mm steel plate in vacuum . . . . . . . . . . . . . . . . . . . . . 137
8.3 Mode shapes for the two fundamental Lamb wave modes overlaid on the phase
velocity dispersion curves. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138
8.4 Phase velocity dispersion curves for the “SH” modes for a 1 mm thick steel plate
in vacuum . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
8.5 (a) Phase velocity and (b) attenuation dispersion curves for a typical leaky Lamb
wave problem, a 1 mm steel plate in water . . . . . . . . . . . . . . . . . . . . . . . 141

7
8.6 The mode shape of a Rayleigh wave propagating on the surface of an aluminum half-
space, shown (a) as a deformed grid and (b) as lines representing the displacement
profiles. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
8.7 The mode shape of a leaky Rayleigh wave propagating on the surface of an aluminum
half-space which is water loaded, shown (a) as a deformed grid and (b) as lines
representing the displacement profiles. . . . . . . . . . . . . . . . . . . . . . . . . . 144
8.8 The (a) in-plane and (b) normal displacement components of a Stoneley wave prop-
agating at the interface of aluminum and steel half-spaces. The vertical axis cor-
responds to the direction perpendicular to the interface and shows a depth of ap-
proximately 5 wavelengths of the Stoneley wave. . . . . . . . . . . . . . . . . . . . 145
8.9 Phase velocity dispersion curves for a 0.1 mm thick epoxy layer on an aluminium
half space. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146
8.10 Phase velocity dispersion curve for the fundamental mode of a 0.1 mm thick epoxy
layer on an aluminium half space. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146
8.11 The grid mode shapes for the fundamental mode of a 0.1 mm thick epoxy layer
on an aluminium half space at (a) 2.0 MHz and (b) 8.0 MHz, showing the change
in behaviour from a mode that is dominated by the properties of the aluminium
properties to one that is dominated by those of the epoxy. . . . . . . . . . . . . . . 147
8.12 The (a) phase velocity and (b) attenuation dispersion curves for a layer of high
performance polyethylene. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 148
8.13 The phase velocity dispersion curves for a aluminum-epoxy-aluminum adhesive joint
when the epoxy is 0.1 mm thick and has a stiffness of 4 GPa and the plates are 1
mm thick. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150
8.14 The (a) phase velocity and (b) attenuation dispersion curves for a aluminum-epoxy-
aluminum adhesive joint in water. The plates are 1 mm thick and the epoxy is 0.1
mm thick. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151
8.15 The phase velocity dispersion curves for the example aluminum-epoxy-aluminum
adhesive joint, when the epoxy has been artificially weakened by (a) lowering the
stiffness to 3 GPa and (b) lowering the stiffness to 1 GPa and reducing the density
by half. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151
8.16 The (a) phase velocity and (b) attenuation dispersion curves for an epoxy layer
embedded in aluminum half-spaces. . . . . . . . . . . . . . . . . . . . . . . . . . . . 153
8.17 The (a) phase velocity and (b) group velocity dispersion curves for modes propa-
gating parallel to the fibre direction in a unidirectional composite laminate. . . . . 155
8.18 (a) Displacement and (b) stress mode shapes for S0 mode parallel to fibres in
unidirectional composite at 1 MHz. . . . . . . . . . . . . . . . . . . . . . . . . . . . 156
8.19 The (a) phase velocity and (b) group velocity dispersion curves for modes propa-
gating at 45 degrees to the fibre direction in a unidirectional composite laminate. . 157
8.20 (a) Displacement and (b) stress mode shapes for S0 mode at 45 degrees to fibres in
unidirectional composite at 1 MHz. . . . . . . . . . . . . . . . . . . . . . . . . . . . 157
8.21 Phase velocity curves for propagation in an 8 ply quasi-isotropic laminate, (a) in
the direction of the fibres in layers 3 and 6 and (b) at 45 to that in (a). . . . . . . 159
8.22 Displacement and stress mode shapes for the (a) S0 mode and (b) A0 mode at 1
MHz in the example 8-ply quasi-isotropic composite plate. . . . . . . . . . . . . . . 160
8.23 (a) Displacement and (b) stress mode shapes for S0 mode propagating at 0 degrees
in an 8-ply 0/90 composite at 1 MHz-mm. . . . . . . . . . . . . . . . . . . . . . . . 161

9.1 Various views of the dispersion curves for a 2 mm diameter steel bar in water: (a)
real wave number, (b) phase velocity, (c) group velocity, and (d) angle of incidence. 165
9.2 Phase velocity dispersion curves for a 2 mm diameter steel steel bar with mode
shapes super-imposed on the (a) fundamental modes and (b) higher order modes. . 167
9.3 Phase velocity and group velocity dispersion curves for an empty 1 mm thick steel
pipe with an inner radius of 2 mm. . . . . . . . . . . . . . . . . . . . . . . . . . . . 168

8
9.4 The effect of increasing the inner radius of a 1 mm thick steel pipe. As the inner
radius increases, the dispersion curves begin to very closely match the dispersion
curves for a plate (shown in section f), except for very low frequencies. . . . . . . . 170
9.5 A comparison of phase velocity and attenuation dispersion curves for three cases:
a fluid filled pipe, a fluid filled pipe with a sink in the centre, and a plate in water. 172
9.6 A comparison of the axial displacement for a fluid filled pipe for a mode that is (a)
predominantly in the fluid and (b) predominantly in the pipe wall. The top of the
graph represents the fluid filled core, the middle section the steel wall of the pipe,
and the bottom section the first 10 mm of the surrounding fluid . . . . . . . . . . . 173
9.7 Phase velocity dispersion curves for a 2 mm diameter steel steel bar immersed in
water with mode shapes super-imposed on the fundamental modes. . . . . . . . . . 174
9.8 As a guided wave travels down a bar, it couples energy into the surrounding fluid,
creating a leaky bulk wave at a characteristic angle. . . . . . . . . . . . . . . . . . 175
9.9 Phase velocity and attenuation dispersion curves for an empty 1 mm thick steel
pipe with a 10 mm inner radius that is surrounded by soft stone (mudstone). . . . 176
9.10 Phase velocity and attenuation dispersion curves for a 1 mm radius steel bar em-
bedded in a stiff material. The arrows and labels correspond to the positions for
which the mode shapes are given in the next figure. The straight lines correspond
to the bulk velocities in the surrounding solid and divide the dispersion curves into
different leaky regimes. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 177
9.11 Mode shapes for a steel bar embedded in a stiff material in (a) a non-leaky region,
(b) a region where only shear waves leak, and (c) a region where both shear and
longitudinal bulk waves leak into the surrounding solid. . . . . . . . . . . . . . . . 177
9.12 Phase velocity and attenuation dispersion curves for a 1 mm thick steel pipe with
a 10 mm inner radius that is lined with visco-elastic grout and is surrounded by
visco-elastic soft stone (mudstone). . . . . . . . . . . . . . . . . . . . . . . . . . . . 178

9
Chapter 1

Introduction

Disperse is an interactive windows program that has been created to generate dispersion curves
for a wide range of systems. It has been designed to enable the user to get a better understanding
of guided wave propagation and allow him or her to develop more efficient non-destructive testing
techniques.

This manual describes how to operate the program and what features it includes. Some more
detailed information on how the dispersion curves are generated is included in the theory section
in Chapter 7 and some examples are explained in Chapters 8 and 9, including step by step
instructions. If you are anxious to get results for a simple case quickly, you may want to refer to
the Quick start pages (2). However, tracing complicated systems will require a significant amount
of guidance from the user, as explained in the section on manual tracing, Section 4.3. Because of
the interaction that is required to trace complicated systems, it is recommended that a new user
practices with simple systems and progresses onto more complicated systems once s/he is used to
Disperse.

Disperse was created by Michael Lowe and Brian Pavlakovic in the Non Destructive Testing
research group in the Department of Mechanical Engineering at Imperial College. It is licensed
by Imperial College Consultants. For technical enquiries or support please contact Michael Lowe
by email at: m.lowe@imperial.ac.uk.

1.1 What Disperse Can Model

Disperse is currently able to handle several different geometries and types of material. These cases
include,

• Cartesian or cylindrical geometries.


• Single or multiple layers.
• Free or leaky systems.

• Elastic or viscoelastic isotropic materials.


• Generally anisotropic materials for non-leaky Cartesian geometries.

Disperse version 2.0.20a


1.2 Limitations 11

Many facilities, both automatic and manual, exist for tracing the dispersion curves and are de-
scribed in Chapter 4.

1.2 Limitations

Disperse solves for modal solutions to the wave propagation problem. These solutions correspond
to the “dispersion curves” of a system. The solution assumes that a continuous wave is propa-
gating in an infinitely long system. Therefore, events that occur over a finite time period or local
phenomena such as defects in a plate cannot be modelled directly.

The following limitations on materials and geometries apply:

• Materials must be isotropic or transversely isotropic if the geometry is cylindrical.


• If an anisotropic structure is defined, then solutions for waves which attenuate by leakage or
material damping may be obtainable, but these have received limited testing. Currently it
is believed that the solutions are working correctly.
• Anisotropic material layers must have a plane of symmetry which is parallel to the plane of
the plate. Thus, when defining the material orientations, not all rotations are permissible.
• The group velocity for waves in anisotropic structures is calculated simply by the usual ω-k
derivative expression; thus the deviation of the direction of the energy velocity with respect
to the phase velocity, which occurs when the propagation is not along a principal axis, is not
calculated.
• If viscous fluids are modelled and they have a very low bulk shear velocity, less than 50 m/s,
then mode shapes are calculated using the approximation that the shear velocity is zero.
• A model capability for surface roughness is present but it is not supported.
• If the structure is embedded or immersed then the power normalisation of the mode shapes
is not strict because the components in the half spaces are neglected in the integration.
• Wave propagation in the circumferential direction of cylindrical structures is limited to
isotropic materials and non-attenuating wave modes. Thus the external boundary must
be free and the materials must be perfectly elastic.

1.3 What Else Disperse Can Do

Once a dispersion curve has been created, Disperse has many ways of examining it in more detail.
The curves can be shown as phase velocity, group velocity, or real wavenumber plots. The program
allows you to zoom in on a section or move around the dispersion curves. Data such as the mode
shapes can be displayed for every traced point. A labelling facility also exists so certain points
can be marked for later study. Some tools are included that help perform certain additional tasks,
such as simulating the dispersion of wave packets as they travel.

These facilities have been designed so that the user will be able to get as much information
from the dispersion curve as possible so that better ultrasonic NDT tests can be developed and
implemented. The authors would appreciate any ideas that you have to better communicate or
use the information held within the dispersion curves.

Disperse version 2.0.20a


1.4 Installing and running Disperse 12

1.4 Installing and running Disperse

Installing on a computer running Windows 95/98/NT/Vista/XP/7: If Disperse is not


already installed on your computer, you will need to take the following steps:

1. If you have a license for Disperse then you will have been supplied with a security key
(“dongle”). If your key was supplied with a previous version of Disperse then your existing
key should still work with the new version. Attach the key to your computer. If your key is
of the parallel port kind, then any other device, such as a printer, can still be plugged into
the key. If you don’t have a license then you will still be able to install and run Disperse but
only in demonstration mode.

2. Load the distribution CD and wait for the auto-install to start. If this does not happen
automatically, then run the program Setup.exe that is located in the root folder of the CD.
The installation will automatically create the folder structure and all necessary files. The
procedure is short and is not complicated. Instructions are also printed in the Read.me file
on the CD. The setup starts with the installation of Disperse. Following that, if you have
purchased a license, then the setup will offer you the options to continue with the installation
of the driver for the security key (dongle) and some additional example files (those listed in
Appendix B of this manual). Please follow the installation in the sequence offered by the
setup program.
3. A .pdf reader is needed in order to read this help manual. If you don’t have a suitable reader
then you will need to obtain and install one.

4. If you have previously installed Disperse, and you are now upgrading to a newer version,
simply install the new files on top of the old ones (use same folder for installation as the
previous one).
5. If you have previously installed Disperse, and you are just adding the driver for the security
key, or upgrading the driver, then you need only run the executable file located in the
dongleInstaller folder inside the Disperse main folder on the CD.
6. Finally, if you have obtained this manual separately from the program, then you should place
the manual in the Help subdirectory of the Disperse installation. If you don’t do this then
the help options when running Disperse will not be able to find the manual.

Installing on multiple computers. Disperse may be installed on any number of computers


but will only be fully active when the security key is in place; without the security key it operates
in demonstration mode in which case it can be used to display previously calculated results or
perform calculations with a very restricted set of materials. The software driver for the key must
also be installed for the key to function.

Example solution files. The installation of Disperse includes the provision of some example
solution files. These contain saved results, which you can open and examine with Disperse. These
are installed in the Examples folder inside the main Disperse folder in the installed location on
the hard disc of your computer. If Disperse has been installed in demonstration mode, then only
a small number of example files will be present. However, in fully licensed mode, a larger number
of example files will be present - those listed in Appendix B of this manual.

Disperse version 2.0.20a


1.5 Display Logic and Mouse Actions 13

Materials database. If you have a license for Disperse then you will be able to create, edit and
save your own database of material properties, see section 3.3. The materials database is stored in
an ASCII file called materials.lst. Prior to version 2.0.20a, this was kept in the Bin subdirectory.
From version 2.0.20a this is kept by default in the directory c:/programdata/dispersedata. However,
you can optionally change the location for storing this file to a directory of your own preference,
see Chapter 10.

Starting Disperse. Disperse can be started by choosing it from the Start menu, double-clicking
on the file disperse.exe in the installation directory or on a desktop shortcut icon, or dragging and
dropping a file onto the appropriate icon. Dragging and dropping a file onto the icon will cause
Disperse to try to load the dropped file.

Please send all bug reports / comments to Michael Lowe, at: m.lowe@imperial.ac.uk

1.5 Display Logic and Mouse Actions

Disperse is designed to be completely compatible with the display and control logic of the popular
windows style applications. It should therefore be very easy for those who are familiar with such
logic to become comfortable with the Disperse user interface.

When starting up, Disperse shows only a limited toolbar at the top of the screen and no windows
other than this main program window. However, once a new geometry (“system”) has been de-
fined, or an existing model file opened, then a window is opened automatically for this model.
Several models can be open concurrently and each will have its own window. In the usual windows
style, any actions relate to the window which is currently selected (eg by clicking on its border).
By default, once dispersion curves have been calculated, the model window shows a graph of the
dispersion curves in frequency-phase velocity space, this being the most popular display prefer-
ence. Other displays, such as frequency-wavenumber, can be selected instead using the Display
pulldown menu.

In addition to the model window,several other windows are used. You will see that other windows,
such as automatic searching and sweeping windows, appear in a minimised form at the bottom of
the main program window. These relate to the separate processes which are called by the main
program as you create you dispersion curves. It is usually not necessary to look in these process
windows but you may want to enlarge them if you need to find out more about the processes, for
example if you are having trouble tracing a mode and you want to see more detail of what the
program is doing. Finally, a separate open window is shown when you choose to display mode
shapes. The modes shapes calculation is again a separate process. The mode shape plots in the
mode shape window relate to the model window and are continuously updated according to the
position of the pointer on the dispersion curves in the model window. If several models are open
concurrently then the mode shape window relates to the currently selected model.

By default the left mouse button is used to point and select in the usual windows manner. However
its function is changed for some of the tasks. For example, if you are using Trace from, then the
left button is used to select the point and automatically start up the trace routine. Similarly it
takes on specific actions you choose from the edit menu or edit toolbar. The left mouse button
always retains the function which you have selected until you select another. To revert to the
default pointer mode, click the right mouse button and select Pointer.

The right button operates a popup menu which offers you some short cuts to popular actions,

Disperse version 2.0.20a


1.6 New Features for Version 2.0 14

the options depending on the current activity. When displaying the dispersion curves, the options
include Zoom, Pointer, Sweep and Search, Trace from, Find minima, Show determinant,
Actions(copy, print, custom zoom), Edit curves, Autoscale. For further details see 5.4

1.6 New Features for Version 2.0

Disperse has been improved extensively in the upgrade from version 1.0 to version 2.0, both in
the calculation procedures and in the capabilities. Users of the previous version should notice
immediately, for example, that the solution speed for version 2.0 is considerably faster. Also
noticeable immediately is the updated Windows style interface with index card dialog boxes and
tree lists. In fact the program has been restructured completely into a modular form in which
the separate activities and features of the program are handled by separate program modules
(DLLs). This greatly improves the flexibility of the program for both maintenance and upgrading.
At the same time, a lot of work has been put into the solution module in order to optimise the
calculation arithmetic for speed, while still retaining the well-proven solution algorithms of the
previous version.

Users who have previously worked with version 1.0 should find the transition to version 2.0 straight-
forward. There have been a few changes to the menu formats, so a few items have moved from
one menu to another, but the general concepts of operation remain the same and it should be
quick to get used to the changes. Indeed we hope that the many improvements and new features
for version 2.0 will be evident very quickly after starting. Some of the major additional features
for version 2.0 are:

• Variable material properties. Material properties can be defined to be a function of frequency,


or indeed a function of some other user-chosen parameter. The function is a polynomial, up
to order 3. See Section 3.3.
• Viscous fluid may be defined. A viscosity constant is specified and this is used to predict
damping losses both of a bulk wave radiating in a fluid, and of shear wave radiation when
a solid is in contact with the fluid. Attenuation of longitudinal waves may also be defined
separately in order to model other loss mechanisms such as scattering. See Section 3.3.1.
• Spring interfaces may be defined between any of the layers, or indeed also at the outer
boundaries of the model. Spring stiffnesses are defined separately in each of the coordinate
directions and may be zero (uncoupled, eg to model free sliding contact), or any user-specified
finite value, or rigid (perfect contact, no spring). See Section 3.3.5.
• New options for external boundary conditions, including rigid boundary, symmetric bound-
ary (so only one half of the thickness of a symmetric plate structure has to be modelled).
See Section 3.2, 3.3.6.
• If a layer structure contains repeated groups of layers, then a super-layer consisting of such
a group may be defined, then used repeatedly in the structure. See Section 3.2.
• Improved definition and solution for anisotropic layers, see Section 3.3.3. The procedure
for defining anisotropic layers has been improved, with clearer and more flexible options
for their definition. Also, anisotropic layers can now be mixed freely with other types of
layers, including fluids. Anisotropic material constants may now also be complex, but see
limitations, Section 1.2.
• Wide choice of solution variables. Conventionally, guided wave solutions are found for vari-
able frequency, real wavenumber and attenuation, while holding constant all other properties

Disperse version 2.0.20a


1.7 Warnings 15

of the system. Disperse version 2.0 allows the solution variables to be chosen in a more flex-
ible manner. Thus, for example, the frequency can be fixed and the modes can be searched
while varying the value of Young’s modulus in one of the layers. See Section 4.4.2.
• Powerful post-solution data extraction tool. After calculating the dispersion curves, it is
now possible to extract detailed information at each point on a mode, perform arithmetic
calculations on it, then plot or export the result. For example, it is possible to extract a
chosen component of stress or displacement at a particular position in the structure. It is
then possible to perform arithmetic operations using one or more such extracted values and
then plot the result. See Section 6.3.
• Simulated signal. A new tool has been added to perform simulations of guided waves as they
propagate. The user defines a wave packet, a position on a dispersion curve and a distance
of propagation. Disperse then predicts the shape of the wave packet after it has travelled
that distance, taking into account the dispersion characteristics of the mode. See Section
6.2.
• Object linking and Embedding (OLE) now supported. For example you can copy a set of
curves and paste into another OLE program such as a word processing program, then later
double click on that object to open Disperse again. Or you can paste something from another
program into Disperse. See Section 6.7.4.

A note concerning the file used to store Disperse results. Disperse version 2.0 creates files using
the file extension .ds2, differing from the Disperse version 1.0 files .dsp. This is to take account
of the improvements to the file format in order to store additional information such as user notes
etc. Disperse version 2.0 can open both the new .ds2 and existing .dsp files, but all files are now
saved as .ds2. The new format .ds2 cannot be opened by version 1.0 of Disperse.

1.7 Warnings

The propagation of guided ultrasonic waves is a large and complex field. There is no way that
one publication or one program can cover this entire field. Some publications that form a good
general background are references [4, 31, 40, 41, 100, 9, 92]. If you have any specific questions
concerning the program or the manual, please contact the authors.

All of the supported cases listed in Section 1.1 have been validated against analytical solutions and
published literature wherever possible. In addition, many of the cases have been validated against
experiment and all of the solutions seem to be logically consistent. However, there may still be
mistakes in the calculation routines. The authors assume no responsibility for the consequences
of using Disperse, regardless of any errors that are present in the results. However, we will make
every effort to correct any deficiencies that are brought to our attention.

Disperse version 2.0.20a


Chapter 2

Getting Started Quickly

The following pages are quick reference guides that you can use to get started in a hurry. It is of
course recommended that you familiarise yourself with the concepts behind Disperse before trying
to run the program. Having a basic understanding of what the dispersion curves should look like
helps dramatically. Further help for the new user can be found in the two examples Chapters (8
and 9) where explanations of the calculation procedure and examples of results are given for a
range of guided wave problems.

Disperse version 2.0.20a


2.1 Quick Start - Flat Isotropic Free Plate 17

2.1 Quick Start - Flat Isotropic Free Plate

To use this quick start sheet, the system you want to examine must fulfill the following criteria.

• The geometry must be FLAT, not cylindrical.


• The materials must all be isotropic.
• The system must be in a vacuum. In other words, no energy can leak into the surrounding medium.

If the system that you want to examine doesn’t fall into these categories, please check other quick start
sheets or the full instruction manual.

In order to trace dispersion curves from scratch, complete the following steps:

1. Start the program.

Windows 95/98/NT/Vista/XP/7: Select Disperse from the Start : Programs menu or double-
click the Disperse.exe file in the directory where Disperse has been installed.

2. Define the System that you want to investigate by


• Selecting File : New (left mouse button to select).
• Making sure that Plate and Structure in vacuum are selected in the menus at the top of
the dialogue box.
• Selecting an isotropic material from the material pull-down menu (to the right of the first
“”).
• Changing the thickness (in mm) by editing the text field to the right of the material name.
• Changing the material constants, if desired, by clicking the “” next to the material and
selecting edit from the popup menu which appears.
• Making sure that “Free boundary” is selected next to the second “”.
• Clicking on the OK button
Refer to the New Geometry section (Section 3.2) for more details.
3. Select Trace : Automatic tracing from the menu, then click on OK. Disperse should search for
the fundamental symmetric and anti-symmetric modes and trace these two modes. Then Disperse
will perform a sweep at a high phase velocity from which it will trace out each mode that it finds.
Pressing the stop button will cause Disperse to abort the automatic tracing (4.2) step that it is
currently executing.
4. Trace any modes that the Automatic Routine missed. If the search for minimum points drew circles
where modes were not traced, you will need to fill in these missing modes. This can be done by:
• Selecting the Trace from point button from the toolbar (the one with the black dot and the
long red arrow). This can also be found on the popup menu which appears if you click on the
right mouse button. Note - if you switched to have a look at your results in one of the other
display forms, then it is best to return to the phase velocity-frequency display before resuming
the search.
• Clicking with the left mouse button on the circles that are not on modes, or on modes that
have stopped prematurely. The direction and extent of tracing from a point can be altered by
changing the target frequency in the text box on the toolbar.
• When finished, click the right mouse button and select Pointer, to return the mouse to its
default mode - otherwise it will continue to attempt to trace each time the left button is
pressed.
5. Record your results.

Disperse version 2.0.20a


2.2 Quick Start - Flat Isotropic Plate in Water 18

• File : Save allows you to save your data (both the material and the dispersion curves). A
.ds2 extension is recommended.
• The plot can be printed by selecting File : Print, or exported in various forms including
postscript or text (which can then be read into a plotting program or spreadsheet) with Tools
: Export Data

2.2 Quick Start - Flat Isotropic Plate in Water

To use this quick start sheet, the system you want to examine must be Cartesian and Isotropic.

In order to trace dispersion curves from scratch, complete the following steps:

1. Start the program.


Windows 95/98/NT/Vista/XP/7: Select Disperse from the Start : Programs menu or double-
click the Disperse.exe file in the directory where Disperse has been installed.
2. Define the System that you want to investigate by
• Selecting File : New (left mouse button to select)
• Making sure that Plate and Structure is embedded are selected in the menus at the top
of the dialogue box.
• Selecting water from the material pull-down menu (to the right of the first “”).
• Selecting an isotropic material to the right of the second “”).
• Selecting water to the right of the third “”).
• Changing the thickness (in mm) of the middle layer by editing the text field to the right of
the material name. The first and third layers are the embedding medium and are thus semi-
infinite; their thicknesses will appear here as default values of 1 (mm) but will be ignored by
the program.
• Changing the material constants of any of the materials, if desired, by clicking the “” next
to the material and selecting edit from the popup menu which appears.
• Clicking on the OK button
Refer to the New Geometry section (Section 3.2) for more details.
3. Select Trace : Automatic tracing from the menu, then click on OK. Disperse should search for
the fundamental symmetric and anti-symmetric modes and trace these two modes. Then Disperse
will perform a sweep at a high phase velocity from which it will trace out each mode that it finds.
Pressing the stop button will cause Disperse to abort the automatic tracing (4.2) step that it is
currently executing.
4. Trace any modes that the Automatic Routine missed. If the search for minimum points drew circles
where modes were not traced, you will need to fill in these missing modes. This can be done by:

• Selecting the Trace from point button from the toolbar (the one with the black dot and the
long red arrow). This can also be found on the popup menu which appears if you click on the
right mouse button. Note - if you switched to have a look at your results in one of the other
display forms, then it is best to return to the phase velocity-frequency display before resuming
the search.
• Clicking with the left mouse button on the circles that are not on modes, or on modes that
have stopped prematurely. The direction and extent of tracing from a point can be altered by
changing the target frequency in the text box on the toolbar.

Disperse version 2.0.20a


2.2 Quick Start - Flat Isotropic Plate in Water 19

• If you think there may still be some missing modes, then you can search further in specific
areas. Click the right mouse button and select Sweep and Search from the popup menu.
Use the left button to drag and outline a rectangular box area to search. The program will
search along a line from bottom left to top right of the box, and will draw circles anywhere it
finds roots. Note that if any of these lie on velocities corresponding to any of the bulk velocities
of the materials, then they are not roots of the guided modes and should be disregarded (See
7.5).
• When finished, click the right mouse button and select Pointer, to return the mouse to its
default mode - otherwise it will continue to attempt to trace each time the left button is
pressed.

5. Record your results.


• File : Save allows you to save your data (both the material and the dispersion curves). A
.ds2 extension is recommended.
• The plot can be printed by selecting File : Print, or exported in various forms including
postscript or text (which can then be read into a plotting program or spreadsheet) with Tools
: Export Data

Disperse version 2.0.20a


Chapter 3

Defining the Structure

Before generating the dispersion curves it is necessary to define the structure. This involves the
choice of a flat plate geometry or a cylindrical geometry, the definition of the material properties
of all of the layers of the system, and the specification of the boundary conditions. This chapter
describes how to define these properties; the next chapter will show how to calculate the curves
once the geometry has been defined.

3.1 Getting Started

The information about the geometry of the system which is to be modelled can be entered in by
the user or it may be read in from a file. In both cases, the type of geometry, the number of layers,
the material properties of each of the layers, and some size parameters are required.

To input this data, you can either define a new geometry (File : New), edit the current geometry
(Trace : Edit Geometry), or load a file from disk (File : Open). If a set of dispersion curves
is already active, Trace : Edit Geometry will replace the current geometry and erase all of the
current dispersion curves. However, File : New (and File : Open) will open new dispersion
curve windows and keep the old data. Thus it is possible to have several windows open, each with
a different set of solutions. This can be useful for example for comparing the results for different
systems.

Choosing File : New or Trace : Edit Geometry will prompt you for all of the material
information that Disperse will need. This information consists of two parts, general information
about the geometry and specific definition of the properties of each layer.

3.2 Defining the Geometry

Choose Plate, Cylinder or Circumferential First you will need to choose a geometry from
the dropdown menu at the top left of the dialogue box. Plate defines a flat plate structure.
Cylinder defines a cylindrical structure with wave propagation along the axis (the z direction).
Circumferential defines a cylindrical structure with wave propagation around the circumference
(Θ direction). Please see Section 1.2 for limitations on these geometries.

Disperse version 2.0.20a


3.2 Defining the Geometry 21

Choose the external boundary conditions Then you need to define, using the dropdown
menu at the top right of the box, what is at the external boundaries of your structure. The struc-
ture may be unbounded (Structure is in vacuum) or it may be immersed in a fluid or embedded
in a solid (Structure is embedded). If Structure is in vacuum is selected, Disperse will au-
tomatically add vacuum half-spaces to the system when the new system is accepted. Technically,
this implies a definition of traction-free surfaces. If Structure is embedded is selected, then
Disperse assumes that the first and last materials which you will define in your list of layers are
semi-infinite half spaces. Each half space may be defined to be liquid or solid materials, or indeed
vacuum.

If you have chosen to model a cylindrical structure with Structure is in vacuum, then an
additional box is opened for you to define an inside cylindrical radius. For example, for an empty
hollow pipe which also has nothing on its outside, you would specify here the inside radius of the
pipe; the subsequent layer information would then define the pipe wall material and thickness.
If the structure is not hollow and empty, then the first layer is not vacuum and so you should
define Structure is embedded. In this case the first layer will define the inner material of the
structure and its radius. Thus a fluid-filled pipe would be modelled by specifying Structure is
embedded then defining the fluid by the first layer in the list with thickness equal to the inner
radius of the pipe, followed by a layer to define the pipe wall. A solid structure, such as a solid
circular bar, can be specified in two ways: (1) Choose Structure is in vacuum, then set zero
inside radius, followed by a layer with thickness equal to the bar radius; (2) Choose Structure is
embedded, then define the first layer with the material and radius of the bar.

Select the materials Following these items in the dialogue box is a series of dropdown menus
offering pre-defined material properties. In the following section we will see how to edit material
properties or define new materials; for the purposes of this section we assume that the pre-defined
materials are sufficient. There is a dropdown menu for each layer, listed in the order of the layers
from the “top” to the “bottom” of a plate structure, or the inside to the outside of a cylindrical
structure. Initially three layers are offered but layers can be added or deleted by clicking on the
“” button to the left of any of the layers and choosing Insert before, Insert after, or Delete.
The first layer always represents the “top” semi-infinite half-space above the plate or the inside
core of the pipe, according to the choice of geometry. Therefore if Structure is in vacuum
has been chosen then the first layer will always be pre-set to Free boundary. Similarly the
last layer always represents the “bottom” semi-infinite half-space below the plate or the infinite
space outside the pipe. The intermediate layers define the finite-thickness layers of the structure.
For each layer it is required to choose a material and specify a thickness. The thickness of the
semi-infinite layers is not used by the program except in the case explained above where the first
layer defines the radius of the core of a cylindrical structure.

Take advantage of repetition in the layer structure There is additionally a facility to


create groups of repeated layer structures. This is convenient when a multilayered structure is
made up of repeated identical groups of layers, as is commonly the case with carbon fibre composite
structures. The use of the Repeated groups facility allows the systems to be defined quickly
and reliably, and also reduces the computation time significantly because Disperse exploits the
repeating patterns when assembling the coefficients for the global matrix. To activate this facility,
click on the “” button to the left of the first layer of the group, then select Repeated groups
from the dropdown menu. The chosen layer will then become the Group header and the indented
layers which appear below it can be used to define the member layers of the group in the usual
manner. The group header may be given a title and the number of times which it repeats must be
entered in the box to the right. The whole layer structure is presented in a windows tree fashion,
so that groups can be viewed in expanded or compacted form by clicking on the + or - button to

Disperse version 2.0.20a


3.2 Defining the Geometry 22

the left of the group name.

Take advantage of symmetry in the layer structure The pre-defined dropdown lists for
first and last layers additionally offer two possibilities which are not available for the intermediate
layers (and indeed there will be further possibilities for defining external boundary conditions
when it comes to editing/creating the layer data). The first option is Free boundary, which
simply specifies the traction-free condition and is therefore identical to a vacuum half-space. The
second option is Symmetric System. This may be used for plate systems (only) which are
geometrically symmetric. If this is selected then the program assumes that the structure defined
from the top down to that layer is repeated in mirror image below that layer. This option is
strongly recommended if the structure has more than one finite thickness layer and is geometrically
symmetric: it reduces the computation which is needed, and it allows the program to search
separately for the symmetric and antisymmetric modes, thereby increasing tracing stability.

Thickness normalisation Sometimes, particularly for parametric studies, it is useful to scale


the thicknesses of the layers of the structure such that the total thickness is a constant. The
Norm. Thick. button scales the thickness of all of the finite layers (neglecting semi-infinite half
spaces) such that the total thickness of the structure is 1 mm. If this button is pressed then the
thickness boxes on the right hand side of the layers definitions are updated to the new scaled
values immediately.

List of pre-defined materials The Edit list button allows you to see the full list of pre-
defined materials and to delete any which are no longer needed. Creating new entries in the list
or modifying existing entries will be explained in the next section.

Special consideration for cylindrical geometries Whenever the material inside a pipe is a
fluid, Disperse may be set to assume that there is a sink at the axis. The presence of a sink means
that bulk waves which leave the inside wall of the pipe and travel towards the centre of the fluid
column are completely absorbed when they arrive at the axis. Without a sink the waves which
arrive at the axis continue to travel across the fluid column and thereby arrive at the opposite wall
of the pipe. Therefore in the absence of a sink the solution will yield ’fluid modes’ corresponding
to resonances of the fluid inside the pipe. These are often very closely spaced compared to the pipe
modes. The sink can be turned on or off in the Trace : Solution type menu. The consequences
of this choice are discussed further in Section 9.2.1.

Several limitations are posed by the nature of the Bessel functions that are used to evaluate the
system matrix for cylindrical systems. The Bessel function routines become unstable for large
arguments, which occur for large radii (in comparison to the thickness), high wavenumber (low
phase velocity), and very slow wave speeds. This instability causes Disperse to stop converging
on roots along a diagonal line. When this problem occurs, you should try modeling the system as
a plate instead of a cylinder. The solutions should remain the same because when the arguments
of the Bessel functions become large enough to cause this instability, the Bessel functions are
behaving as damped versions of their Cartesian counterparts and provide very similar results. A
general rule of thumb derived from experience of tracing curves and explained in Section 9.1.6
indicates that a plate case can usually be substituted if the minimum frequency of interest is
greater than the value given by the equation,
vlong
fmin = (3.1)
rinner

Disperse version 2.0.20a


3.3 Defining a Layer 23

where fmin is the minimum frequency in MHz, rinner is the inner radius in mm, vlong is the bulk
longitudinal velocity in mm/µs. Therefore, if the target application will only use frequencies–wall
thicknesses above 0.25 mm-MHz, a plate case can only be substituted if the ratio of radius to the
wall thickness is greater than about 20.

The leaky cylindrical case has been tested against some experiments but there is little data in
the literature for non-axisymmetric modes leaking into solid materials, so use these results with
caution.

3.2.1 Some Simple Examples

Here are some examples of how to define the geometry of some simple structures, using the pre-
defined materials:

Lamb waves in a steel plate in vacuum: select Plate, Structure is in vacuum. The first
layer will be pre-set to free boundary and the third layer will show free boundary as the
default. All you need to do is to choose steel as the second layer. See the example file
plate-steel.ds2.
Leaky Lamb waves in a steel plate in water: Select Plate, Structure is embedded and
define three layers: (1) water (this is semi-infinite), (2) steel, and (3) water (again semi-
infinite). See the example file plate-steel-in-water.ds2.

A Rayleigh wave on a steel half-space: Select Plate, Structure is embedded. Define the
first layer as a free boundary, then either (1) define the second layer as steel and delete the
third layer, or (2) define the second and third layers as steel (this would allow you to see
mode shapes in the finite thickness layer of steel without having to request the option for
mode shapes in the half spaces). See the example file ray-alum.ds2.

Waves travelling along a water-filled pipe , the pipe being also immersed in water:] Select
Cylinder, Structure is embedded. Define three layers: (1) water with thickness equal to
the inner radius of the pipe, (2) the pipe material together with its wall thickness, (3) water
(this will be infinite). See the example file cyl-h20f10-stl-h20.ds2.
Waves travelling along a solid bar in vacuum: Select Cylinder, Structure is in vacuum.
Set the inside radius to zero, then specify the first finite layer with the bar material and its
radius, then finish with free boundary. See the example file bar-steel-2.ds2.
Waves in an adhesive joint of two identical plates , the whole joint being in vacuum: se-
lect Plate, Structure is in vacuum. The first layer will be pre-set to free boundary.
Define the second layer with the properties of the plate and the third layer with the prop-
erties of the adhesive, but specify a thickness which is half of the actual thickness of the
adhesive layer. Choose Symmetric System for the fourth and final layer. See the example
file plate-adhesive-joint.ds2.

3.3 Defining a Layer

The following sections allow you to examine or edit the pre-defined material properties or define
new ones. The material properties are stored in the material.lst file - this is a text file and may be
edited externally (see further information in Chapter 10) as well as through the Disperse options

Disperse version 2.0.20a


3.3 Defining a Layer 24

which follow. Additional materials can be appended to the list and the constants can be changed
whenever a material is being edited.

New Material If the material that you would like to model is not listed, there are two similar
ways to define a new material. One method is to choose an existing material from a layer dropdown
menu that is similar to the material that you want to model, then edit it, then save it as the new
material. Alternatively, you can select the material called Undefined from the dropdown menu,
which then opens the procedures to create a new material from scratch. Both approaches use
Edit, followed by the same procedures.

Editing a Material To edit a material, click on the “” button to the left of the chosen layer
and select Edit from the pulldown menu. A card index Material definition menu will appear.
This has separate cards to define different types of materials. There are some common options to
all of these, as follows:

Name. Enter the full name of the material.

Abbrev.. Enter an abbreviated name which will be used by the program when space is limited.
Add to List. Click on this if you want to add your material to the list of pre-defined materials.

The procedure for material definition has been designed to give a wide choice of input parameters
as well showing how they relate. Thus, for example, an isotropic solid can be defined by its bulk
velocities or its moduli or by the mixture of its shear velocity and Poisson’s ratio. Immediately it
has been defined by any one of these the equivalent values converted to the other two definitions can
be viewed. Furthermore, clicking on the tab for anisotropic materials will show these constants
in place in any of the chosen anisotropic matrix forms. Internally the program uses the bulk
velocities for the calculations; accordingly references to the material properties normally specify
them in this form by default. Detailed options on the tabs are as follows:

3.3.1 Liquid

Fluids are treated internally as isotropic solids. For a lossless fluid only the longitudinal bulk wave
is modelled; the shear bulk wave is omitted. The required properties are therefore just the bulk
compression wave speed and the density.

For a viscous fluid both longitudinal and shear waves are modelled so that the isotropic solid
represents the properties of a Stokes fluid with Newtonian viscosity. The single viscosity constant
which is given by the user (it can be expressed either as the dynamic or the kinematic constant)
is then used internally to define longitudinal and shear velocities and longitudinal and shear
attenuations. You may optionally choose to neglect the damping which the viscosity causes in
the longitudinal wave by removing the check Velocity affects long. atten.. In that case the
viscosity will only cause attenuation in the bulk shear wave; the bulk longitudinal wave will be
perfectly lossless. If you want to specify longitudinal attenuation which is additional to any given
by the viscosity then simply enter this in the Long. atten. data field. The definition of the
longitudinal attenuation is as explained for Hysteretic structural damping in the context of
isotropic solids in the next section. Theoretical details of the treatment for viscous fluids are given
in 7.1.2.

Disperse version 2.0.20a


3.3 Defining a Layer 25

Values are constant sets the parameters to be fixed values for the whole solution. Value
variation is automatic recognizes that the parameter is not constant for the whole solution but
allows Disperse to set it automatically according to other criteria - for example Disperse needs to
set this internally according to a frequency-dependent function when it is modeling a viscous fluid.
If Values depend on frequency is chosen then buttons appear to enable the density, velocity
and attenuation properties to be defined as polynomial functions of the frequency up to order 3.
When the coefficients of a polynomial have been entered the polynomial is displayed graphically.
The graph can be copied to the clipboard for exporting by using the functions on the right mouse
button, as described in 6.7.4. Note that the effect of viscosity is itself frequency dependent and
this is dealt with internally using the viscosity constant. The option Values depend on user
variable allows the user to define the properties to be functions of another variable, other than
frequency. This is explained in 4.4.2.

Fluid layers are usually encountered as semi-infinite half spaces in which structures are immersed;
however it is possible to include a fluid as an internal layer, but note restrictions in 1.2.

It is also worth remembering that the user may choose explicitly to define a fluid using the solid
properties tab instead of the fluids tab. If the shear wave speed is set to be very low then the
solutions for a lossless fluid defined in this way are very similar to those obtained using the formal
fluids definition (for example use a shear wave speed of 10 m/sec), although the searching can be
less stable. This could be useful if it is necessary to get around the restrictions for fluids as internal
layers. Another reason for defining a fluid using the solids tab could be to define a complex lossy
fluid for which the viscous model is not adequate.

3.3.2 Isotropic

Isotropic solids are defined by the density and either the longitudinal and shear wave speeds or
the Young’s modulus and Poisson’s ratio. If there is No material damping then then this is all
that needs to be defined.

Attenuation caused by material damping may be specified in several ways. In each case the
damping values correspond to the exponential decay of plane waves in an infinite medium and
they are defined independently for the longitudinal and shear waves. the options are as follows:

• Hysteretic structural damping defines a damping loss which is a constant per wavelength
travelled. It may be expressed in units of either Nepers per wavelength or dB per wavelength.
A neper is an exponential decay constant, such that in the following equation,
A(location2) = A(location1) e−αn
the amplitude of the wave, A, decreases by the factor e−αn if α is in nepers per wavelength,
and location 2 is n wavelengths (λ) downstream from location 1. For this type of damping it
follows that the loss per unit distance travelled increases as the frequency increases and the
wavelength diminishes. This is the default description which fits directly with the derivations
presented in Chapter 7.
• Hysteretic damping (per meter) for convenience enables the bulk attenuation to be
defined as a loss per unit distance at a given frequency. The program then converts this to
the appropriate loss per wavelength. From here on the model is identical to that described
in the previous item.
• Kelvin-Voigt viscous damping defines the dashpot damping model, that is the damping
force is proportional to the particle velocity. This model results in a bulk wave attenuation

Disperse version 2.0.20a


3.3 Defining a Layer 26

which is a stronger function of frequency than the Hysteretic damping. The loss per wave-
length is now no longer constant but increases linearly with frequency. Accordingly if the
loss is expressed per unit distance then it increases with the square of the frequency. To
define the attenuation you must specify a loss per unit distance at a particular frequency.
Disperse then automatically scales this appropriately to the frequency every time it makes
a calculation.

When an attenuation mechanism is defined then a data box for a value of surface roughness (this
is for both surfaces) is also given. This is a specialist option and should normally be left as
zero. The algorithm for the roughness calculation is from [48, 49] and it uses a complex value of
the layer thickness to yield an approximate solution for the losses due to scattering at the rough
surfaces. This approximation is known to be valid only for certain ranges of modes and positions
on the curves, as discussed in the papers. It should not be used without first carefully studying the
references. The complex thickness has been implemented in Disperse such that it functions for flat
plates of one or more layers, although only the single layer case has been tested (by comparison
with the solutions in the above papers). It does not function for cylindrical geometries.

Values do not vary with frequency sets the parameters to be fixed values for the whole
solution. Values change automatically with frequency is pre-selected in this box by the
program if properties have been set such that it has to recalculate the material constants at each
frequency rather than simply look up the constants. This is the case for example if Kelvin-Voigt
viscous damping has been selected. If Values depend on frequency is chosen then buttons
appear to enable the density, velocity and attenuation properties to be defined as polynomial
functions of the frequency up to order 3. The option Values depend on user variable allows
the user to define the properties to be functions of another variable, other than frequency. This
is explained in 4.4.2. When the coefficients of a polynomial have been entered the polynomial
is displayed graphically. The graph can be copied to the clipboard for exporting by using the
functions on the right mouse button, as described in Section 5.4.2.

3.3.3 Anisotropic

Disperse allows the specification of anisotropic materials in a number of ways, and can also carry
out transformations from one axis system to another. Thus the user may define the anisotropic
material in a convenient axis system and then transform it to lie in a chosen orientation as a layer
in the structure. Disperse also converts material properties, as far as is possible, from one type
to another. Thus, for example, if a material is defined to be transversely isotropic, and then the
option for orthotropic is chosen, then Disperse automatically populates the orthotropic matrix
with the equivalent coefficients. Equivalence relationships between different types of anisotropic
materials are given in Appendix A.

The new user is advised to go through this present section together with the examples for
anisotropic layers which are given in Section 8.7. In outline, the procedure for defining an
anisotropic layer is:

1. Choose a type of anisotropy (transversely isotropic, orthotropic etc) from the pull-down
menu at the top left of the definition card.
2. Optionally choose a default material from the list of stored materials to populate the data
boxes with an initial set of constants. Important - see note later about how these are stored
and recalled.

Disperse version 2.0.20a


3.3 Defining a Layer 27

Figure 3.1: A Bond transformation matrix is used to rotate an anisotropic material. The orien-
tation is specified by a rotation about the x̂ axis, followed by a rotation about the ŷ axis and a
second rotation about the x̂ axis. The angles are specified in degrees.

3. Enter the anisotropic material constants as required (or edit existing ones if starting from a
default material), assuming the material to be lying in a convenient orientation. For example
a transversely isotropic material would be defined most conveniently when the fibers align
with one of the global axes.
4. Enter the transformation angles so as to rotate the material to the orientation in which you
would like it in the layer.
5. Complete other information, such as density, as required.

Detailed information and an example relating to these steps follows. Theory relating to the
propagation of waves in anisotropic plates is covered elsewhere in the manual, in Section 7.4.1.

The axis system which is used by Disperse for its calculations and the procedure for the transfor-
mations is illustrated in Figure 3.1.

Disperse uses the z axis as the axis of propagation of the waves. This choice was made so that
the same axis could be used for propagation in both flat plate and cylindrical analyses, since z is
the conventional cylindrical axis direction (note this differs from the convention which was used
by Disperse version 1). The x axis is then the direction normal to the layer; in a flat plate this is
the normal to the plate, in a cylinder this is the radial direction, also called the R direction. The
Y axis is then the direction normal to the XZ plane; in a flat plate this is in the plane of the layer
and normal to the direction of propagation, in a cylinder this is the circumferential direction, also
called the θ direction.

The stiffness of an anisotropic material is often defined using a Cij matrix. For further details on
anisotropic properties see [4, 63] and the summary of relations in Appendix A. The Cij matrix
relates the stress vector σ to the strain vector  according to the following completely general
relationship:

Disperse version 2.0.20a


3.3 Defining a Layer 28

    
σxx C11 C12 C13 C14 C15 C16 xx

 σyy  
  C21 C22 C23 C24 C25 C26 
 yy 

 σzz  
= C31 C32 C33 C34 C35 C36  zz 
   (3.2)

 σyz  
  C41 C42 C43 C44 C45 C46 
 yz 

 σzx   C51 C52 C53 C54 C55 C56  zx 
σxy C61 C62 C63 C64 C65 C66 xy

The Cij matrix is thus always symmetric and there is a maximum of 21 independent constants.
Clearly in particular cases such as transversely isotropic, orthotropic etc, only the limited set of
independent constants need to be defined, as requested in the data boxes. The other constants
are either zero or are defined internally by the program using the given constants.

An alternative way of expressing the stiffness information is to use the Engineering constants
which in fact define the properties according to a flexibility matrix. The coefficients are composed
of E (Young’s modulus), G (shear modulus) and ν (Poisson’s ratio). The matrix of engineering
constants for an orthotropic material (and therefore also for a transversely isotropic material
because this is a special case of an orthotropic material) is:

1 −ν21 −ν31
 
0 0 0
  
xx E11 E22 E33 σxx
−ν12 1 −ν32
 yy   E11 E22 E33 0 0 0 
σyy 
   −ν13 −ν23 1
 
 zz  
= E11 E22 E33 0 0 0 
σzz 
(3.3)
  
1
yz   0 0 0 0 0 σyz
  
   G23  
zx 1 σzx
  0 0 0 0 0
  
 G13 
xy 1 σxy
0 0 0 0 0 G12

Again this matrix must be symmetric; therefore although 12 constants are shown here there are
only 9 independent constants once the off-diagonal symmetric terms are equated.

The type of anisotropy can be chosen from the following:

Transversely Isotropic (Engineering constants, along X) . 5 constants. Hexagonal sym-


metry class. This assumes a material whose axis of isotropy is defined to be in the X
direction, i.e. the direction normal to the layer. For example, if this material is used with-
out any subsequent rotation, then it could represent a unidirectional fiber material with
the fibers aligned in the direction normal to the layer. Alternatively, the material could be
entered in this orientation and then rotated to another axis system for the analysis; this
will be illustrated in the example which follows. The key point to remember is that the
propagation direction of the wave modes will always be the Z direction. The Engineering
constants which are required consist of three moduli and two Poisson’s ratios.
Transversely Isotropic (Engineering constants, along Z) . Needs 5 constants. This is
identical to the previous item except that it assumes a material whose axis of isotropy is
defined to be in the Z direction. Thus, for example, it can be used to model a unidirectional
fiber material with the fibers aligned in the direction of wave propagation, without the need
for any rotation. The Engineering constants which are required consist of three moduli and
two Poisson’s ratios.
Transversely Isotropic (Cij’s, along X) . Same as first item but defined by Cij coefficients
instead of engineering constants. The relationship between the Cij coefficients and the
engineering constants is given in Appendix A.

Disperse version 2.0.20a


3.3 Defining a Layer 29

Transversely Isotropic (Cij’s, along Z) . Same as second item but defined by Cij coefficients
instead of engineering constants.
Transversely Isotropic (Cij’s, lossy) . Needs 5 complex constants. The real parts describe
the stiffness in the usual way. The imaginary parts describe the material damping. The
axis of isotropy is defined to be in the Z direction. The modeling of damping in this way
in anisotropic materials is discussed, for example, in [24]. Note, however, that Disperse
needs the opposite sign of each imaginary coefficient to that given in [24]. This is because
of the difference in the convention which is used in the direction of the wave equation.
Thus the imaginary constants which are entered here should normally be negative. The
relations between the imaginary constants, the wave attenuation and the equivalent isotropic
attenuation constants are given in Appendix A. The outcome of this definition of damping
is that the attenuation of bulk waves is a constant value per wavelength, regardless of the
frequency (hysteretic). Other damping models may be defined by entering the imaginary
constants as functions of the frequency - see Section 3.3.4.
Transversely Isotropic (components) . Calculates 5 constants from constituent material
components. This is a convenient way to define a fibrous material by defining separately
each of the two constituent materials and then stating their relative proportions. The fiber
material and the matrix material are each assumed to be isotropic; the anisotropy only en-
ters because of the lay-up. The Fiber fraction is the volume fraction of the fibre material
- or the area fraction in a cross section normal to the fibers. The orientation angles define
the alignment of the fibers using the usual Bond transformations; if all three angles are zero
then the fibers are aligned along the Z axis.
Orthotropic (Engineering constants) . Needs 9 constants. Orthorhombic symmetry class.
The Engineering constants which are required consist of six moduli and three Poisson’s ratios.
The stiffness equation is the full arbitrary axis expression with engineering coefficients given
in equation 3.3. The relationship between the engineering constants and the Cij constants
is given in Appendix A.
Orthotropic (Cij’s, elastic) . Same as previous item but defined by Cij coefficients instead of
engineering constants. The relationship between the 9 Cij constants and the engineering
constants is given in Appendix A.
Orthotropic (Cij’s, lossy) . Needs 9 complex constants. The real parts describe the stiffness
in the usual way. The imaginary parts describe the material damping. The modeling of
damping in this way in anisotropic materials is discussed, for example, in [24]. Note, however,
that Disperse needs the opposite sign of each imaginary coefficient to that given in [24].
This is because of the difference in the convention which is used in the direction of the
wave equation. Thus the imaginary constants which are entered here should normally be
negative. The relations between the imaginary constants, the wave attenuation and the
equivalent isotropic attenuation constants are given in Appendix A. The outcome of this
definition of damping is that the attenuation of bulk waves is a constant value per wavelength,
regardless of the frequency (hysteretic). Other damping models may be defined by entering
the imaginary constants as functions of the frequency - see Section 3.3.4.
Orthotropic (Cij’s, compliances) . Needs 9 constants. The matrix of compliance coefficients
is simply the inverse of the Cij matrix.
Trigonal (32,3m, Cij’s, along Z) . Needs 6 constants. This material satisfies the 32 and 3m
symmetry classes as defined in [4], p.198. The principal axis is along the propagation direc-
tion Z.
Cubic (Cij’s) . Needs 3 constants. This material has C11 = C22 = C33 , C12 = C13 = C23 , and
C44 = C55 = C66 , so the constants are the same for orientation in any of the three principal
axes (XYZ).

Disperse version 2.0.20a


3.3 Defining a Layer 30

Isotropic (Cij’s) . Needs 2 constants. This is included in order that isotropic materials may be
defined using Cij matrix coefficients. Orientation is clearly arbitrary.

Once you have defined an anisotropic material, you may examine the slowness of the surface of
the defined material by pressing the appropriate button. This option is helpful for checking if the
values for the constants are reasonable. The slowness graph can be copied to the clipboard for
exporting by using the functions on the right mouse button, as described in Section 6.7.4.

The material properties are entered in whatever convenient form they are available, according
to the choices listed above, then it is rotated so that it is aligned appropriately for the analysis,
bearing in mind that the wave propagation direction is always the Z direction. The orientation of
the material when it is placed as a layer in the structure is thus defined by three angles. These
three angles are used to generate a Bond transformation matrix (See [4] p.82). The three rotation
angles are defined as follows:

• Phi rotates around the X (or 1) direction, that is the axis which is perpendicular to the
plane of the layer, i.e. around the axis that goes through the thickness of the layer. The
direction of rotation follows the right hand screw rule.
• Theta rotates around the Y (or 2) axis which is parallel to the plane of the layer and
perpendicular to the direction of propagation. Again the direction of rotation follows the
right hand screw rule.
• Psi specifies another rotation about the axis perpendicular to the plane of the layer (X (or
1). Again the direction of rotation follows the right hand screw rule.

Using these three rotation angles allows any orientation to be specified. For example, if you
defined a transversely isotropic material with the axis of isotropy (fibers) in the (Z) direction, a
rotation of 0,0,0 gives you propagation along the fibers, 0,0,90 across the fibers, and 0,90,0 in the
plane of isotropy (such that the cross-sections of the fibers at the interface between the layers
would be circles). A rotation of 90,90,-90 simply spins material around the axis of propagation (Z)
and would have no effect on a transversely isotropic material whose fibers are in the X direction.
However, note that not all rotations may be permissible, see Section 1.2.

Note that when you save a material in the materials file it is saved in the native orientation in
which it was defined. This is important to remember when selecting the material later as the
default material. The following example illustrates the point. You define a transversely isotropic
material with its axis of isotropy along Z, then save it. Later you want to define a new material and
you select the type: Transversely isotropic (Engr along X) and you take the saved material as the
default. Disperse automatically converts from one material type to another as well as is possible
so it converts the constants of the saved material to the new desired input format retaining its
original direction; it does NOT rotate the saved material to put its axis of isotropy along X.

Example of defining an anisotropic layer.

We consider a simple unidirectional carbon/epoxy composite laminate sheet 1mm thick. The
material is designated T300/914, consisting of 32 parallel plies of T300 carbon fibre in a 914
epoxy organic matrix. It can thus be considered to be a single layer. Propagation in this material
was studied by Simon ([95]).

The task is to enter the appropriate parameters defining the material, then to rotate it to the
desired orientation such that the waves will propagate parallel to the fibres. As listed above, the

Disperse version 2.0.20a


3.3 Defining a Layer 31

Figure 3.2: A fibrous material is typically defined in the literature with the X axis parallel to the
fibres.

material data can be entered either as engineering constants (where Eij are the elastic moduli,
Gij are the shear moduli and νij are the Poisson’s ratios), or as elastic constants (Cij ), together
with the density (ρ). Clearly the subscripts i and j must be explicitly defined with respect to the
material reference co-ordinates. In much of the literature the reference co-ordinate system which
is used to define materials is such that the axis of symmetry is aligned with the X, or 1 direction,
as shown in Figure 3.2.

The material in this example has indeed been defined in this way. Thus its fibres are aligned in
the X direction, and the material constants, in that native coordinate system are:

C11 = 143.8GP a C12 = C13 = 6.2GP a C22 = C33 = 13.3GP a


C23 = 6.5GP a C44 = 3.6GP a C55 = C66 = 5.7GP a ρ = 1560kg/m3 (3.4)

Clearly the best way for us to put this material into our Disperse model is to enter the material
constants in the local coordinate system and then to rotate the material to the desired orientation
for the analysis.

Select File : New to get the Geometry Definition form, then since this is a single layer we can
accept the default system given which shows an undefined layer with free boundaries above and
below. To define the material for this layer, left click the “” button on the left of the undefined
layer and select “Edit” from the popup menu. This then displays the Material Definition form,
from which we select the “Anisotropic” tab. Type in an appropriate name (“Graphite-epoxy-
T300-194”) and abbreviation (“T300-194”) in the edit boxes at the top of the form. Now we
need to select the appropriate material and parameter type from the left hand list-edit box. This
material is transversely isotropic, that is to say it is isotropic in the plane which is perpendicular
to the fibres. We may therefore use one of the transversely isotropic options. The material has
been provided with properties defined by Cij values with the coordinate system such that the fibre
axis is parallel to X. Therefore the clear choice for input is “Trans.Isotropic (Cij’s - along x)”. It
is important to understand at this stage that the “along x” refers to the axis of symmetry of the

Disperse version 2.0.20a


3.3 Defining a Layer 32

Figure 3.3: View of how Disperse sees the example material after the constants have been entered
using “Trans.Isotropic (Cij’s - along x)” but before any rotation angles have been specified.

Figure 3.4: View of the example material after the constants have been entered and a theta
rotation of 90 degrees has been applied.

material, as it has been provided, and has nothing to do with the direction of propagation. Enter
the constants in the boxes provided and ensure of course that the units are correct (These can be
changed by clicking on the down arrow next to the units).

Having entered the material parameters, and thus defined the material, it is now necessary to
rotate the material to define it in the desired direction of propagation. This is achieved by means
of the three “orientation” angles. Currently with the default setting of zero for each of the
angles Disperse sees the material as shown in Figure 3.3 and if this geometry were to be accepted
dispersion curves would be calculated for propagation in the Z direction with the fibres normal to
the direction of propagation, as shown in the figure. For reference the rotation angles phi, theta
and psi are also shown in the figure.

First of all, let us set up the layer for wave propagation along the fibres. Clearly we require a
rotation of 90 degrees in theta and this is entered in the appropriate data box to produce the
system shown in Figure 3.4. The propagation direction in the Z axis is thus parallel to the fibres
as required.

Disperse version 2.0.20a


3.3 Defining a Layer 33

Figure 3.5: View of the defined material after a theta rotation of 90 degrees followed by a psi
rotation of 45 degrees.

As an alternative let us now set up the layer for wave propagation at, say, 45 degrees to the fibre
direction in the median plane of the plate. To do this we additionally enter 45 in the psi rotation
edit box. Disperse thus rotates the material by 90 degrees in theta, as before, followed by 45
degrees in psi. This defines the layer shown in Figure 3.5.

This example has been somewhat laboured, since it is a likely source of error or misunderstanding.

Before saving the material, the slowness curve can be intuitively checked. In this case, since
the fibre direction is the direction of greatest stiffness, the Z axis (the propagation direction)
should indicate the minimum slowness of the longitudinal wave, while the X axis should indicate
the maximum slowness of the longitudinal wave. The quasi-shear waves should have the same
slowness in both Z and X directions, which should be greater than that of the longitudinal wave.
The Z-Y plane should appear the same, because both X and Y directions lie transverse to the
fibre direction and have the same stiffness.

Having satisfied ourselves that the material is correctly defined it can be added to the material
list by clicking on the button at the bottom of the form. Note that the rotations are also saved,
so that if various rotations about the psi axis are anticipated in future use, the material is best
saved with the 90 theta rotation.

Finally, click the OK button to return to the geometry definition where the thickness of the layer
(1mm) is entered.

3.3.4 Variation of properties with frequency

Since the solution of every point on a dispersion curve relates to a single value of frequency, it
is permissible to solve the dispersion relations for material properties which vary with frequency.
Indeed in principle the properties may vary according to any function of the frequency, provided
that it is smooth. Disperse offers a powerful option in this respect, by allowing any of the material
properties to be defined by a polynomial function up to 3rd order. This possibility has been
introduced earlier by specific data entry procedures (Section 3.3.2), mainly focused on providing an
easy way to define different material damping models. However, a polynomial function definition

Disperse version 2.0.20a


3.3 Defining a Layer 34

may be entered at any location where material data is entered. This could include, for example:

• Direct definition of frequency-dependent damping properties without using the dialogue box.
• Definition of frequency-dependent wave speeds or stiffness constants for isotropic elastic
materials.
• Definition of frequency-dependence of any anisotropic material properties, including engi-
neering constants, Cij stiffness constants, and Cij imaginary attenuation constants.

The variation function is defined independently for each property which is entered, so affording
great flexibility of input.

The form of the polynomial function is:

y = A + Bf + Cf 2 + Df 3 (3.5)

in which y is the material property which is being defined, f is the frequency in MHz, and A, B, C,
D are the polynomial coefficients. The polynomial is entered using the syntax: A;B;C;D directly
into the data entry box. It is not necessary to enter all of the polynomial coefficients; indeed the
usual definition of a property whose value does not vary with frequency corresponds simply with
the coefficient ”A” on its own. The procedure is illustrated by the following examples:

• Shear velocity. An entry of 3;0.01;0.002;0.0003 defines a shear velocity equal to 3 + 0.01f +


0.002f 2 + 0.0.0003f 3 m/ms.
• Shear velocity. An entry of 3;0;0.002 defines a shear velocity equal to 3 + 0.002f 2 m/ms.
• Shear velocity. An entry of 3 defines a shear velocity of 3 m/ms, without variation with
frequency.
• C11 real coefficient (stiffness) in an orthotropic material. An entry of 100;3 defines this
material stiffness coefficient to be 100 + 3f GPa.
• C11 imaginary coefficient (damping) in an orthotropic material. An entry of -0.1 defines
damping which follows the default hysteretic model such that bulk waves would attenuate at
a constant value per wavelength travelled, for any frequency (see Section 3.3.2 and Appendix
A); thus the attenuation per unit distance would increase linearly with the frequency .
• C11 imaginary coefficient (damping) in an orthotropic material. An entry of 0;-0.1 defines
damping which follows the Kelvin-Voigt viscous model such that the attenuation per wave-
length of bulk waves would increase as a linear function of the frequency; thus the attenuation
per unit distance would increase with the square of the frequency.

Once the material has been defined and the curve tracing window is open, the frequency functions
are displayed in the Geometry docking notes (Section 5.1). The values are listed using the same
A;B;C;D shorthand code.

3.3.5 Interface

In most cases it is reasonable to assume that the layers of a multilayered structure are perfectly
connected (“welded”) together so that there is continuity of the displacements and stresses across

Disperse version 2.0.20a


3.4 Opening an Existing File 35

the interfaces. However it can be useful to be able to model some forms of imperfect interfaces, for
example to represent rough surfaces in contact or interfaces which are free to slide. the Interface
tab allows such conditions to be defined. Although a boundary condition such as this strictly has
no thickness, conceptually Disperse deals with it as a special kind of layer. Thus it becomes one
of the layers in the multilayer system.

Separate specifications are given for each of the three possible directions: normal to the interface
(X direction), parallel to both the interface and the propagation direction (Z direction), and
parallel to the interface but normal to the propagation direction (Y direction), see Figure 3.1 for
directions. Clearly only those directions which are relevant to a particular problem need to be
considered, for example the Y direction stiffness plays no part in the Lamb wave solutions.

The option Rigid defines a perfect connection in the given direction at the interface and the option
Free defines complete disconnection in the given direction. The third possibility, User, allows a
finite stiffness to be defined. The stiffness is given in N/m per square meter of the interface.

3.3.6 Boundary

In addition to the possibilities of defining the structure to be in vacuum or embedded, there are
some other boundary conditions which can be specified at the top and/or bottom surface of the
structure. As with the spring interfaces, Disperse treats these external boundary conditions as
special layers in the multilayered system.

Free boundary sets the surface to be completely free in all three directions ; this is the same
as a vacuum boundary. Fixed boundary sets the surface to be rigid in all three directions.
Mixed boundary allows different conditions to specified in each of the three directions, in each
case either free or rigid. Symmetric boundary can be used where a structure has geometric
symmetry about a centre line. In this case it is only necessary to define one half of the structure
(the top half) and define a symmetric boundary at the bottom of the layers list (which is the centre
line). Symmetric boundary can only be used in flat plate cases; it has no meaning for cylindrical
geometries.

Once the material constants have been modified (or entered), you can add the new material to the
list of standard materials by pressing the Add to List button. The material that is currently being
edited will be appended to the material.lst file and will be available for future uses of Disperse.

3.4 Opening an Existing File

If a file was created during a previous session, it can be reopened by choosing File : Open or by
specifying the file name on the command line when Disperse is originally run. If some modes have
already been traced and saved, these modes will be read into memory and will appear.

Once the geometry is defined you are ready to search for modes. This will be explained in the
next chapter.

Disperse version 2.0.20a


Chapter 4

Tracing the Dispersion Curves

Tracing dispersion curves forms the core of Disperse activities. This section describes how to do
that, including some hints and guidelines for efficient analysis. This chapter is divided into areas
that approximate the order normally encountered. First, the automatic routine is described: this
covers all that is needed for the new user working with a relatively simple structure. Second, the
manual tracing methods are explained: these will be needed for the more advanced problems when
the automatic routine does not find all of the modes. Finally, some special solution options for
the expert user are explained.

If you have not already defined the geometry and material properties of the structure, then you
will need to do so before proceeding with this chapter. See how to do that in the previous chapter.

4.1 Overview

Dispersion curves describe which guided waves can exist in a given system. To find these possible
solutions we generally search for a superposition of bulk waves that will satisfy all of the neces-
sary boundary conditions. Although it is possible to derive equations for each case encountered,
Disperse usually uses a more general method called the Global Matrix method, which can relate
partial wave amplitudes in each layer to the stresses and displacements at the interfaces between
the layers. For more information, please refer to Chapter 7, which describes the wave model, or
references [51, 69] which describe various matrix techniques. Disperse searches for points where the
characteristic equation is satisfied. This condition usually correlates to guided wave modes. How-
ever, some other cases, such as when the phase velocity is equal one of the layer’s bulk velocities
can also cause the determinant to be singular, as explained in Section 7.5.

To find a starting solution and then trace out a mode (connecting solutions), Disperse uses the
following basic logic,

• A certain system is loaded, including the geometry and material properties.


• A line is swept in frequency, phase velocity, attenuation space.

• The minimum points along this line are found.


• The program converges at each of these minima searching for a root.

Disperse version 2.0.20a


4.2 Automatic Tracing of Dispersion Curves 37

• One of these converged points is chosen as a starting point and a mode is traced from it.
• Extrapolation routines predict the location to search for each new point as the mode is
traced.
• This mode is saved and/or investigated.

The rest of this section is dedicated to describing each of these processes in more detail and giving
some hints on tracing difficult modes.

4.2 Automatic Tracing of Dispersion Curves

To generate dispersion curves for the geometry that you have defined, several techniques may
be employed. The easiest method, especially for simple systems like a plate in vacuum, is to
choose the Trace : Automatic tracing option (or press F7). This routine makes sweeps along
several lines of the frequency-phase velocity space, looking for valid roots. The locations of the
sweep lines are chosen such that all of the modes are likely to be found in the common cases such
as Lamb waves. For more complicated problems the automatic sweeps may not find all of the
roots in which case the full set will need to be completed later using the manual tracing options.
You can see the results of these sweeps and the locations of the sweep lines as the small circles
appear to indicate roots or function minima which have been found. Having found the roots, the
automatic routine uses these as starting points to trace the modes. As the automatic routine
proceeds, the curves which are being traced are by default displayed in frequency-phase velocity
space. Frequency-phase velocity is the most commonly understood form of the dispersion curves
and so is used in Disperse as the default space for displays in general, and for the descriptions in
the manual. But of course the curves which have been found can be displayed instantly in other
forms such as wavenumber, group velocity etc, as explained in 5. In fact internally the program
performs the tracing calculations in frequency-wavenumber space.

The automatic routine has been designed to detect reliably the modes of the most common types of
system, such as Lamb waves, since this addresses the vast majority of the usage of Disperse. Given
the very wide range of possible systems and the mathematical complexity and conditioning of the
equations, it is not realistic to expect it always to find all of the curves in the more complicated
systems. This is especially true in cases where the attenuation is very high, such as strongly
leaky modes. Therefore, the user should not expect the automatic procedure to be 100 percent
reliable. Auto-tracing generally provides a good starting point for generating dispersion curves
but afterward, many curves may need to be traced manually, see Section 4.3.

Choosing Automatic tracing initially pops up a window asking for certain parameters. These
parameters include:

Trace a function of - this allows you to choose the principal variable for the solution. This will
almost always be frequency, which is the default. However you can choose another property
such as one of the material properties of the layers to be a variable, in order to perform a
parametric study. This will be explained in Section 4.4.2.
Top frequency - This determines the upper frequency limit for the calculations. Disperse sug-
gests a default value here, estimated according to the layer properties and dimensions.
Trace fundamental modes - if this is selected then the automatic routine will perform a sweep
which aims specifically to detect the modes which can exist down to zero frequency. These
are the a0 and s0 modes in the case of the Lamb wave solutions.

Disperse version 2.0.20a


4.2 Automatic Tracing of Dispersion Curves 38

Trace Higher order modes - if this is selected then the automatic routine will perform a sweep
which aims specifically to detect the modes which extend to high phase velocities. These are
the “cutoff” modes, a1, a2, etc, and s1, s2, etc in the case of the Lamb wave solutions. The
routine searches for the roots at one and a half times the greatest bulk velocity and then
traces any found roots to the maximum velocity.

Trace back to zero wavenumber - if this is selected then the automatic routine calculates this
low wavenumber information. If you commonly display the dispersion curves in wavenumber-
frequency space or in group velocity frequency space you will want to leave this option on,
otherwise you may turn it off. If selected, this option will take all of the higher order mode
curves and start new traces that continue backward until they hit a negative wavenumber.

Trace symmetric/antisymmetric/both - This option is only offered for some very simple
symmetric systems, such as a single isotropic plate in vacuum or immersed in a fluid, for
which specific analytical expressions can be used in place of the usual general purpose matrix
method. In these cases the symmetric and antisymmetric modes can be evaluated separately.
Trace Lamb and/or SH modes - This option is only offered for flat plate structures. By de-
fault the routine limits the calculations to the lamb type modes, in which there is no dis-
placement in the direction normal to the plane defined by the direction of propagation and
the normal to the plate. However, if SH modes are selected then the routine will search for
the completely uncoupled solutions in which the there is displacement only in the direction
normal to that plane. These are the Love modes in a single layer isotropic plate.

Trace zero/one order modes - This option is only offered for cylindrical structures. The zero
order modes are those which are axially symmetric, that is to say each displacement or
stress value is identical for all locations around the circumference of the structure. The one
order modes have one sine wave of variation around the circumference of the structure. The
automatic routine does not search for modes with higher order than one; if such modes are
required then they must be found using the manual search methods.

Trace/do not trace T(0,-) modes - this option is only offered for cylindrical structures. The
T(0,-) modes are the cylindrical equivalent of the SH modes in plates. The T(0,1) mode
in a simple unbounded single layer pipe is perfect torsion, involving displacements in the
circumferential direction only.

Vary attenuation or sweep with zero attenuation - this option is only offered when calcu-
lating dispersion curves for an attenuative system, that is if there is a fluid first or last
semi-infinite layer, in which case there can be attenuation due to leakage, or if any of the
materials have damping losses. If you choose to sweep with zero attenuation then the routine
will perform a single sweep assuming zero attenuation. This is quickest and it can often find
minima of the function which it can then pursue to solve properly for the complex roots.
However, sometimes the function only exhibits a minimum when it is evaluated very close to
the root, in which case the frequency, wavenumber/velocity and attenuation values all have
to be quite close to their solution values. If you choose to sweep with “vary attenuation”
then two sweeps are made, one at zero attenuation and one at a fixed fraction of whatever
value is set to be the “maximum attenuation” for the search (this is set by the Trace limits
in Trace : Convergence parameters, see Section 4.3.3). This takes longer but much
improves the chances of finding local minima. The default, which is to vary the attenuation
when searching for the fundamental modes, but leave the attenuation zero when searching
for the higher order modes, is a compromise which generally works reliably.
Convergence parameters - The increment is the step size which controls the number of points
used to calculate a dispersion curve. By default the routine takes these constant steps of
wavenumber while tracing each curve. If you increase this number then each curve will be
composed of a larger number of points. This can be useful if the plot of a curve is not smooth

Disperse version 2.0.20a


4.3 Manually Tracing Dispersion Curves 39

enough when you zoom in to look at a region of particular interest. Varying this step size
can also help the stability of solution in difficult cases. The Edit button opens a dialogue
box for adjusting other parameters of the solution. See Section 4.3.3 of for more detail. If
you are a first time user, we recommend that you leave all the convergence parameters at
the default values.

Once the routine has started, it can be stopped at any time, by pressing on the stop sign in the
toolbar, or by choosing Trace : Stop all jobs. Jobs can also be stopped selectively using the
Jobs tab in the docking notes (window to the left of the curves display). Disperse will ask you
if you want to cancel the remaining jobs. Answer yes if you like to exit the automatic tracing
routine or answer no if you would like to stop only the current step and continue with the next
steps.

When the automatic routine has finished there may be some places where curves are incomplete
or where only the minima are shown but tracing has not been successful. The minima are shown
as red or green circles; this is described in the following section on manual tracing. If the plot is
incomplete in this way, it is suggested that you,

• Select the Trace from selected button on the toolbar (it is the one that has a red arrow
that runs through a black dot).
• Click with the left mouse button on a circle which doesn’t correspond to an already traced
mode or on the end of a partially traced mode. The tracing limit is set by the text box on
the toolbar that is close to the Trace from selected button. If the tracing limit is set to
be lower than the value where you want to start tracing then the tracing will proceed in the
backwards direction.
• Repeat for other incomplete modes.

If there is a repeated problem, you may want to look at the output box, which by default appears
as a minimised window. Some of the convergence parameters may need to be adjusted as described
in the manual tracing section (Section 4.3).

The automatic tracing routine attempts to name some of the modes. The names are shown
whenever a point on a mode is selected, using the pointer. This option works relatively well for
simple Lamb wave cases, but has problems with more complicated systems. These names should
therefore be treated with caution. They can be changed by choosing Tools : Labels : Edit
mode names, then pointing at the curve whose name you would like to change.

Individual spurious points can occur at a location where the phase velocity of a mode is equal
to one of the bulk velocities of the materials of the system. These spurious points appear most
noticeably as sudden peaks on the group velocity curves. They can be removed by selecting Edit
: Delete : Next point selected and clicking on the spurious point. See Section 7.5.

4.3 Manually Tracing Dispersion Curves

If you are trying to create dispersion curves for complicated systems or the Automated tracing
routine is having trouble then you will have to trace the dispersion curves manually as described
in the following section. To find the curves, you will first need to make a sweep to find initial
points on a dispersion curve and then trace modes out from those starting points. Background

Disperse version 2.0.20a


4.3 Manually Tracing Dispersion Curves 40

information on our method of sweeping and tracing can be found in the theory section (Section
7)of the manual.

4.3.1 Sweeping

In order for Disperse to initially find roots, it has to perform a sweep in frequency, wavenumber,
attenuation space.

The easiest way for you to do this manually is with the Sweep and Search facility; this is
started from the menu which appears when you click the right mouse button. The window must
be showing frequency - phase velocity or wavenumber-frequency space (this generally means that
at least one mode will already need to have been traced). You then select the line for the sweep
by dragging a box across part of the plot. Place the mouse at one location on the plot, hold
down the left button, then drag to another location and release, thus creating a rectangular box
on the screen. Disperse will then search for roots along a line from the bottom left to the top
right of the box. Any minima or roots which are found will be shown by small circles. A green
circle represents a location where a minimum of the function has been found but it has not passed
the criteria to be accepted as a root (for example it my just be a local minimum, not a zero); a
red circle represents a location where the search algorithm has satisfactorily found a root. The
attenuation value for the calculations is taken to be the “current” value. This is the value most
recently selected elsewhere in the program, for example if you previously performed a sweep at a
certain value of attenuation then that would be the current value, or if you clicked with the pointer
on a mode then the value of attenuation at that position on the mode would be the current value.
When the program is started the value is set initially to zero. If Disperse has problems converging
on roots you should refer to Section 4.3.3 for advice on adjusting the search parameters.

If the geometry is cylindrical then each sweep operation is limited to one circumferential order.
This is indicated, and can be changed, by an integer on the toolbar immediately to the right of the
text Order. Similarly, certain simple problems with symmetric geometry, such as Lamb waves,
are solved using specific solution functions which find separately the symmetric and antisymmetric
modes. In such cases the toolbar indicates a choice of Symmetric or Antisymmetric.

A more general way to perform all sorts of sweeps is offered by selecting Trace : Sweep, or press
F9. You can choose which parameter to vary in the sweep, which to keep constant and their values,
and the range over which to sweep. You can limit the search to symmetric or antisymmetric modes
if the geometry is a simple plate. You can limit the search to a specific circumferential order if the
geometry is cylindrical. For example, for a plate immersed in water, you may select Frequency
(constant phase velocity), From 1 to 10 MHz, Holding the following constant: phase
velocity = 10 km/s, attenuation = 1 np/m. Then the program will search for roots varying only
the value of the frequency, from 1 to 10 MHz, and holding the other two parameters constant at
the specified values. Any roots which are found will appear as circles along the appropriate line
on the plot. You can also plot the characteristic function along the line of the sweep by selecting
the menu item Display : Sweep results, see Section 5.3.5.

Default values for the various limits will be suggested; these are based on the dimensions and
geometry of the structure. However, this is where a little experience helps. For example, predicting
attenuation values is frequently difficult. Therefore, if you want to generate curves for a leaky
system or if your materials are attenuative, you may want to change the attenuation values at
which you sweep, once you have an idea where they are likely to fall.

To help with the search for elusive roots in attenuative systems you can select Perform 2-D
attenuation sweep. With this option, Disperse performs multiple sweeps, each time increasing

Disperse version 2.0.20a


4.3 Manually Tracing Dispersion Curves 41

the value of the attenuation. By default it takes a minimum value of attenuation of zero and
a maximum value which it calculates from the material properties to be a likely sensible upper
limit. It then sweeps separately for each of five values of attenuation, from the minimum to the
maximum in equal intervals. The values of the minimum and the maximum attenuation can be
set by the Trace limits in the Trace : Convergence parameters menu: choose attenuation as
the variable and then set the minimum and maximum. Thus minima of the function are identified
within a two-dimensional space. Although this type of sweep may take quite a while, it can be
helpful to find troublesome roots.

Find minima If you would like to find the locations of the minima of the characteristic function
over an area of the solution space rather than just a single line, you can use the Find minima
routine, which is selected from the right hand mouse button. The find minima routine searches for
minima in the absolute value of the determinant of the global matrix without trying to converge
on roots at each of the minima. This should be used on a velocity or wavenumber display. The
attenuation value for the calculations is taken to be the “current” value. This is the value most
recently selected elsewhere in the program, for example if you previously performed a sweep at a
certain value of attenuation then that would be the current value, or if you clicked with the pointer
on a mode then the value of attenuation at that position on the mode would be the current value.
When the program is started the value is set initially to zero. Select this option then drag a
box over a rectangular area to define the region for the search. Any minima which are found are
indicated by small green circles in the usual way. These minima can be used as starting points for
traces by selecting one of them and choosing Trace : Trace from Selected Point.

Note 1. In general, the frequency sweep is the most reliable method because it can find many
modes and it is not susceptible to problems of finding zeros at the bulk wave speeds.
However, to find the fundamental modes, it is usually easiest to do a phase velocity
sweep at relatively low frequency and ignore any roots that correspond to the bulk
velocities.

Note 2. Some waves travel ’backwards’ (The phase and group velocities have opposite signs).
For these modes the attenuation will be negative since we define our direction of
propagation with respect to the phase velocity, not the group velocity.

4.3.2 Tracing

Once you have some converged roots (either from performing a sweep or from previous traces of
curves), you are ready to trace out a mode. There are two ways in which you can do this.

Trace: Trace from swept point . If you have performed a sweep and found one or more roots,
then you can choose to trace from these. Select this option from the menu (or press F10) to
bring up a window showing details of all of the roots which have been found by the sweep.
If you keep to the default on the top pull-down menu of this window, One mode, then
you just need to click on the button of the mode you want to trace. Alternatively you can
select Multiple mode in which case the window format changes and you can identify any
number of the modes which you want. In either case you may also indicate the limit of the
frequency, wavenumber, attenuation or phase velocity to which you want to trace. If you
want to trace in both directions from the root then enter the lower limit in the Reaches:
box and the upper limit in the And also to box.
Trace from point If you just want to trace a single mode at a time, then the most convenient
way is to use the Trace from point facility. This is found either as a button on the tool
bar (black dot with red line running through it) or on the right hand mouse button menu.

Disperse version 2.0.20a


4.3 Manually Tracing Dispersion Curves 42

Select this option, then click on the point where you want to start tracing. This can be
a converged root (red circle), a minimum (green circle) or any point on a curve which has
already been traced. Tracing from a point on an existing curve in this way can be useful if
you want to continue a trace that was prematurely stopped. The limit for the trace is defined
on the tool bar: select frequency (default), wavenumber, attenuation or phase velocity, then
enter the limit value in the box immediately to the right. Tracing may go in either direction,
depending on whether the limit is higher or lower than the starting point.

Once the tracing parameters have been chosen, Disperse proceeds in the following order,

• A small step step is taken in the step variable. By default the step is normally in wavenumber
but can be in one of the other variables, as defined by the Convergence parameters
described in 4.3.3. The algorithm then iterates the other variable (or two other variables
if there is attenuation) until a converged solution is found. This or these are the iteration
variable(s).
• Disperse extrapolates from these first two points to a third and then from the first and third
to a fourth point.
• Once the curve has been started by these small steps, the algorithm proceeds in a series of
steps of fixed size. On each step a quadratic curve fitting algorithm is used to predict the
initial values of the iteration variable(s).
• The stepping continues until the user-defined limit is reached.

The progress of the trace is indicated in two places: (1) an updating text string at the bottom
left of the screen shows the step number and solution values; (2) information is given on the Jobs
card in the window to the left of the main trace window. A trace can be stopped by pressing the
STOP button on the main tool bar, or by selecting Trace : Stop Jobs.

If two modes approach very closely to each other, the mode’s shape can usually be used to see if
Disperse stayed on the correct mode or if it jumped to another mode that passes nearby. Disperse
tries to detect when it is jumping between two modes and restarts a trace when this happens.
However, it can chose the wrong root to start on. In this case you will want to split the drawn
mode, delete the incorrect portion, sweep for a new zero in the region where the mode should have
gone, trace the rest of the mode, and then combine the two modes together. These procedures are
described in Section 5.4.3 on editing a curve.

A useful function, Show determinant , is offered on the right button menu for those wanting to
look at more detail of the numerical values and the matrices being used in the calculations of the
individual roots or points on the curves. This is really a debug facility and should not normally
be required but it can be useful for the expert user who wants to extract extra detailed numerical
information. The information should be interpreted with reference to 7. Select this option then
click at a point on a curve. The detailed information at that point will be shown in a window.

4.3.3 Converging on Roots

As a user, you will never directly interact with the routines that converge on a root because they
are automatically called by the sweep and tracing routines. However, it is important that you
understand the processes behind the search routines, so that you can assess (and conquer) more
complex problems.

Disperse version 2.0.20a


4.3 Manually Tracing Dispersion Curves 43

Disperse looks for roots by holding one variable constant (this is the variable called the step) and
iterating the other variable (or two variables if there is attenuation) until a minimum is found that
is within the tolerances set. By default the step is normally the real wavenumber but any of the
other parameters can be selected to be the step if desired. The iteration uses a five point bisection
algorithm which is not particularly fast to converge but is extremely robust, robustness being
essential because of the problems which can be encountered when two or more roots are in close
proximity. Disperse then checks to see whether this minimum is a root, rather than just a local
minimum of the function, by detecting if there is at least a ninety degree phase shift in complex
space between points that span the suspected root. Further explanation of these procedures is
given in 7 (see in particular 7.8) and [51]. The parameters relating to these searches are chosen
automatically and in most cases do not need to be understood or changed. All of the values are
scaled by thickness so they should remain similar for most systems. Also, the increments are scaled
depending on the type of variable on which the parameter is acting. However if roots are proving
difficult to find then it is necessary to intervene and alter the parameters manually. Experience
suggests that, unless you know already what values are going to work best, you are most likely
to be successful if you modify the values slowly, for example by no more than 50 percent at a
time. To change the search variables or values select Trace : Convergence parameters, or
press F11. The parameters which appear, and some suggestions to help with setting them are:

Step type This is the primary search variable. When tracing a mode the program steps along the
curve using this variable, solving for the other variables at each location on the curve. This
is explained in more detail in the paragraphs preceding. For simple problems such as Lamb
waves the choice of wavenumber works very well because of the nature of the dispersion
curves in frequency-wavenumber space. If you look at these curves for Lamb waves you can
see that it is possible to increment the wavenumber and move steadily along any one of
them without encountering severe changes in slope, the possibility of a dual valued solution
or other numerical difficulties. Compare this with the curves in phase velocity-frequency
space where stepping in either frequency or phase velocity would lead to infinite slopes of
the curves, and thus multi-valued results, at some positions. However some more difficult
problems have curves which are not simple in any of the possible solution spaces. Then
it is necessary to switch to another variable for part of the solution. For example, curves
sometimes loop around and cross over themselves in the shape of a lower case letter “e”
in the frequency-wavenumber space. Then it would be best to use the wavenumber for the
steps along the parts of the curve which are most parallel to the wavenumber axis, and the
frequency for the steps where the curve is most parallel to the frequency axis. If you are
having trouble tracing a curve then it is helpful to look at the curve in this way and identify
the variable which is most likely to work at the location of difficulty. In fact, in addition to
the straightforward choices of frequency, real wavenumber, attenuation and phase velocity,
Disperse can use a hybrid parameter called Freq-wavn combo. This steps along the curve
in the frequency-wavenumber space, taking steps which are measured along the tangent
to the curve. Thus each step increments a combination of frequency and wavenumber, the
proportion depending on the slope of the curve. This scheme can often automatically traverse
some quite difficult shapes such as loops. Since it is as reliable as wavenumber stepping for
the simple problems and quite powerful for many difficult cases, it has been set as the default.
Increment This value sets the length of the step. Its units are those of the step variable . The
value should be reduced when tracing areas of dispersion curves that change quickly in order
for the algorithm to be able to track the changes. But beware that making this value too
small causes problems with the smoothness of the group velocity curves. This is because
the group velocity is calculated by numerical differentiation of the wavenumber curve; any
small errors, the amplification increasing as the step size decreases. The overall shapes of
the group velocity curves are not compromised but ’noise’ may be present. Another tip is
that it can actually help to increase the step length sometimes. If you are having trouble
tracing because one curve won’t cross another but simply joins it or is diverted because of

Disperse version 2.0.20a


4.3 Manually Tracing Dispersion Curves 44

it, then try increasing the step size. There are procedures in the algorithm for avoiding
crossing problems (these generally work extremely well) and they can be helped by taking
larger steps.
Iteration type This specifies the parameter which will be iterated at each step in order to find the
solution. If a step has been taken in wavenumber then it makes sense to iterate in frequency.
Consider for example the Lamb wave solution with steps in wavenumber: iteration of the
frequency then searches along a path which is approximately normal to the direction of the
steps. Clearly you should never choose to iterate the same variable as is used for the steps.
However, if you have chosen Freq-wavn combo for the step type then you should also
choose this for the iteration type. This is because Disperse will automatically iterate in
another combination of frequency and wavenumber which is normal to the direction of the
curve. Thus the steps move along in a direction tangential to the curve, whatever its slope,
and the iterations are always along a line which is normal to the curve.
Search width This value specifies the step size of the iteration variable. Disperse starts by
taking steps of this size of the iteration variable, looking for a minimum of the characteristic
function. Having found a minimum it then performs a local search. Therefore this search
width does not affect the final accuracy, however it does affect the resolving power to find
roots, particularly when they are close. If the value is too large then the algorithm may
miss the minimum altogether. For example sometimes the “valley” of a minimum is very
narrow and a large value of this iteration step can take the search away onto a plateau which
does not have a slope towards the root. A large value can also cause the program to find
a different nearby curve instead of the desired curve. On the other hand if the value is too
small then it may require an unacceptable number of iterations before the algorithm can
reach the minimum of the function. In general it is a good idea to choose the value always
to be smaller than the Increment. It should be reduced if there is a problem of the tracing
routines jumping to other modes.
Tolerance This specifies the accuracy to which the root should be found. After the algorithm
has iterated to find a minimum it performs a five-point bisection sequence to converge more
precisely on the root. It takes the point where it has found the minimum and the point
on either side of this, then bisects each of the adjacent intervals, so reducing the interval
between points by a factor of two. It identifies a new minimum from amongst the five points
and repeats the process. When the interval has been reduced to the value of the Tolerance
then it stops and performs a convergence check. The convergence check determines whether
the final point which has been found is a root (i.e. the value of the characteristic function
is acceptably close to zero) or just a local minimum. Thus the Tolerance does determine
the accuracy of the solution and should be set to a small value if precise results are needed.
You should avoid very small values because of problems of numerical rounding: we suggest a
sensible lower limit of about 1.0e-8 but this is problem-specific. In addition, the value must
always be smaller than the search width, otherwise roots will not be found.
Iteration type (2nd set) (only present if the modes may have attenuation) A second set of the
three parameters Iteration type, Search width and Tolerance, is required if the modes may
have attenuation. Disperse identifies whether to ask for this second set according to the
geometry and materials. If the structure has one or more fluid or solid half-spaces, or any
one of the materials has non-zero damping properties, then attenuation can be present in
the solution and the characteristic function becomes a function of three variables. The root
finding algorithm now has to add another dimension to its search. The procedure is to search
in the first iteration variable (as explained above) first, then perform the same kind of search
in the second iteration variable, then return to the first iteration variable, and so on. When
searching in one iteration variable the other is kept constant. Each search proceeds until
the bisection interval diminishes to the value of the Tolerance and results in a minimum
of the function which hopefully is smaller then the previous. The convergence criterion is

Disperse version 2.0.20a


4.3 Manually Tracing Dispersion Curves 45

checked after each search and used to determine whether the end of the process has been
reached. If wavenumber, frequency, and/or Freq-wavn combo have been specified for the
step type and the first iteration type (as is usually the case) then a sensible initial choice
of this second Iteration type is attenuation. Clearly it is important that the two iteration
types are different.

Trace limits These lower and upper limits in frequency, real wavenumber, attenuation or phase
velocity allow the user to set the region of interest for the solution. Disperse will not search
for roots outside the ranges specified here. By default Disperse starts with values here which
it expects to be sensible. These are then reset if the user chooses values outside the ranges
either in this menu or elsewhere in the program (eg by choosing a trace limit which exceeds
the maximum here). One particular parameter to keep an eye on is the attenuation. The
minimum and maximum values of attenuation which are set here determine the parameters
for the 2-D sweep, see Section 4.3.1.

If when tracing, the routine fails, you will need to make some adjustments. The best choice comes
mainly from experience,but a good starting point is to look at the output on the Jobs tab in the
docking notes (window to the left of the curves display). This will say at which step the trace
failed and suggest if it is likely that the real or imaginary step lengths should be increased. Some
possible choices are:

Fails on Step 1 Try increasing the real or imaginary tolerance since for the first two steps the
actual tolerance is five times smaller than normal in order to obtain extra accuracy for the
initial trajectory.
Fails on Step 2-4 Try decreasing the increment since at this stage we are still using simple
extrapolation because we do not yet have enough points for the more reliable quadratic
extrapolation scheme.

Poor convergence, attenuating system Try increasing the imaginary step length to some-
thing very large.
Converging to bulk velocities Try decreasing the real and imaginary step lengths and/or in-
creasing the increment.

Stops after 2000 steps A maximum of 2000 steps in a single trace is set within the program to
avoid indefinite tracing which could happen when the program is left to run automatically,
so this is a natural stop. If you need to go further then simply click on the last point and it
will continue tracing.
Stops with error code 0 This indicates poor conditioning of the matrix, for example at one of
the bulk velocities; it can help to decrease the iteration search width considerably.

You may find that there are places where you just cannot continue to trace a curve. For example,
in leaky systems there are often discontinuities at the bulk velocities of the outer medium where
the system switches between the different domains where both the longitudinal and shear partial
waves leak, where only the shear waves leak, and where there is no leakage. Modes which cross
these domain boundaries are actually connected, but since Disperse only looks for propagating
solutions, these solutions are not found. Furthermore, modes such as S1 are discontinuous in
attenuative systems, because they change from travelling ’forward’ to travelling ’backward’ at
which point there is an abrupt change in the sign of the attenuation.

Disperse version 2.0.20a


4.4 Special Solution Options 46

4.4 Special Solution Options

4.4.1 Types of Solution

There are various choices which can be made regarding the characteristic function which is used
for the solution. Disperse makes a selection of these by default and it is usually not necessary to
pay attention to them. Certainly the new user is advised to stick with the defaults to start with.
To make changes to these settings, select the menu Trace : Solution type. This will show a
dialogue box; its items are as follows:

Also trace non-propagating roots - Disperse has been developed initially for guided wave
NDE and thus has a focus on modes which propagate, that is to say, although modes maybe
complex their wavenumbers are significantly real. However, modes whose wavenumbers are
significantly imaginary can also exist, even in a simple isotropic perfectly elastic plate in
vacuum, these being the non-propagating roots. More information about these can be found
for example in [4, p. 83]. It can be useful to know about these modes when studying localized
fields at discontinuities where there is mode conversion. By default Disperse does not look
for these modes but this option allows such a search.
Also trace roots with negative wavenumbers - By default Disperse considers only wavenum-
bers whose real part is positive, that is waves whose phase propagates in the positive Z
direction. In general the solutions for waves which travel in the negative direction are of
course identical and are therefore not pursued. However it maybe desirable to allow for
solutions in both the positive and negative wavenumber domains, for example when looking
for non-propagating modes in which the the imaginary wavenumber dominates. Note that
it is not necessary to select this option in order to find the backward travelling modes such
as the Lamb waves S1 or A2. These have positive phase velocity (by definition, as stated
above) and negative group velocity, but the wavenumber remains real.

Frequency is real/complex - Only relevant for systems where there can be attenuation, i.e.
if there is material damping and/or the structure is bounded by a solid or a liquid. The
complex form of the wave equation can be expressed in terms of a complex wavenumber or a
complex frequency or both. A complex wavenumber with real frequency is nearly always used
in the literature and is the default offered by Disperse. However, solutions using complex
frequency are fundamentally different (they cannot be mapped from one to the other) and
in some special cases may be more representative. This is discussed in detail in [79, 8],
see also 7.3.4. If the complex frequency option is chosen then Disperse solves using a real
wavenumber and a complex frequency. To display the complex frequency results see Section
5.3.2. The possibility of both frequency and wavenumber is currently not offered. Unless
you are familiar with this specialist topic it is recommended that you keep to the default of
complex wavenumber.
Trace/do not trace SH modes - This option is only relevant for flat plate structures. The
Lamb type modes have no displacement in the direction normal to the plane defined by
the direction of propagation and the normal to the plate. In contrast the SH modes have
displacement only in the direction normal to that plane - these are the Love modes in a
single layer isotropic plate. The SH and the Lamb modes are completely uncoupled in an
isotropic plate, in which case the full set of solutions is made up of the solutions using only
SH and the solutions using only Lamb. The options here allow you to solve for one or other
of these exclusive sets, for both separately (separate calculations are performed sequentially),
or everything in a single analysis in which case all of the equations are present thus allowing
the solution of coupled modes. The advantages of finding the modes using the uncoupled

Disperse version 2.0.20a


4.4 Special Solution Options 47

solutions, where this is possible, are improved speed and elimination of the risk of jumping
between modes of the different types when tracing.
Trace/do not trace T(0,-) modes - This option is relevant only for cylindrical structures. The
T(0,-) modes are the cylindrical equivalent of the SH modes in plates. The T(0,1) mode
in a simple unbounded single layer pipe is perfect torsion, involving displacements in the
circumferential direction only. However this is something of a special mode - most modes
involve coupling between the circumferential and other components. You can select from
options of exclusive or coupled solutions in the same way as was described for the previous
item.
Place sink at origin - This option is relevant only for cylindrical structures with liquid cores.
By default Disperse assumes that there is a sink at the origin, that is to say, all partial waves
which leave the inside surface of the solid and propagate into the fluid core vanish when they
arrive at the axis; they do not cross the core to insonify the solid on the opposite side. The
presence of a sink prevents standing waves from being created in the fluid. Of course it also
introduces attenuation into the system because the partial waves radiating into the fluid are
lost. If you are mainly interested in what will happen in the pipe, and are looking for fast
moving modes, you will probably want to keep this option on because the fluid modes can
be numerous and can clutter the diagram. If you are interested in the standing waves and
what will propagate as a whole, then you will probably want to turn this option off. The
significance of this choice is discussed in Section 9.2.1. Note that this option is for fluid cores
only; it has no effect if you have a solid or vacuum core.

Choice of the characteristic function Disperse automatically chooses the appropriate form
of the characteristic function and there should normally be no need to change this. Indeed
under most license arrangements it will be possible to make only limited changes here. For
information, the selection showing two asterisks (**) is that which Disperse identifies as the
most appropriate, one asterisk (*) denotes that the function matches the coordinate system,
and no asterisk denotes a clearly inappropriate function.

Cartesian (Global matrix) This is the general purpose function which is discussed in
detail in 7.5. It is valid for all of the types of flat plate problems covered by Disperse
and is used by default unless there is a specific function which could be used to advantage
in its place (for example the Analytical leaky Lamb waves). In such cases this menu
can be used to override that selection with the global matrix method if desired.
Cartesian (general anisotropic) This is a developmental option reserved for some spe-
cific anisotropic applications.
Cylindrical (zero order developmental) This is a developmental option reserved for
some specific cylindrical applications.
Cylindrical (V1 Fortran routines) This is the default function for cylindrical applica-
tions.
Analytical Lamb waves Only for (strict) Lamb waves. Rather than the general purpose
Global matrix function, Disperse can use an analytical solution for the characteristic
equation for the special case of a single flat isotropic layer in vacuum (Lamb waves).
See detail in 7.7. This is chosen by default because it significantly increases the speed of
solution. Also, because it solves the symmetric and antisymmetric modes separately, it
completely avoids mode crossing and so improves the stability of the calculations. An
advantage for the display of the curves is that Disperse is able to show the symmetric
and antisymmetric modes with different colours.
Analytical leaky Lamb waves Only for waves in a single flat isotropic layer immersed
in a perfect fluid. As described for the previous item, Disperse also has an analytical
solution for this case.

Disperse version 2.0.20a


4.4 Special Solution Options 48

Circumferential waves Only for waves which travel in the circumferential direction in a
cylindrical structure.

Job priority boost This specifies the priority to be given to the calculation job compared to
that of the user interface. By default the job priority is set to the lowest value in order for
the user interface to respond well. If this priority is increased then the calculation job may
run a little faster but the user interface will become less responsive. In our own experience
the increase in calculation speed is modest and is usually not worth the loss in interface
response. We therefore recommend the default setting.

4.4.2 Material Property Variation

This is a new feature for version 2.0. The preceding description in this chapter addressed the
solution of the guided wave functions in terms of the modal variables frequency, wavenumber and
attenuation. While the solution finds roots with these as variables, the other parameters, such as
the material properties and thicknesses of the layers, are assumed to remain constant. In most
cases this is appropriate. However it can be useful sometimes to allow for the variation of one or
more of these other parameters. For example, you might want to see how the phase velocity of a
particular mode at a particular frequency varies as function of the Poisson’s ratio of the material.
Disperse now has powerful and flexible features to calculate such solutions. The features can be
considered in three categories: Continuously variable properties, Properties vary between
modes, and Properties vary as a function of frequency or user variable. Each of these
will be explained in turn:

Continuously variable properties For a normal calculation of a set of dispersion curves, the
frequency, wavenumber, and (possibly) attenuation are allowed to vary, while the material
properties and thicknesses of the layers are held constant. However, it is often beneficial to
determine how the dispersion curves change as the material properties change. Therefore
Disperse allows you to set a material property to become a variable in the solution, then the
dispersion curves can be traced as a function of it. The choice of any new variable(s) is quite
flexible (density, bulk velocity, thickness etc of each layer), but there is a restriction that the
number of variables has to be equal to the number which the root searching algorithm is
expecting - i.e. 2 variables for problems without attenuation, 3 variables for problems with
attenuation). Therefore if one of the material “constants” becomes a variable then one of
the usual variables (frequency, wavenumber etc) must be chosen to become a constant for
the duration of the solution. These are called Continuously variable properties because
they are set for the whole solution and thus apply for all of the calculations and all of the
curves which are displayed in the window and stored in the file.
Properties vary between modes The values of the material properties can be configured to
allow them to be the same for all points along a particular curve but to vary from one curve
to another. For example this allows the results of a parametric study to be shown on the
same graph and stored in the same file. These parameters are therefore variable for the whole
solution set but actually they are held as constants while each curve is being calculated. An
example where this kind of variable already existed even in version 1.0 is the circumferential
order of the modes in a cylinder: the circumferential order is a constant at all positions along
any curve but may vary from curve to curve. The generalization of this idea here simply
extends the concept to other parameters of the layers.

Properties vary as a function of frequency or user variable The variation of the material
properties with frequency was already introduced in 3.3. This simply requires the search
algorithm of Disperse to look up the material properties each time it does a calculation,

Disperse version 2.0.20a


4.4 Special Solution Options 49

rather than assuming that they remain the same for the whole set of curves. Since any
single calculation in the search works at a particular value of frequency, it is straightforward
for the material database to return a value which is appropriate to the frequency of the
current calculation. This type of variable therefore does does not actually change the nature
of the searching or tracing, only the data values which they use. As an example of the use
of this idea, the attenuation constants of a material could be set up so that they vary with
frequency and better represent particular experimental data for the real material.
This idea is extended by allowing the material properties to be linked to the value of a
generic user variable rather than the frequency. The user can then control how the user
variable behaves. It can be set to be a fixed value for each mode or it can even be set to
be a continuously variable property. As an example of this idea, a study of the effects of
temperature on a particular dispersion curve may be performed, in which the the density,
longitudinal velocity and viscosity of a fluid vary with the temperature. These properties
can then be set to vary according to a user variable and a series of curves calculated, each
with a different value of the user variable.

The following paragraphs explain how to use these facilities. It should be noted that tracing
curves with variable material properties is much more difficult than the traditional method of
tracing curves and should be considered to be an advanced option. Two areas of the program are
affected:

Definition of the layers data - Establishing frequency (or user dependent properties) is per-
formed in the material edit dialog when the geometry is originally being defined, see 3.3.
Frequency dependent properties can only be set for isotropic materials, including liquids.
On the ’Liquid’ and ’Isotropic’ tabs, there are pulldown boxes with four options: values are
constant with frequency, values change automatically with frequency, values depend on fre-
quency, and values depend on user variable. The option, values are constant with frequency,
indicates that the values are set. The second option, values change automatically, will be
automatically set by Disperse for certain types of viscosity. It indicates that certain internal
quantities (for example the attenuation per wavelength) will need to be recalculated for each
frequency, however the constants that the user has entered will not change. The third op-
tion, values depend on frequency, allows the user to specify how the material properties will
change with frequency. By clicking on the property label, which should be converted into a
button when this option is selected, a cubic function can be specified for the frequency be-
havior. Similarly, the final option, values depend on user variable, allows the user to specify
a cubic equation describing how the property changes as a function of the user variable. The
value of the user variable will be set later when preparing to trace the curves.
Specification of the variables in the Trace menu - There are three places where tracing us-
ing variable properties can be set up:
Trace : Automatic tracing This is the easiest method. When you select automatic trac-
ing in the usual way the pulldown menu at the top of the dialogue box shows its default
Trace a function of Frequency. Simply change this to the variable of your choice
from the pulldown menu. You will see that the pulldown menu is automatically popu-
lated with the possible parameters relating to your specific structure. The variables will
be of the type which you used to define the material; thus if you defined the material
using Young’s modulus and Poisson’s ratio then these will be the available variables
rather than, say, the bulk velocities. Ranges for the variation of the selected material
property and the phase velocity must be specified. On selecting OK, the selected ma-
terial property will automatically be set to be variable and the frequency will be set to
be constant (and equal to the value specified in the appropriate box). The frequency
must be set to a constant value or else the dispersion curve would become a plane (in

Disperse version 2.0.20a


4.4 Special Solution Options 50

frequency - material property space) and could not be converged upon by the solution
routines. Before tracing a set of dispersion curves as function of a material property, it
is helpful to trace them as a function of frequency so that an appropriate frequency can
be chosen and the nature of all of the roots is relatively well known. An example of this
type solution can be seen by defining an adhesive joint (Free boundary, 1.0 mm thick
Aluminum layer, 0.05 mm thick epoxy layer defined as Young’s modulus and Poisson’s
ratio, symmetric system) and then varying the modulus of the epoxy layer at 0.5 MHz.
The effect on the phase velocity (although small) can be observed.
Trace : Change variables To set up the choices for the trace manually, you will need
to use both this menu and the next menu (“Change variable values”). This menu
establishes which parameters are variable, then the next menu will set values to those
which you have decided to fix. There are three categories of variables, as described
earlier: continuously variable properties, fixed for a mode but varying between modes,
and fixed. To change the nature of a property simply change the column in which
it resides by using the arrow buttons. If you have set the materials to be dependent
on a user variable, then you will see that “User variable” appears here as one of the
properties which you may move. Some limitations should be kept in mind as you change
these variables:
• If a property has ever been variable, for example when tracing a previous curve,
then it can no longer become fixed but will be limited to the middle category of
being fixed for a curve but varying between curves. This restriction results from
the way in which the modes are stored.
• Phase velocity and Group velocity are considered to be derived quantities and are
therefore always variable provided that the frequency or wavenumber are variable.
• If there is no attenuation present in the system, then there should always be exactly
two items that are continuously variable. This is because the plotting routines can
only handle dispersion curves that are lines. If only one item is variable, then the
dispersion curves become points and if three items are variable then the dispersion
curves become planes.
• If attenuation is present, there should be exactly three items that are continuously
variable, one of which should be attenuation.
Trace : Change variable values Once the parameters have been placed in their chosen
positions in each of the three categories in the “change variables” menu, data values
need to be set for those in the middle category, that is those which are fixed for the
curve but vary between curves. The values which are set here will remain in effect for
subsequent sweep and trace operations. However, please note that if you click on a
point on an existing curve then the values are changed immediately to those of that
curve.

Disperse version 2.0.20a


Chapter 5

Displaying the Dispersion Curves

The user interface of Disperse can best be described as an interactive graphing program. In
addition to displaying the data in various ways, it can respond to your requests and start new
calculations. The options available for displaying are described in this chapter. You should also
look at the next chapter, 6, which presents some of the additional processing which you can
perform once you have found your curves, for example calculating and plotting mode shapes, and
simulating a time domain signal.

5.1 The Basics

Internally, the calculated dispersion curves are stored in 7 dimensional data sets (i.e. frequency,
wave number, attenuation, two variable dimensions, phase velocity, and group velocity), which
can be plotted in various ways, or can be copied, printed, or exported.

The Display menu contains the options to change how these data are displayed. When you first
start using Disperse it is not necessary to get into the detail of the many possibilities for display;
instead you would be wise just to stay with the first four choices on the menu which display the
dispersion curves in the most popularly used forms:

• Phase Velocity vs. frequency in MHz.


• Group Velocity vs. frequency in MHz.

• Attenuation vs. frequency (or real wavenumber)


• Frequency vs. Real Wavenumber

Choosing one of these options from the menu immediately displays all the curves which you
have calculated in the chosen format. At the same time it clears any plots which were showing
previously. At the beginning this is probably all you need to know about the display, although it
would be helpful to read Section 5.4 very soon.

You will also see that the display window has a separate box attached to its left side which contains
docking notes, showing Geometry, Notes and Jobs. You can click the tabs to move between

Disperse version 2.0.20a


5.2 Default two-dimensional Display Types 52

these choices; you can also click on and drag the divider bar between the docking notes part of
the window and the main graphical display in order to change the width of the docking notes.
For example it can be useful to reduce the size of the docking notes when you want the maximum
area to display the curves. Geometry shows the structure which is being analyzed, using a tree
format. Click the + symbol to show expanded detail of any of the layers. Notes allows you to
enter your own text notes. These are stored with the current file and so can be viewed, and if
desired edited, in later sessions using this file. Jobs shows the status of the searching and tracing
jobs which are running or which have run most recently. For further information about this see
Chapter 4.

Details of the graphical display options follow.

5.2 Default two-dimensional Display Types

This section elaborates the first four display options on the Display menu.

5.2.1 Phase Velocity

The phase velocity is plotted by selecting Display : Phase velocity. This projection is the view
that is most frequently shown. The phase velocity of a guided wave, which describes the rate at
which individual crests of the wave move, is related to the real wave number as vph = ω/ξ, where ω
is the circular frequency (= 2π times the cyclic frequency) and ξ is the real wave number. Because
of this relation, the phase velocity display essentially shows the same information as the real wave
number display and so these two displays can be used interchangeably, according to preference.
However, the phase velocity view is more convenient to use for realistic ultrasonic testing. It is
also the easiest when comparing the phase velocity to the bulk velocities in the layers to study
the behaviour of leaky waves.

When studying waves which travel in the circumferential direction in a cylindrical structure, it is
important to be aware that the fundamental velocity quantity is the angular velocity. Thus the
linear (tangential) velocity varies linearly with the radius at which it is calculated. For consistency
with other kinds of solutions, Disperse plots the velocity as the linear quantity, and the convention
is that this is calculated at the radius of the first interface in the structure. If the angular velocity
is required (radians per second), then this can be obtained by dividing the phase velocity value by
the radius of the first interface. This can be done using the Tools : Calculate function facility,
see Section 5.3.2 below.

5.2.2 Group Velocity

The group velocity is plotted by selecting Display : Group velocity. This projection displays
the speed at which a guided wave packet (or envelope) travels. This rate is determined by how
quickly the energy of the wave will progress down the structure and will always be smaller than the
fastest bulk wave present in the system. Disperse calculates the group velocity from the derivative
of the dispersion curves such that
∂ω
vgr =
∂k
where vgr represents the group velocity, ω the angular frequency, and k the real wave number.

Disperse version 2.0.20a


5.2 Default two-dimensional Display Types 53

The above relation is the well-known conventional definition, although what is not so well known is
that it is strictly only valid for systems that do not include attenuation. In attenuative systems, it
is no longer valid, although it is usually reasonably accurate if the attenuation is small. When the
attenuation is large then physically impossible solutions may result, for example the group velocity
calculated in this way may tend to infinity as a mode reaches cut-off. An accurate alternative is to
calculate the energy velocity which is identical to the group velocity when there is no attenuation
and properly gives the velocity of a wave packet when there is attenuation, although it does take
significantly longer to calculate. Section 5.3.2 explains how to do this, the equation is given in 7.10,
and details may be found in [7] explain the theoretical background. Therefore if your curves only
have mild attenuation in the regions of interest you are advised simply to use the group velocity
and ignore the excursions at any locations of high attenuation. However if you need accurate
results for attenuating waves, then calculate the energy velocity.

The group velocity can also be used to predict how much dispersion will occur for a given wave
packet. Any wave that lasts for a finite time period will excite a range of frequencies, the size of
which depends on the number of cycles in the wave packet, as well as the shape of the packet.
Each of the frequency components in the wave packet will travel at a different group velocity. If
the difference in velocities over the range of frequencies is great, there will be a lot of dispersion
as the different frequency components arrive at different times. This spread will cause the signal
to change shape as it propagates. However, if the group velocities are very similar over the
generated frequency range, the packet will maintain the same shape during its entire propagation
length. Therefore, for practical testing situations, we usually prefer to use a centre frequency that
corresponds to a maximum in group velocity so that the predominant frequency components all
travel coherently. Disperse does in fact offer a facility to simulate the effects of dispersion in a
wave packet. This is explained in Section 6.2.

Currently the calculation of group velocity is also unrepresentative of the wave packet velocity
if the structure is anisotropic. This is because the energy of a wave packet does not necessarily
follow the phase direction (sometimes referred to as skewed phase). Since Disperse only works in
a single plane through the plate, that of the phase, it therefore does not incorporate this effect
at present. The implication is that the group velocity should be correct if the orientation being
studied aligns with a principal plane of the plate, otherwise it is not correct.

5.2.3 Attenuation

The attenuation is plotted by selecting Display : Attenuation. This plot shows the decay of
the guided wave. In a perfectly elastic system which is surrounded by vacuum, the attenuation is
always be zero and therefore does not need to be plotted. However, whenever damping materials
are present, or the system is surrounded by a liquid or solid medium, there is a mechanism for
energy to be lost from the waves and so the solution in general includes attenuation (although note
strictly that there are cases of guided waves in an immersed or embedded structure which do not
leak and therefore do not have attenuation). The attenuation is is a property of the guided wave,
just as much as is frequency or velocity. The model which Disperse uses by default accounts for
the attenuation by incorporating a complex wavenumber, whose units are nepers per meter. An
attenuation of κ nepers per meter means that a wave of unit amplitude is reduced to an amplitude
of e−κ after travelling one meter. However, by default Disperse shows the attenuation in units of
decibels (dB) per meter since this is the more popular unit with NDE people. They are simply
related: 1 dB = 0.115 nepers, 1 neper = 8.69 dB. If you prefer nepers per meter as the default
then you can change to that using the Display : Preferences menu, see 5.5. You can also set
the display in nepers per meter, or indeed nepers or dB per wavelength (np/wl or dB/wl), by
using the Select axes facility, see Section 5.3.2.

Disperse version 2.0.20a


5.3 Other Forms of Display 54

5.2.4 Real Wave Number

The real wave number is plotted by selecting Display : Real wave number. This plot displays
the relationship between the temporally and spatially varying wave characteristics of the guided
mode along the direction of propagation. The wave number, which is described in Section 7.3.4,
is inversely related to the wavelength of the guided wave by the equation λ = 2π/ξ, where λ is the
wavelength and ξ is the real wave number in radians per meter. The tracing routines explained
in Section 4 operate in real wave number – frequency space because the dispersion curves appear
more linear in this projection than in any other projection that contains the same information.

Note that when studying waves which travel in the circumferential direction in a cylindrical struc-
ture, the wavelength is that at the radius of the first interface of the structure - see the explanation
in Section 5.2.1 above.

5.3 Other Forms of Display

5.3.1 Second Graph

The Display : Second graph option causes Disperse to draw two graphs in the same window.
The top graph will contain the information that was already showing in the window prior to
choosing this option. The smaller bottom graph contains the projection that was chosen from the
Display : Second graph menu, that is, Phase velocity, Group velocity or Attenuation.
The y axis of the second graph can be changed to other possible axes by modifying the “Y2” axis
with the Select axes facility, see Section 5.3.2. The two graphs share a common x axis.

Points that are selected on one of the graphs will be marked on both of the graphs so that common
points can be identified. In addition, any zooming that is performed on one of the graphs will also
affect the other graph, since the x axis is shared.

You can revert to a single axis by choosing Display : Second graph : Hide .

5.3.2 Select Axes

The Display : Select axes menu allows you to select alternative axes to the four common ones
which were described above.

It is simple to use this facility: you just select your choice for the x axis and the y axis separately
from the selections on the pull-down menus. If the window is currently showing a second axis then
you are given an extra pulldown menu for the second axis. Most of the options are self-evident
and need no explanation. Some of the options are explained here:

Frequency-thickness, Wavenumber-thickness, Attenuation-thickness - It is well known


that the frequency axis of the Lamb wave curves can be scaled with the thickness of the
plate. Therefore a scale of Frequency-thickness (often referred to as “fd”) is convenient.
Similarly the thickness can be used to scale the wavenumber or attenuation in many cases.
Normalized frequency - This sets the axis to the frequency-thickness product and then divides
by the shear wave speed in the plate. This is a dimensionless quantity sometimes used in

Disperse version 2.0.20a


5.3 Other Forms of Display 55

the literature. Clearly this is appropriate only for single layer structures.
Attenuation for complex frequency - Complex frequency is a specialist topic, see explanation
in Section 4.4.1.
Determinant - This displays the determinant which is the result of the evaluation of the char-
acteristic function. The modes exist at values of the frequency, real wavenumber and atten-
uation for which the determinant is zero. Therefore if you plot the determinant it can help
to see where roots may exist - this is really a diagnostic aid if modes are proving hard to
find; in most cases Disperse finds the roots automatically and there is absolutely no need to
plot the determinant. The determinant is in fact a complex quantity itself, even when the
solution has no attenuation, so you have the choices of real, imaginary, absolute or phase to
look at.
Angle of incidence - Disperse allows the user to display the dispersion curves in terms of the
angle of an incoming wave that that would be necessary to excite the guided mode. If the
semi-infinite half-spaces are not vacuum, then the angle selected by Inc. Angle (dft) is that
of a longitudinal wave incident on the structure from the top half space. If the half-spaces
are vacuum, then Disperse assumes the angle for a longitudinal wave in a perspex (lucite)
wedge. You can also choose the material to be used for the angle calculation: Perspex uses
the longitudinal wave in perspex, Air uses the longitudinal wave in air, Water uses the
longitudinal wave in water, Vl and Vs use respectively the longitudinal and shear waves in
the top half-space.
The angle of incidence plots are mainly used to assist the generation of guided waves. The
plot displays the angle at which an incident wave should insonify the system in order to
generate a guided wave, as commonly done using perspex wedges or local water immersion.
The value is calculated by simply applying Snell’s law so that the wave number components
of the incident wave and the desired guided mode are equal in the direction parallel to the
interface. This wave number matching can be expressed as,
 
−1 vinc
angle = sin (5.1)
vph

where vinc is the velocity of the incident bulk wave and vph is the phase velocity of the
guided wave.
Group slowness - This is the inverse of the group velocity. This quantity can be convenient to
indicate the time that it takes for a guided wave packet to travel to a given location.
Phase slowness - This is the inverse of the phase velocity.
Function result - You can choose, as one of the axes, the results of a calculated function. First,
of course, you need to calculate the function: this is accessed either by selecting Display
: Derived data or Tools : Calculate function; both bring up the same dialogue box,
although with different starting defaults. For details of how to calculate a function, see
Section 6.3. This increases enormously the possibilities for plotting. So if you want to plot
your results using some complicated axis function in order to compare with a graph in the
literature, there is no need for tedious calculations; just use this versatile facility.
Energy velocity - This is only available in the pulldown list once you have calculated the energy
velocity. The energy velocity is identical to the group velocity if there is no attenuation, and
the proper alternative instead of the group velocity if there is attenuation; see the discussion
in the earlier Section 5.2.2. Calculation of the energy velocity requires an integration of
various mode shape quantities across the thickness of the structure and is quite laborious.
It is offered as one of the standard function calculations, see the Section 6.3, accessed either
by Display : Derived data or by Tools : Calculate function.

Disperse version 2.0.20a


5.3 Other Forms of Display 56

Additional information can be added to the plots by specifying a Colour axis. This is again set
using a pull-down menu with a list of options. This causes the curves to be coloured according to
the colour axis choice. Each point on each curve is then given a colour which depends on the value
of the chosen parameter at that point. Thus for example you can display phase velocity with a
colour scale showing the attenuation of the modes.

5.3.3 Hide Curves

This feature, selected by Display : Hide curves, allows you to limit the display to a subset of
the calculated modes. The modes which are not shown are not deleted, but are simply not shown.
The options are as follows:

Hide next selected - After selecting this, you simply click with the left mouse button on each
of the modes which you want to hide. You can also choose this menu item from the mouse
right button menu Edit curves : Hide curve.
Hide all but selected - First click with the left mouse button on the mode you want to keep in
the display - this selects it. Then select this option and all the other modes will be hidden.
Show all - This returns all hidden modes, roots and minima to the visible state.
Show only modes - This hides the minima, roots (these are the small circles, see 4.3) and bulk
velocities so that only the modes are plotted.
Select types to show - This brings up a window where you can explicitly select items to show
or not to show. Most are self-evident. Native modes are modes which were created in the
current data set, that is to say, in the usual way. This name is to distinguish them from
Foreign modes; these are modes which have been copied and pasted from another data set.
This is easily done, for example so that curves from different analyses can be compared on
the same plot - see Section 6.7.4 for details. You can delete all the foreign modes by selecting
Edit : Delete : Foreign modes.
Hide other orders/symmetry - Only relevant if the set of curves is separable into subsets of
different cylindrical orders or symmetric and antisymmetric modes. First click with the left
mouse button on a mode of the type you want to keep in the display. Then select this option
and all the modes which are not of this type will be hidden.

5.3.4 Derived Data

This opens a powerful utility which allows you to define a function which you may then apply to
the dispersion curves which you have solved. The utility can be accessed either from this menu
item or by choosing Tools : Calculate function from the Tools menu. Both routes bring up
the same dialogue box but start with different default menu choices. The default when called from
this Display menu is simply to display the value of an item of derived data. The use of this utility
is described with the other tools in Section 6.3.

5.3.5 Sweep Results

This is really a diagnostic aid if modes are proving hard to find; it should not normally be necessary
to use it. After you have performed a sweep (see Section 4.3), this command will plot the absolute

Disperse version 2.0.20a


5.4 Using the Display 57

value of the (complex) characteristic function along the line of the sweep. The modes exist at
locations where the determinant is zero, but remember that the frequency, real wavenumber(or
velocity) and attenuation must all have appropriate values for the existence of a root, therefore a
sweep may only show minima rather than zeroes if the sweep line does not actually pass through
the solution point. You can think of this as a line in the three dimensional solution space which
may pass near a point but does not necessarily pass through it. For more details about the solution
and finding roots, see Sections 7, 4.3, and 4.3.3. Plotting the determinant can help to see where
roots may exist if you are having trouble finding them. Alternative to the absolute value of the
determinant, if you want to plot the real or imaginary parts you can do so by using the Display
: Select axes command, see Section 5.3.2.

5.4 Using the Display

This section covers several items relating to the the use of the display, including the menu items
Display : Custom zoom, Display : Unzoom and Display : Autoscale, some hot key
commands, some actions which can be accessed using the right mouse button, and relevant display
items on the toolbar. When using the toolbar you might find it helpful to keep an eye on the
bottom left of the Disperse window before clicking on a tool button: you will see there a text hint
for the button.

5.4.1 Zooming and Moving

Disperse provides several ways to navigate about the screen and change the dimensions of what
is plotted. Note that some of these choices change the function of the mouse, and accordingly
the shape of the pointer on the screen. These changes stay in effect until they are redefined by
another choice. You can return to the default pointer by selecting Pointer from the right mouse
button menu.

AutoScaling - AutoScaling is the standard way to set the plot limits. The program does ev-
erything for you; it calculates the minimum and maximum values and sets the graph limits
accordingly. Disperse uses this by default unless you choose to set the scales yourself, eg by
zooming. You can return to this option by selecting Display : Autoscale or by selecting
Autoscale from the menu attached to the right mouse button or by pressing the F4 key.
Note that if you have traced backwards towards zero wavenumber, then the non fundamental
modes will have very high phase velocities at their cut-off frequencies and the phase velocity
plots will be very uninteresting. Therefore the Autoscale function chooses a display cut-off
of 12 km/s when plotting the phase velocity.

Custom zoom - It is also possible to set specific window values, by choosing Display : Custom
zoom, or selecting from the right mouse button Actions : Custom zoom. In this way
the minimum and maximum values of the X and Y axes can be individually set. Optionally
you can also choose on this dialogue box to leave the y axis to be scaled automatically while
setting specific values for the x axis.

Zoom - To enter the ’zoom’ mode, select Zoom from the popup menu attached to the right
mouse button or press the F2 key. When in the ’zoom’ mode, the pointer will appear as a
magnifying glass. Then click and drag with the left mouse button to outline an area that
will become the new x-y graph limits.

Disperse version 2.0.20a


5.4 Using the Display 58

Note 1. If the new area is less that 10 pixels big, Disperse will move to the geometrically
closest point and show the values at that point.
Note 2. The zoom window must begin within the plot limits, however the end point
may extend beyond the graph borders. When the end point is outside the
previous graph limits, the new limits will also be outside the previous limits.

Unzoom - Choose Display : Unzoom or press the F3 key to return to the previous zoom
position.

Pointer - Choose Pointer from the right mouse button menu or press F5 to return from one
of the pointer actions such as Zoom to the default mode of the pointer. This will show a
conventional arrow rather than the special pointers shown for the various functions.
Moving - You can nudge the zoom box by l0 percent in any direction simply by holding down
the Control key and then pressing any of the Arrow keys. This option is helpful when you
have zoomed in on a section and would like to move to a nearby new section of the curve
without AutoScaling and re-zooming in a different area.

5.4.2 Using the Pointer in the Display Window

The default role of the mouse pointer is the Pointer mode; this displays the conventional arrow
on the screen. When used in the display window, the Pointer is used to select modes or points
on modes, as will be explained in this section.

If the pointer has previously been set to perform a separate function, such as zooming or editing,
then as a matter of habit it should afterwards be reset to pointer, by selecting Pointer from the
right mouse button menu. If this is not done then you may inadvertently continue to perform the
other function when all you want to do is select a point or mode.

To select a curve - Curves need to be selected for various reasons including colouring, editing,
and individual copying. Simply click with the left button anywhere on or near the curve.
To select a point on a curve - The main reason for pointing at a particular point on a curve
is to see numerical values for the point. A second reason is for editing, see Section 5.4.3.
Selection of a point is identical to selection of a curve except that you click on or near
the point. Disperse always jumps to the nearest point to the location where you click.
When a point has been selected then Disperse displays the numerical values for the point
in two places. The display immediately above the graph shows the name of the curve, if
named, and the values corresponding to the current x and y axes. The display right at
the bottom of the Disperse window shows the name and all of the key parameters of the
mode. The number in square brackets is the number of the point: each curve is stored as
a list of sequentially numbered points starting with point number 0. Disperse attempts to
name curves automatically with the conventional names in the simple systems such as Lamb
modes, bar modes or pipe modes. In most cases it gets these right but you should be aware
that because of the wide range of possibilities with different materials and dimensions this
naming is not rigorous and should be taken simply as a starting attempt. You can change
these names or introduce new names for unnamed modes by using the Tools : Labels :
Edit mode names menu, see Section 6.8.
To move the pointer quickly along a curve - Simply hold the left button down and drag the
pointer along the curve. The numerical display is updated continuously.

Disperse version 2.0.20a


5.4 Using the Display 59

To move the pointer one point at a time - Once you have clicked on a point on a curve, you
can move along the curve one point at a time by holding down the Shift key and pressing
the right or left Arrow key. The right key moves in the increasing point number direction,
the left key in the decreasing direction. Note that these are not necessarily the right and
left directions on the screen.

5.4.3 Toolbar and Edit Options

The purpose of editing is to modify attributes of the dispersion curves. You can simply change
the colour, or the label, you can delete individual points or whole curves, or you can split curves
into two parts or join two curves into one.

Change curve colour indexCurve!Colour - This changes the colour of the chosen curve as it
appears in the graphical display on the screen. This option is helpful to distinguish various
modes as you switch between different formats of displays. Choose a colour from the pull
down menu at the top right of the toolbar, then either select from the right button menu
Edit curves : Change colour or click on the paint brush button on the toolbar (this is
the second button to the left of the colour pull down menu). Now when you move the mouse
over the plot area the pointer appears as a paint brush and any curve which you select is
immediately changed to you chosen colour. This function remains active until you select
another or return to Pointer mode.
Change all curves colour - To change the colour of all of the modes to a single colour, choose
the colour from the colour pull down menu on the toolbar, then click the button immediately
to the left of that menu.
Create a new label - You can display text labels of your choice next to the curves. The toolbar
button to activate this shows a picture of a label. The description of how to use the facility
is given is Section 6.8.

Delete an entire mode - This deletes an entire curve. Simply click on the Eraser button on
the toolbar then click on the mode or modes which you wish to delete. Remember this
function remains active until you select another or return to Pointer mode! You can also
delete modes using the Edit : Delete menu; this allows you to delete the next mode you
click or the last mode you calculated or the currently selected mode. If you delete a mode
by mistake then this menu also allows you to undelete it - but note there is only one level
of undelete; it works just on the most recent deletion and it should be applied immediately
after making the mistake.
Delete a point on a mode - This deletes a single point on the curve. This can be useful if
there is a spurious point, for example where crossing a bulk velocity. Click on the Delete
point button on the toolbar - this shows a small eraser on a red line and is immediately to
the right of the Erase mode button. Now click on the point or points which you want to
delete. Remember this function remains active until you select another or return to Pointer
mode! You can also access this function, or undelete the most recently deleted point, via
the Edit : Delete menu.

Split one mode into two - This function breaks a curve into two portions. The most common
reason to want to do this is to delete a large part of a curve: you simply divide the curve into
two and then delete the unwanted part. Click on the Split button on the toolbar (picture
showing square brackets and arrows pointing outwards), or select from the right button menu
Edit curves : Split, then click on the point where you want the split.

Disperse version 2.0.20a


5.5 Preferences 60

Combine two modes into one - This function joins two mode together. This is useful if you
traced the mode in two parts and then want to joint the two parts into a single curve.
Click on the Combine button on the toolbar (picture showing square brackets and arrows
pointing inwards), or select from the right button menu Edit curves : Combine. Now
click and hold down the left button on the point at the end of one curve. Now, keeping the
button down, drag the mouse until it points at the end of the other curve, now let go. The
curves will be combined immediately into a single curve.

5.5 Preferences

Choosing Display : Preferences allows you to modify some of the settings for printing, screen
display and exporting of data. The settings which you choose are saved in the computer registry
and so remain in force for future sessions by the same user.

The first group of options control primarily how Disperse sends your results to the printer:

Print colour - Disperse is initially configured to use a black and white printer. Therefore, all of
the solid lines are printed in black and the screen colour scale is converted into a linear gray
scale. However, by changing this option, Disperse can print in gray scale (where all of the
colours (including solid colour lines) are directly converted into their gray scale equivalents),
simply all in black, or in full colour.
Clipboard colour - This gives control separately for the colouring of the curves when they are
copied onto the clipboard. the options are the same as for the previous item. See Section
6.7.4 for details on copying and pasting.
Print line width - This allows some control over the width of the lines as they appear on the
printed graphs.
Geometry notes - This determines whether the information about the structure (the layers and
their material properties) should be included or not on the printout.
Orientation - The printed curves can be in portrait or landscape orientation on the page, ac-
cording to this choice.

The second group of options can be used to set preferences for the graphical screen display:

Colour axis - This options sets the colour scheme that is used whenever a colour axis is defined,
see Section 5.3.2. Each segment of the curve is coloured according to its value.
Attenuation type - This determines whether by default the attenuation axis should be displayed
using a dB/m scale or a nepers/m scale. This default is used when you choose to plot
the attenuation using the menu Display : Attenuation or Display : Second graph :
Attenuation. Of course having set the default you can still choose specifically to plot the
attenuation in any of the alternative forms by using the Display : Select axes options
(details in Section 5.3.2). The two default scale options are related as follows: 1 dB = 0.115
nepers, 1 neper = 8.69 dB; for further details of these units see Section 5.2.3
Axis units - The default units for plotting velocity axes can be set here. This will be used for
all velocity scales unless you choose a specific scale with different units using the Display :
Select axes options (details in Section 5.3.2)

Disperse version 2.0.20a


5.5 Preferences 61

Axis scaling - It is well known that the frequency axis of the Lamb wave dispersion curves for a
flat plate can be scaled by the thickness so that the results for a particular thickness of plate
can be used for any other thickness. Thus a point at 1 MHz on a curve for a 1 mm thick plate
is also correct in every way (including its mode shapes) for 0.5 MHz with a 2 mm thick plate.
It therefore often makes good sense to plot the frequency axis scale as frequency-thickness,
often referred to as “fd”. If fd is chosen then Disperse automatically re-scales in this way,
taking into account the thickness of the layer. If the plate consists of multiple layers then it
can still be valid to perform such a scale. However the limitation is that the results remain
valid only for equal scaling of the thickness of each of the layers. For example a 1 mm thick
layer of copper on a 3 mm thick layer of steel gives valid results also for a 2 mm thick layer
of copper on a 6 mm thick layer of steel, simply by scaling the frequency by a factor of 2;
conversely the results for the 1 mm thick layer of copper on 3 mm of steel cannot be used
for a thicker or thinner layer of copper on the same layer of steel. When the plate consists
of multiple layers Disperse uses the total thickness of the plate to perform the scaling. You
may also choose your own scale factor: select Thickness*value, then enter your scale factor
in the box immediately below. Please note that Disperse only applies these scaling choices
to the results from allowable geometries - for example f*d scaling cannot be applied when
the structure is cylindrical.

The third group of options sets the parameters for exporting data (6.7.2) in text form to be read
by other programs, and also for copying text data (6.7.4) onto the clipboard using Edit : Copy
or Edit : Copy all:

Title header - If No text headers is selected then the text file (or clipboard) starts with
an integer which is the number of lines of the data which follow. Immediately after this
the numerical data is listed without any other descriptive information. Only axis titles
additionally puts a text title at the top of each column of data to say what it is. Include
text headers (default) further adds the mode names on the first line of the file.
Point header - Include point count (default) puts the number of lines of data which follow
at the start of the data (see item above). No point count omits this. Thus if No point
count is selected here and No text headers is selected from the previous item, then the
text file contains only data, no additional text information. This can be useful if you want
to import the data into a program which cannot handle the text information.
Layout - If more than one curve is being exported then Side by side puts the data side by side
in the file; alternatively Consecutive puts the curves one after the other in the file. This is
only relevant to exporting to file - copying to clipboard is always side by side.
Selection - All points exports all of the points of the curve or of all curves; Points on graph
exports only those which appear in the current display window.

Disperse version 2.0.20a


Chapter 6

Tools for Processing the Results

Disperse provides several tools to help with the interpretation of the results. These include: the
display of the mode shapes, that is the distributions of displacements and stresses through the
thickness of the structure; the simulation of wave packet signals to show the effects of dispersion
as the waves travel; the possibility to calculate user-defined functions using the results, thereby
extracting all sorts of detailed information from the solutions. This chapter also describes the
options for data transfer into or out of the program: saving, exporting, printing, copying and
pasting.

6.1 Mode Shapes

The mode shapes are the distributions of the field quantities (displacements, stresses, power flow
etc.) through the thickness of the structure. These are properties of the waves just as much as
are the velocity/frequency/attenuation dispersion curves. In fact the mode shapes are calculated
from the same equations as are used for calculating the dispersion curves, and a mode shape
applies strictly to a specific position on a curve. Details of this theory are given in Chapter 7 and
particular details about mode shapes are in Section 7.10.

Mode shapes are very useful for improving understanding of the nature of a wave and how best to
exploit it. All users are strongly recommended to make extensive use of this facility. For example,
in the application of NDE (non destructive evaluation), it is useful to determine the depths in the
plate where the wave has significant stresses since these will be the depths where it has greatest
sensitivity to defects. Even more basic, as an example for beginners, the extensional nature of the
S0 mode can easily be understood by looking at the distribution of the displacements through the
thickness of the plate. Or try plotting the displacements of the Rayleigh wave to show clearly how
it is localized next to the surface of the plate. A particularly helpful form of display for visualizing
the waves is the animated grid display - in our experience, even the most advanced users make
use of this. You can, for example, show the A0 waves leaking from a plate which is immersed in
water and thus see how the attenuation, which is caused by the leakage, varies as you move along
the dispersion curve. Details of how to run that specific example are given later in this section.

Some very important points regarding mode shapes must be understood:

1. Mode shapes naturally have arbitrary scales. Disperse never deals with guided waves of

Disperse version 2.0.20a


6.1 Mode Shapes 63

particular amplitudes; its function is to evaluate the conditions for the existence and the
properties of waves which can propagate in a structure. This of course includes the dispersion
curves and the shapes of the displacement and stress distributions, but it does not include
the amplitudes of any of the quantities. To study amplitudes, one would have to consider
also the strength and manner of the excitation which is used to generate the wave. Therefore
mode shapes must be regarded strictly as shapes or distributions; a value on a mode shape
plot can only be used to compare with another value relating to the same mode shape, for
example at another point on the curve or in another of the field components.
2. The mode shapes vary as you move along the modes. For example the S0 mode has simple
extensional behaviour at very low frequency but its shape changes gradually as the frequency
is increased until ultimately it becomes the Rayleigh wave. Therefore the mode shapes are
properties, not just of a mode, but of a particular point on a mode. With reference to the
previous note, it follows that the values on the mode shape curves cannot be compared from
one location on a curve to another - only the SHAPES can be compared.
3. Therefore, with regard to these two previous notes, Disperse has to adopt some arbitrary
scale when plotting the mode shapes, yet one which has consistent characteristics so that
there are no sudden jumps in the scale as you move along a dispersion curve. With this in
mind we have chosen to normalize the mode shapes to the power flow in the direction of the
wave propagation. Thus all of the displacement and stress values which are shown on the plot
are those for a wave which has unit power flow along the Z direction for a plate or cylindrical
structure, or the q (Θ) direction for circumferential waves. The power flow in a plate is per
unit (m) distance in the direction normal to the plane of calculation (the Y direction);
for modes along a pipe it is for the whole circumference of the pipe; for circumferential
waves it is per unit (m) distance in the direction normal to the plane of calculation (the Z
direction). However there are currently two exceptions to this normalisation: (1) the power
flow calculation is approximate but not strict for structures which are embedded or immersed
- see Section 7.10), (2) values in the Grid display are not normalised.

The mode shape module is activated by clicking on the Mode shapes tool bar button, which is a
picture of a cantilever and is immediately to the left of the STOP button, or by selecting Tools :
Mode Shapes. The mode shape module opens another window, while still leaving the dispersion
curve window active.

The mode shape display shows the mode shape for whatever is the current position of the pointer
in the active dispersion curve display. Therefore to plot a mode shape, you need to open the mode
shape module and then move the mouse over to the dispersion curves and select a point on a curve
by clicking with the left mouse button. The mode shape display will immediately update to show
the shapes for that particular point. Furthermore, you can observe the variation of the shapes
along a mode by “sliding” the pointer along the dispersion curve. Click on a point with the left
mouse button and hold the button down, then move the pointer along the curve. You will see
that the selected point and the mode shapes are updated continuously.

Mode shapes may be printed using the File : Print function (Printing is described in Section
6.7.5. The mode shapes information may also be copied to the clipboard for exporting by using
the functions on the right mouse button, as described in Section 6.7.4.

Some general notes about mode shapes:

• It is possible for a mode shape to be all zeros, for example the normal displacement of an
SH type mode.

Disperse version 2.0.20a


6.1 Mode Shapes 64

• Sometimes, for example when calculating zero order torsional modes which are decoupled
from the longitudinal and flexural modes, some of the mode shape components which should
be zero will not be identically zero, but only approximately zero.
• Displaying mode shapes at frequencies above 90 MHz-mm can cause numerical problems
which may result in crashing of the program. In general, when the dispersion curves can
be calculated using the analytical routines rather than the global matrix method, the mode
shape solutions are stable to much higher frequencies. The analytical routines are discussed
in Sections 4.4.1 and 7.7.
• The accuracy of the mode shape calculations depends on the accuracy of the solution of the
dispersion curves. If the dispersion curves have not been solved to a high degree of accuracy,
then errors may be evident in the mode shapes. For example, a normal or shear stress value
may not be exactly zero at a free surface. To improve the accuracy of the mode shapes, you
must first calculate the dispersion curves using tighter tolerances. These can be set in the
Trace : Convergence parameters menu, see Section 4.3.3.

The mode shapes module shows a split window with the shapes on the right hand side and the
choices for the display on the left hand side. The choices are set out on three cards: Lines, Grid
and Power. Each of these will be explained in turn in the following paragraphs.

6.1.1 Lines

This option simply displays graphs of selected field quantities across the thickness of the plate.
Identify the quantity or quantities which you want to plot from the list in the left hand window,
and you will see them plotted in the graphical display. The top of the display is the top surface of
the plate or the inner radius of the cylinder and the bottom of the display is the bottom surface
or outer radius. The amplitude is positive to the right. Any number of the field quantities can
be shown at once. The location of the mode shape on the dispersion curve is always displayed by
text information beneath the graph window. The axis directions for the components are: x (or r)
is the direction normal to the plate surface (or the radial direction for a cylinder), y (or q) is the
direction normal to the plane of the calculations (or the circumferential direction for a cylinder),
z is the direction of the propagation of the wave (in the case of a cylinder this is the axis of the
cylinder). These are illustrated in Figures 7.1 demonstrates the construction of a five layer flat
plate system and Figure 7.2.

The phases of the mode shapes are of course arbitrary. However the phase differences between
components can be useful information, for example the two in-plane displacement components
of (strict) Lamb waves differ in phase by 90 degrees. There can also be phase variation of a
component across the thickness of the structure if the mode has attenuation. The rules which
Disperse uses for plotting the mode shapes takes the phases into account as follows:

• The in-plane displacement (Z direction) at the 2nd plotting point from the top surface of the
first layer is taken as the reference and is defined to have zero phase (the 2nd point is taken
because sometimes the surface value may be zero). Thus its plotted value is its magnitude
and it lies on the positive real axis of the complex plane. The Z displacements for all other
points through the thickness are then plotted with this phase. In the case of strict Lamb
waves all of these other points will also lie on the real axis (0 or 180 degree phase). In cases
of attenuating modes, these other points may not be on the real axis but may differ in phase
from the first point; the plot therefore only shows their real parts, that is their projections
onto the real (zero phase) axis.

Disperse version 2.0.20a


6.1 Mode Shapes 65

• Each of the other components of stress, displacement, or strain is treated in a similar manner,
so that all components are plotted with phases such that the point on the top surface lies on
the real axis (0 or 180 degree phase). All of the plots therefore show the shapes of modes in
a consistent manner, however in this format the phase differences between the components
are not seen.

• All of the phase information, that is the phase differences between modes and any phase
variation through the thickness, can be seen by choosing Display as phase. This is done
either by selecting the box at the end of the list of components, which has this label, or via the
options in the Advanced menu which is reached via the Grid card and will be described
shortly. When this is set then all of the mode shape components which are selected are
plotted as phases. Each point on each phase plot shows the phase delay in degrees with
respect to the reference (the Z displacement at the top surface).
• The energy and power flow components are plotted simply as real values(i.e. on the real
axis). They are calculated using integrals over a complete cycle, and so there is no phase
information to be considered.

The display shows the graphs in different colours to aid recognition. You can see specifically which
component is being displayed by a graph, and the numerical values of the graph, by clicking on
the curve with the left mouse button. This will bring up text information above the graph. You
can move the mouse along the curve to pick off values at any position. If you want to take the
values from the graph to use in another program then the easiest way is to click on the graph,
then choose from the right mouse button menu Actions : Copy text. This will put the x-y
graph information in text form on the clipboard. Further information about exporting is given in
Section 6.7.4.

The points on the mode shapes are of course calculated at discrete positions through the thickness.
You can change the number of points through the thickness by using the Advanced menu which
is reached via the Grid card and will be described in the following paragraph for the Grid options.

Another useful option, in cases where the structure is immersed or embedded, is to show the waves
in the half spaces. This is also accessed via the Advanced menu.

Most of the field quantities are displacement, stress and strain components and need no further
explanation. The units of these quantities are all in Newtons, meters, seconds. Explanation of the
less obvious quantities follows:

Strain energy density - This is the strain energy per unit volume which of course varies from
position to position through the thickness. It represents the stored energy given by the
products of the stress and strain components. This quantity is cyclic in time; the display
shows the value integrated over one cycle. The definition is given in Section 7.10.
Kinetic energy density - This is the kinetic energy per unit volume, representing motion of
the material. It is calculated from the particle velocities which in turn are derived from the
displacements and the frequency, as defined in Section 7.10. It is cyclic in time and so again
the integral over one cycle is shown.
Total energy density - The energy at a point in the material is composed of a combination of
strain energy and kinetic energy. However the addition must take into account the differences
in phase between these two contributions. The strain energy and the kinetic energy are each
cyclic in time, but the total energy density is not. If there is no attenuation then the total
energy density is constant in time; if there is attenuation then it decreases exponentially.

Disperse version 2.0.20a


6.1 Mode Shapes 66

Power flow - The power flow of a wave is the rate at which it propagates energy along a par-
ticular direction. Disperse calculates the power flow per unit area at each point through
the thickness of the structure; this is given by the Poynting vector and it has a direction
as well as a value. Details are given in 7.10. If you want to see a plot of the vectors with
their directions then you can get one using the Power card, see later in this section. The
choices for the line plots here are the resolved vectors in the axial (Z), radial (R), or circum-
ferential (q) directions. Note that if there is no leakage or damping in the system then the
power flows entirely in the direction of the wave propagation and is zero in the other two
directions. Remember also that the power flow, like all other mode shape components, is
scaled to an arbitrary level, as discussed earlier in this section; the actual amplitude of the
power in practice will of course depend on the strength of the excitation. The scale which
Disperse uses for normalization is in fact the power flow itself: the amplitudes of all of the
mode shape components which it plots are such that if you resolve the power flow vectors
into the propagation direction of the wave, and integrate them through the section of the
structure, then the result is 1.0. However, beware of the scaling of waves which travel along
cylindrical structures. The power flow which is shown in the mode shape plots (and in the
Tools : Calculate function facility) is expressed per radian of circumference, whereas the
power normalisation is based on the whole circumference. Thus to get the integrated power
flow result of 1.0, it is necessary to integrate through the thickness of the pipe and then
multiply the result by 2 ∗ Π.
Display as phase - The displacements, stresses and strains vary harmonically both in time and
space. They therefore have both amplitudes and phases. By default the line plots show the
amplitude information; in fact they assign zero phase to the first point at the top layer of
the structure when plotting each component. But there can be a problem with this because
such plots do not show phase differences between each of the components. For example the
displacements in the X and Z directions in a Lamb wave differ by 90 degrees in phase. This
option allows you to see these phases instead of the amplitudes. The phase information
which is plotted is the phase angle of the component at the first point at the top layer of the
structure. Note that if there is no attenuation then phases do not vary through the thickness
(except for changes of sign, i.e. 180 degree shifts); thus the wavefronts are always normal to
the Z direction. However if there is attenuation then you will see drifts of phases across the
thickness.
Display in half-spaces - If the top or bottom semi-infinite half-space is not vacuum, Disperse
allows you to show the mode shape in a portion of this half space. See Advanced menu
which is reached via the Grid card in following paragraphs.

6.1.2 Grid

The grid display does not give any numerical information but it is very useful for visualizing the
waves. Therefore it is recommended for regular use, particularly for new users. It uses just the
displacement components of the mode shapes to plot a deformed shape of the structure; this can
be viewed either stationary or animated as if a wave was passing along the structure. A couple of
notes regarding the grid displays:

• The scale of the displacements is of course greatly exaggerated in order to see the displaced
shapes. Therefore the amplitudes do not relate to the power normalized modes in the way
that all of the line plots do.
• The grid itself is only present to help the display. It has no significance in the calculations -
i.e. it does not signify any kind of discretization in the solution.

Disperse version 2.0.20a


6.1 Mode Shapes 67

To illustrate the use of the grid display, we start with an interesting example. Open the example
file plate-steel-in-water.ds2; this is for the leaky Lamb modes in a steel plate immersed in water.
Select Mode shapes and the Grid card. Click on the A0 mode at about 0.7 MHz. Click the
Animate box. You will see a moving image of the bending motion of this mode. Now, for an
even more interesting image, we will look at the waves leaking into the water. Click Advanced,
then choose Display in half spaces from the second pull-down menu and enter the value 5 in
the Half space!Display box just below. Click OK.

The following paragraphs describe the options for the grid display. Note that the Advanced
options may be used to change the settings for the Lines plots as well.

Plane of display - This pull-down menu only appears if the geometry is cylindrical. It simply
offers you a choice of the plane in which you want to plot the mode shapes. The RZ plane
shows you a view from the side of the cylinder so that the vertical direction on the screen is
the R coordinate (positive upwards) and the horizontal direction is the Z coordinate. The
axis of the cylinder thus lies below the image. The RQ plane gives you a three dimensional
perspective view of the cylinder. The RQ plane (sin soln) plots the shapes using sine and
cosine functions, showing the non-zero-order modes to spiral around the pipe. The RQ
plane (exp soln) plots the shapes using complex exponential functions. These describe the
superposition of two identical waves which spiral in opposite directions around the pipe;
the summation of the two opposite spirals have a cancelling effect such that the image is no
longer a spiral. These forms of display are both equally valid, although it may be argued that
the single spiral is a more fundamental entity than the double spiral. For plate structures,
the view is always the XZ plane, that is to say the view from the side of the plate. The top
of the plate is at the top of the image and the wave travels in the horizontal direction on the
screen from the left to the right.
Show undeformed - When this toggle is selected, the mesh in its undeformed position is drawn
in colour behind the deformed mesh. If Disperse is configured to print in black and white (see
the Display : Preferences menu (5.5), you will usually want to not show the undeformed
grid because it can make the printed copy appear confusing.

Animate - This toggle controls whether the grid mode shape should be animated. If selected
the mode shape display is updated over a series of time steps in order to show the motion
of the wave. Note that displaying animated mode shapes can use a lot of processing power.
Therefore selecting different points on a dispersion curve when animated mode shapes are
active can sometimes cause the program to behave erratically.

Grid colour - This pull-down menu allows you to choose between three options for the colour
of the undeformed and deformed grids. If you choose to display the deformed grid in colour
then a red-blue colour scale is used such that the colour varies according to the amplitude
of the displacement.
Amplitude of deformed grid - This slider bar and text box control control the scale of the
displacements in the deformed grid.
Constant wavelength or distance - If constant wavelength is selected then the same number
of wavelengths is shown on the display whatever the position on the dispersion curves. If
constant distance is selected then the display shows a constant distance of the structure,
regardless of the wavelength.

The Advanced button opens a window with some further settings for the grid and line mode
shapes. These options are described as follows.

Disperse version 2.0.20a


6.1 Mode Shapes 68

Display as amplitude or phase - This option applies only to the Lines mode shapes. It sets
the display to show either the amplitude or the phase of the components which are being
plotted. Setting the phase here has the same effect as clicking the Phase box at the bottom
of the Lines menu. See discussion of amplitude and phase of Line plots earlier in this
section.
Points per layer - This parameter controls the number of points which Disperse uses through
the thickness of each layer in order to generate the mode shape plot. The first and last
points on the layer are of course at the top and bottom surfaces, so the number of intervals
between points is this number minus one. For example if you set this number to 101 then
the plate thickness will be divided into 100 intervals for the plot. The maximum number
of points added up for all of the layers must not exceed 1000. Note also that Disperse may
change this value automatically if it is not an integer multiple of the grid size (described
below).
Animation delay - This applies only to animated mode shapes. It sets the time interval between
frames of the animation, in units of milliseconds. For example if you set this to 1000 then
each frame will be shown for 1 second before updating to the next.
Time step - This applies only to animated mode shapes. The mode shape is incremented by this
proportion of a cycle with each frame. For example if you set this to 0.2 then each frame
will advance 0.2 of a cycle, so it will take 5 frames to describe the whole cycle.
Grid line every... - This value is used to specify the density of the grid. The grid is drawn with
a horizontal line every n points through the thickness of the layer. Thus for example if you
have specified 101 points through the layer thickness (see earlier item Points per layer),
and now you specify Grid line every 20 points, then the display will show 5 grid intervals
through the thickness of the layer. Note that 101 points through the layer gives 100 intervals
between points and so is a convenient number for the division. After you have chosen this
parameter, Disperse automatically chooses the grid density in the horizontal direction such
that grid elements are approximately square.
Number of cycles - The number of cycles simply sets how many cycles of the wave should be
shown by the deformed grid when Constant wavelength has been selected.
Display in half spaces or just finite layers - If the structure is surrounded by fluid or solid
(i.e. it is not bounded by vacuum) then in general the mode involves displacements and
stresses in the surrounding medium (the half spaces). By default Disperse only plots the
mode shapes in the finite thickness layers of the structure, but this option sets the plot to
include also the shapes in part of the half spaces. If you choose this then you should also
consider how much of the half space to show; see the next item.
Half space display - This only applies if you have chosen to show the mode shapes in the half
spaces, see previous item. Since the half spaces above and below a plate and outside a
cylinder are infinite in extent, it is necessary to define how much of the half space to show.
The value entered here specifies the ratio of the amount of the half space which is shown to
the total thickness of the structural layers. Thus for example if you have a 1 mm thick plate
immersed in water, then a value of 5 will display 5 mm of water on each side of the plate.
Partial wave - As discussed earlier in this section, mode shapes are properties of the system and
as such their amplitudes are arbitrary. When calculating the mode shapes, Disperse starts by
assuming unit amplitude of one of the partial waves in the system, let us say the “reference”
partial wave. Once this wave is assumed, it is possible to solve for the amplitudes and phases
of all of the other partial waves with respect to this wave (for explanation of partial waves
see Chapter 7 and of application to mode shape calculation see Section 7.10). Disperse then
calculates the power flow (see earlier discussion in this section) and scales all of the partial
wave amplitudes, displacements, stresses etc. to give unit power flow. By default Disperse

Disperse version 2.0.20a


6.1 Mode Shapes 69

chooses the reference partial wave to be the downward travelling longitudinal partial wave
(L+) in the first finite thickness layer (or the half space in the special case of a =Rayleigh
wave), except in cases where the wave motion is entirely normal to the solution plane (eg
Love waves in a plate or the fundamental torsional mode in a pipe) when the downward
shear horizontal partial wave (SH+) is chosen. This menu item allows you to alter that
choice to any other of the partial waves; you can also specify which layer it is in in the
text box immediately beneath. The labelling system for the partial waves is shown for flat
plate systems in Figure 7.1 and for cylindrical systems in Figure 7.2. In theory it makes
no difference whatsoever which partial wave is chosen as the reference because their inter-
relation is a property of the system and all are scaled for unit power flow. In practice also it
is extremely unusual to have to change this. However there can be circumstances when there
are numerical difficulties because of this choice. This can arise if the reference partial wave
happens to play very little part in the solution, in which case there can be poor numerical
conditioning. It is suggested that the default should always be used unless you have reason
to suspect that this choice is causing problems, such as noisy mode shape plots.

Finally, the Export Movie button writes a movie file (.avi) of an animation of the current mode
shape. You can then use this in other programs, for example it can easily be imported into a
Powerpoint presentation.

6.1.3 Power

The final card of the mode shapes window displays two kinds of vectors relating to the mode
shapes in the structure: power flow and partial waves. The meanings of these have been described
in recent paragraphs in this section; the description here will therefore be limited just to the use
of these options. As with all the mode shape plots, the display relates to the current location of
the pointer in the dispersion curves; this is shown in text at the bottom of the Disperse window.

Power flow - The power flow plot displays the power flow vectors through the thickness of the
structure. The thickness is divided up into a number of equally spaced points and the power
flow vector is calculated and plotted at each. The direction of each line corresponds to the
direction of the power flow vector and the length corresponds to its magnitude.
Partial waves - This item is only available for flat plate geometries. The plot shows a vector
illustration of the partial waves which compose the guided wave. The orientation of the plate
is the same as for the grid mode shapes and as shown in Figure 7.1 - the figure also shows
the labelling system for the partial waves. Click on any of the vectors with the left mouse
button to show more information about it. The text which then appears shows the name of
the partial wave, its amplitude, then the symbol <, then its phase in degrees. The lengths
and directions of the vectors in the plot correspond to the amplitudes and directions of the
partial waves. If the partial wave is evanescent (inhomogeneous) then it is plotted in the
horizontal direction; note that this can result in overlaying vectors in the plot. The phase
angles are arbitrary in an absolute sense; the important information which these values give
is the phase differences between the partial waves. Note that currently the amplitudes of
the partial waves are not power normalised; therefore they can only be compared with each
other and not with the components of the mode shapes.

Disperse version 2.0.20a


6.2 Simulated signal 70

6.2 Simulated signal

It is well known that the shapes of the wave packet signals of dispersive waves change as they travel
- the change is always to spread out the signal as the faster components of the signal separate
from the slower components. This has the detrimental effects both of extending the signal length,
so reducing the spatial resolution, and reducing the amplitude. A useful discussion about these
effects is given in reference [102]. Thus developers of techniques for long range guided wave non
destructive testing are normally keen to use modes with minimum dispersion in order to retain
the shape of the wave packet over long distances. However it is not always possible to avoid the
dispersive regions of curves completely; furthermore, there can be mode conversions of the chosen
mode to other, dispersive, modes so that dispersive signals may often be present. It is therefore
useful to be able to predict the shapes of the wave packets in order to know what to expect with
these dispersive signals. This Disperse tool is implemented in order to make such predictions.

The concept of the tool is simple. The user defines first the Excitation. This consists of the
specification of the mode and frequency, and the selection of one option from a list of possible
types of wave packets. Disperse then shows a graphical prediction of the wave packet as generated
and also after it has propagated a chosen distance. As a second possibility, the user may study
what the signal would like like if some or all of the other modes which can exist at the chosen
frequency are also present; this is done using the Multi-mode options.

Start by selecting from the menu Tools : Simulated signal. A dialog window will appear, with
two tabs. We consider each in turn.

6.2.1 Excitation

The information on the Excitation tab is used to specify the wave packet which is to be generated.
The wave packet is constructed by taking a sinusoidal carrier wave and then multiplying it by a
temporal window function. The mode is chosen by clicking with the left mouse button on any of
the dispersion curves in the set which is currently displayed. The position on the curve also sets
the carrier frequency for the wave packet. The information to be selected is then as follows.

• Frequency This sets the frequency of the carrier wave in the wave packet. This is auto-
matically set when clicking on the mode itself, but additionally it may be set or modified
here.
• Cycles This sets the number of cycles of the signal within the wave packet window.
• Amp mod This sets the shape of the wave packet window which is used to multiply the
carrier signal. There are several choices:
– Rectangular This is the simplest window, comprising just a rectangular shape; thus
the carrier wave is turned on abruptly at the beginning of the window and turned off
at the end of it. This creates the simple form of signal which is generated by many
proprietary tone burst generators. However such a wave packet has significant side lobes
in the frequency domain due to the abrupt start and finish which widen its effective
bandwidth.
– Gaussian The Gaussian function is an ideal window for a narrow band signal, having
the same shape, without side lobes, in the frequency domain.
– Triangular Another simple shape: the window amplitude rises linearly to the maxi-
mum then falls linearly to zero.

Disperse version 2.0.20a


6.2 Simulated signal 71

– Hanning This function consists simply of a single negative cosine cycle with an am-
plitude offset such that the start and end values are zero. This is very similar in shape
and performance to the the Gaussian function although it does have small side lobes.
Its principal attraction is that it has clearly defined locations for the start and finish
(the Gaussian function has exponential variation at the start and finish and so the ends
are not easily defined) and so is easy to implement.
– Taper This a combination of triangular and rectangular functions. It uses a linear rise
at the start of the window, then a constant rectangular part, then a linear fall at the
end. The length of time for the rectangular part is from 30 % of the total time until 70
% of the total time.
– Flat top Gaussian This is similar to the taper window except that the rise and fall
use the Gaussian function instead of the triangular function.
– Flat top sin This is similar to the taper window except that the rise and fall use the
Hanning function instead of the triangular function.
– Flat top ring down This uses a Gaussian function for the rise during the first 10 %
of the total time; then a constant rectangular part until 60 % of the total time, then a
Gaussian fall during the remaining 40 % of the total time.
• Prop dist This sets the propagation distance for the simulation.

Once these parameters have been set, the display to the right of the window shows two signals.
The first signal, starting at zero time, is the initial signal as generated. Thus the form of the
signal can be seen exactly as it is defined, without having travelled any distance and therefore
without any dispersion. The second signal is the signal after the wave packet has traveled the
chosen propagation distance along the plate. This therefore shows the effects of any dispersion
over this distance of travel.

The right mouse button offers some tools to help with the graphical display; these are the same
tools used elsewhere in the program. The Zoom and Autoscale functions are as described
in Section 5.4.1. The Copy options under Actions are as described in Section 6.7.4. Select
Pointer, then left-click at any position on the signal to show the time and amplitude value. The
Edit curves options are present only because this is the general utility function; in this particular
application they serve no useful purpose. The simulated signal plots may also be printed using
the menu function File : Print, as discussed in Section 6.7.5.

The parameters described so far cover the majority use of the function. However there are some
more specialist options available, which are reached by clicking the Advanced button. This brings
up another dialog window. Several of the functions in this window are the same as in the parent
window, but are repeated here for convenience; these are: Frequency, Cycles and Amp mod.
The new parameters are as follows:

• Delay This introduces an initial time delay to the generated signal. It thus follows that all
of the signals are delayed by this amount.

• Max amp By default the generated signal has a maximum amplitude of 1.0 Volt (2.0 Volts
peak to peak). The maximum amplitude may be altered by entering an alternative value
here.
• Sin/Cos By default the carrier signal is a sine function; this option may be used to redefine
it to be a cosine function. The difference between the sine and cosine options is of course
only in the phase of the carrier signal. Each of them is implemented as a function of the
time since the start of the window function.

Disperse version 2.0.20a


6.2 Simulated signal 72

• Frequency modulation This option enables the carrier signal to be modulated in frequency
before applying the window function. The options are:
– Do not modulate frequency This is the default option, in which no frequency mod-
ulation is applied. With this choice the carrier mode remains as a single frequency
continuous sine (or cosine) tone.
– Use frequency modulation This modulates the frequency within the signal envelope
using the window function of time which follows. Thus, for example, a Gaussian fre-
quency modulation defines a signal whose frequency starts low, then rises to a peak,
then falls again. For definition of the window parameters, see the Window and Freq
span items which follow.
– Use frequency chirp In concept this is the same as the frequency modulation, except
that only the first half of the window function is used. Thus, for example, if a Gaussian
window is used then the frequency of the signal starts low, then rises to a peak.
• Window This option is only used if the frequency is modulated. The window functions
which are offered here are the same functions as those described earlier for the amplitude
modulation, but here they are used to modulate the frequency of the signal as a function of
time. Thus at any given time the frequency of the signal is determined from the amplitude
of the window at that time.
• Freq span This option is also only used if the frequency is modulated. This determines
the range of frequency for the modulation. The range is centred on the nominal carrier
frequency. Thus, for example, if the nominal frequency is set at 1 MHz and the span is set
at 200 kHz then the frequency ranges from 0.9 MHz to 1.1 MHz.

6.2.2 Multi-mode

The information on the Multi-mode tab is used to observe other modes which may also be present
at the chosen frequency. The purpose is to show the arrival times and the signal shapes of any or
all of the other modes which could be present. It is important to appreciate however that there
is absolutely no significance in the amplitudes of these modes. That is to say, Disperse does not
calculate how much of these modes might be excited by any particular transducer arrangement,
nor the reflection factors from the end of a plate. The options are:

• Pitch-catch/Pulse-echo This defines the nature of the multi-mode simulation. If Pitch-


catch is selected, then Disperse assumes a configuration of two transducers at separate
positions along the structure. The distance of separation is as specified on the Excitation
tab. The signal is generated by one of them and received by the other. All of the modes
which are selected for the simulation are assumed to start concurrently at the generating
transducer, then the simulation predicts what is received by the receiving transducer. All of
the modes are generated using the same frequency and wave packet information as defined
on the Excitation tab. Alternatively, if Pulse-echo is selected, then Disperse assumes that
there is a single transducer which acts both as generator and receiver. The transducer is
situated at a distance from an end of the structure; the value of this distance is as specified on
the Excitation tab. The single mode which was selected on the Excitation tab is assumed
to be excited and travels, alone, to the end of the structure. At the moment of reflection
from the end of the structure it is assumed that all of the other modes which are selected
are generated; furthermore the wave packet which is used to generate these other modes is
that of the incident mode at the moment of arrival at the end of the structure - this may
be significantly different from the initial wave packet because of dispersion. Disperse then
predicts the multiple modes which are received when they reach the transducer. Note that

Disperse version 2.0.20a


6.3 Calculate function 73

Disperse 3 does not predict the reflection coefficients which would be needed to know the
actual amplitudes of the reflections of the different modes from the end of any particular
structure.
• Mode Below this title word is a list of all of the modes which have been calculated for
the structure. A tick in any of the boxes in this list turns on the relevant mode. Thus all
modes which are ticked are included in the simulation. However, note that not all modes
will necessarily appear in the display: only those which exist at the chosen frequency can be
generated.
• Amp Each of the modes is assumed by default to be generated with an amplitude of 1.0,
but the amplitudes may be changed by modifying the values in the Amp data box list.

When the multi-mode information has been defined, the results are displayed graphically in two
displays. The upper display shows all of the selected modes as separate time traces superimposed
on the same graph, each with a different colour, so enabling the contributions from the different
modes to be identified. The lower display shows the proper superposition of all of these signals,
that is to say the signals are all added and the display shows the single graph of their sum. This
is of course representative of the addition of the signals which would occur in practice.

6.3 Calculate function

This is a powerful utility which allows you to define a function which you may then apply to the
dispersion curves which you have solved. The utility can be accessed either from this menu item
or by choosing Display : Derived data from the Display menu. Both routes bring up the same
dialogue box but start with different default menu choices.

What sort of function might we want to calculate and why is this tool useful? Initially it may not
be clear what sort of things we could do. As examples, consider four things we might want to do:
(1) calculate the group slowness from the dispersion curves, given by the reciprocal of the group
velocity; (2) calculate the angle of incidence for exciting the modes in a material which is to be
placed in contact with the plate (eg a wedge transducer), given by Snell’s law; (3) calculate the
ratio of the in-plane displacement divided by the normal displacement at the surface of a plate so
as to determine how readily the waves in the plate might radiate energy if it was to be immersed
in a fluid; (4) plot dispersion curves using some obscure dimensionless axes in order to compare
with the curves in a published article. All of these can be done using the Calculate function
option.

On selection of the Calculate function option a dialog box appears, offering several methods
of defining the function. The large data box across the middle of the dialog box offers the most
general method of defining a function by typing in an equation to be evaluated. However, there
are various shortcuts and predefined functions which are quicker and easier to use to begin with
and which will suit most requirements. The new user is therefore advised to start with these
predefined routes and move on to the general equation method only when necessary. Accordingly
we give our description also in that order.

6.3.1 Input Using Menu Options

Start with the box at the top right, which offers a choice of several types of calculation, as follows:

Disperse version 2.0.20a


6.3 Calculate function 74

Take value: This simply extracts the value of some chosen quantity at each position along the
dispersion curves. Some of these values, the state variables, are already in memory and are simply
extracted. Others, the mode shape components, relate to a chosen position on the structure and
will need to be calculated for each point on each dispersion curve. A third category, integrated
components, requires the integration of a quantity across the thickness of the structure, again
performed at each point on each dispersion curve. The available choices of these three kinds of
variables are listed in the following paragraphs. In each case the abbreviated symbol in angle
brackets is the name given to the variable for the equation function. When you select one of these
variables you will see this name appear in the large data box across the middle of the dialog box,
as the program assembles the chosen equation.

The state variable quantities are:

• Frequency <Vfq> The frequency (M Hz).


• Wavenumber <Vk > (NB there is a space after the “k”) The real wavenumber (cycles/mm).
• Attenuation <Vat> The attenuation (N epers/m).
• Phase velocity <Vph> The phase velocity (km/sec).

• Group velocity <Vgr> The group velocity (km/sec).


• Wavelength <Vwv> The wavelength (mm).
• Current x axis <AXIS[0]> Whatever variable is currently plotted on the x axis of the
graph display.

• Current y axis <AXIS[1]> Whatever variable is currently plotted on the y axis of the
graph display.

The mode shape quantities are a little different in definition, because it is necessary now to define
additionally the location of the quantity in the layered structure. Accordingly when any of these
quantities are requested two extra data boxes appear for the definition of the location. First, the
available quantities are:

• Strain energy density <VEs> The strain energy density, see section 6.1.1.

• Kinetic energy density <VEk> The kinetic energy density, see section 6.1.1.
• Total energy density <VEt> The total energy density, see section 6.1.1.
• Normal displacement <Vdx> The displacement in the direction normal to the layer (X
for plate, radial for cylinder), see section 6.1.1 (nm).

• Perpendicular displacement <Vdy> The displacement in the direction normal to the


plane of calculation (Y for plate, circumferential (q) for cylinder), see Section 6.1.1 (nm).
• In-plane displacement <Vdz> The displacement in the direction of propagation of the
wave (Z), see Section 6.1.1 (nm).

• Stress xx <Vxx> The direct stress xx (GP a).


• Stress yy <Vyy> The direct stress yy (GP a).
• Stress zz <Vzz> The direct stress zz (GP a).

Disperse version 2.0.20a


6.3 Calculate function 75

• Stress xy <Vxy> The shear stress xy (GP a).


• Stress yz <Vyz> The shear stress yz (GP a).
• Stress zx <Vzx> The shear stress zx (GP a).
• strain x <Vex> The direct strain xx (m/m).

• strain y <Vey> The direct strain yy (m/m).


• strain z <Vez> The direct strain zz (m/m).

The data box immediately to the right of the quantity box defines the layer in which the quantity
is to be evaluated. The order of the layers is simply the order in which the structure is defined.
Note that the numbering system starts with zero. The next data box to the right defines the
position in the layer, offering several self-evident choices. Thus the location in the structure
is defined. Further possibilities for the position in the layer can be achieved using the general
equation function which will be explained later.

Finally, the available integrated components are:

• Energy velocity <VEV> The energy velocity (km/sec). This indicates the speed of
propagation of a wave packet. It is the same as the group velocity if there is no attenuation
in the solution, see Section 5.2.2.
• Strain energy <VES> This is the integration of the strain energy density over the thick-
ness of the structure. The result thus consists of an integral both over one cycle and over
the total thickness.
• Kinetic energy <VEK> This is the integration of the kinetic energy density over the
thickness of the structure. The result thus consists of an integral both over one cycle and
over the total thickness.
• Total energy <VET> This is the sum of the strain energy and the kinetic energy.
• Power flow 1 <VE1> This is the amplitude of the X direction component of the Poynting
power flow vector, see Sections 6.1.1, 6.1.3 and 7.10 (km/sec).
• Power flow 2 <VE2> This is the amplitude of the Y direction component of the Poynting
power flow vector, see Sections 6.1.1, 6.1.3 and 7.10 (km/sec).
• Power flow 3 <VE3> This is the amplitude of the Z direction component of the Poynting
power flow vector, see Sections 6.1.1, 6.1.3 and 7.10 (km/sec).

Predefined equation: A number of useful functions have already been predefined using these
tools. The main predefined functions are:

• Phase slowness. This is the reciprocal of the phase velocity.


• Group slowness. This is the reciprocal of the group velocity.

• Wavelength. This is the reciprocal of the real wavenumber.


• Attenuation per wavelength. This is the attenuation per metre divided by the wavenum-
ber.

Disperse version 2.0.20a


6.3 Calculate function 76

• Incidence angle. This is the angle of incidence of an incoming wave in a chosen mate-
rial such that its component of wavenumber along the structure matches that of the mode
(Coincidence principle and Snell’s law). This is useful for identifying the angles to use to
excite the mode with wedge angle transducers or immersion coupling. Several materials are
offered; for alternatives simply choose one of these, then edit the velocity of the coupling
material which appears in the equation box.
• Root verification (energy). This is the strain energy divided by the kinetic energy. Since
both are integrated over one cycle and over the whole thickness of the structure, they should
be equal, at least for the cases without damping or leakage.
• Root verification (velocity). This is the energy velocity divided by the group velocity.
These quantities should be the same for problems without damping or leakage.

Take ratio: With this option you can take the ratio of any two values calculated in the manner
just described. Two data boxes are now offered. Each can be set up independently in exactly the
same way as for the previous (Take value) option. Then the result of the function is the ratio of
these two values.

Sum: This works in the same way as Take ratio, but instead adds the values.

Subtraction: This works in the same way as Take ratio, but instead subtracts the values.

Multiplication: This works in the same way as Take ratio, but instead multiplies the values.

6.3.2 Input Using General Equation

The most general way to define a function is to enter it in the central equation box in the Calculate
function dialog. The equation editor interprets a range of parameters (the data information), by
abbreviation codes, and mathematical functions. A parameter is defined in angle brackets < >
using three pieces of information, as follows:

<a[b,c]>

In which:

• “a” is the abbreviation code for the quantity which is to be evaluated. These are all of the
abbreviation codes in the preceding explanations. For example the frequency is <Vfq>.

• “b” is the number of the layer. This is as described in the previous section. The top half
space is layer 0, the next is layer 1, and so on.
• “c” may be the position in the layer, in which case it is a real number between 0.0 ( = the
top of the layer) and 1.0 ( = the bottom of the layer). Alternatively, “c” may itself represent
a function from the following list:

– max This takes the maximum value from all points through the thickness of the layer.
– min This takes the minimum value from all points through the thickness of the layer.

Disperse version 2.0.20a


6.3 Calculate function 77

– amax This calculates the absolute value at each point through the thickness of the
layer, then takes the maximum of these.
– amaxi This calculates the absolute value at each point through the thickness of the
layer, then takes the index of the maximum.
– int This integrates the variable through the thickness of the structure.
– aint This integrates the absolute value of the variable through the thickness of the
structure.
– x,real This takes the real part of the variable at position x in the layer, where 0 ≤ x ≤
1. For example the expression 0.25,real takes the real value at a position 25 % down
from the top of the layer. See full example of function using this at end of this section.
– x,imag This is similar to the previous but takes the imaginary part of the variable.
– x,abs This is similar to the previous but takes the absolute value of the variable.

Additionally, the bulk velocities of the materials of any of the layers may be extracted by the use
of <Vl0[n]> for the longitudinal velocity or <Vs0[n]> for the shear velocity, where n is the layer
number.

The available mathematical functions, which must be written in lower case, are:

• +, -, *, / These are the fundamental operations add, subtract, multiply, divide.


• ∧ Raise to the power, eg x∧2 is x squared.
• ( ) Brackets can be put around parts of the equations, and can be nested, to control the
order of calculation. The use of brackets is strongly recommended.
• sin, cos Sine and cosine, arguments in radians.
• asin, acos Arc sine and arc cosine, results given in radians.
• sqrt Square root.
• log Logarithm, base 10.
• abs Absolute value.
• pi Gives the value of Pi, =3.14159...

We finish with a few examples.

1. To calculate the coincidence angle for a longitudinal wave in the upper half space to match
the wavenumber of the guided wave in the structure (eg immersion coupling), we could use
the function:
asin(<Vl0[0]>*<Vk >/<Vfq>)*180.0/pi
2. To calculate the ratio of the in-plane displacement divided by the normal displacement at
the surface of a plate (eg so as to determine how readily the waves in the plate might radiate
energy if it was to be immersed in a fluid), we could use the function:
<Vdx[1,0.00]>/<Vdz[1,0.00]>
3. To extract the absolute value of the stress component σzz at a position one tenth of the
depth from the top surface of the epoxy layer in a metal/epoxy/metal lap joint, we could
use the function:
<Vzz[2,0.10,abs]>

Disperse version 2.0.20a


6.4 Show Bulk Velocities 78

6.3.3 Sampling Information

The two data boxes at the bottom of the Calculate function dialog are used to control the
sampling for the calculations:

• Points through thickness: This is used only when a function has to perform an integration
through the thickness of the structure. The integration is performed using the trapezium
rule and this number of subdivisions through the thickness.
• Calculating at every: This determines how many points on the dispersion curves are
used. If a value of 1 is given then every point on the dispersion curves is used, if 2 is given
then every other point, and so on. If the value is not 1 then the intermediate points are
calculated by interpolation. An alternative method to reduce the number of points is to use
the Resample data option before calculating the function, see Section 6.6.

6.4 Show Bulk Velocities

When working with leaky systems it is helpful to be able to see where the bulk velocities lie on
the dispersion curves. The Tools : Show Bulk Velocities option allows the user to see this
information. Lines are drawn at the shear and longitudinal bulk velocities for all of the isotropic
materials in the system. The lines are plotted in any of the velocity or wavenumber graphical
displays. Selecting a point on one of the bulk velocity lines will indicate the bulk velocity to which
it corresponds, via a display string which appears above the plot.

The bulk velocity lines can be removed by choosing Edit : Delete : Bulk velocity lines.

6.5 Verify Solution

The selection of Tools : Verify solution opens up a dialog which offers several tools which can
be useful for checking the validity of a solution. A couple of examples of when such a check may
be useful are (1) when there is some doubt about whether a mode which has been calculated is
really a correct mode of the structure, or instead is maybe a mathematical artifact, or (2) to check
an upgraded version of the program against stored old solutions. The options for the utility are
as follows:

• Modes/points to verify This determines the number of points on the currently displayed
dispersion curves which are to be verified. The first three choices limit the check to the
currently selected mode: Verify selected point checks only the currently selected point;
Verify selected mode (10 points) picks out for checking an even distribution of 10 points
along the current mode; Verify selected mode (all points) checks al points on the current
mode. The final three choices on the menu perform in the same way, but for all of the modes
in the current display.
• Check convergence This uses the usual Disperse algorithm to check that the solution point
satisfies the convergence criterion. The algorithm is explained in Section 7.8.
• Check energy consistency This calculates the strain energy divided by the kinetic energy.
Since both are integrated over one cycle and over the whole thickness of the structure, they
should be equal, at least for the cases without damping or leakage. See Section 6.3.

Disperse version 2.0.20a


6.6 Resample Data 79

• Check energy/group velocity This makes two separate calculations, one of the group
velocity (see Section 5.2.2) and the other of the energy velocity (see Section 5.3.2). This is
quite a powerful comparison because the calculations are entirely different: the group velocity
is calculated directly from the dispersion curve whereas the energy velocity is calculated
from the mode shapes. But note that this check is only relevant in cases where there is no
attenuation; this is because if there is attenuation then the expression for the group velocity
is not strictly correct. See also limitation for energy velocity in Section 1.2.
• Re-assemble as more layers and re-check If a single layer structure is re-modelled using
two layers of the identical material, and each of half the thickness, then the results should
turn out to be the same. This option uses this concept to perform a check.

• Verification tolerance (%) This sets the threshold for acceptance of the verification for
the Check energy consistency, Check energy/group velocity and Re-assemble as
more layers and re-check options. If the value being checked differs from the value being
compared by more than this percentage, then an error is recorded. This number can be
changed from the default 0.5 % by clicking the arrow to the right of the box, then clicking
the small up or down arrows in the small window which appears.

Once the verification parameters have been selected, click on the Go button to perform the
checking. A summary of the results will appear in the text box at the bottom of the dialog
window.

6.6 Resample Data

The calculation of the dispersion curves is performed by default using a constant step size in the
direction tangential to each curve in the frequency-real-wavenumber space. This provides a good
distribution of points on the curve for the graphical displays. However it is often more convenient
for the user to have some other distribution of points, for example constant steps in frequency,
or indeed values at specific round numbers of frequency, wavenumber or velocity. Furthermore, if
very large numbers of steps have been taken in order to find difficult roots then it can be useful
to re-sample the curves with a smaller number of points in order to reduce the volume of data.
The Resample data function performs this task.

The function is very easy to use. First calculate the dispersion curves in the usual way, save the
file (it is a good idea to do this in case you want to return to the original data if you are not happy
with your choice of re-sampling), then open the function by selecting Tools : Resample data.
There are two pieces of information to provide:

• Resample This sets the re-sampling axis, offering a choice of Frequency, Wavenumber,
Phase velocity or Group velocity. The re-sampling is performed in constant steps along
the chosen axis. Thus if you choose to re-sample in frequency then the curves will be re-
sampled such that all of the points are evenly spaced in frequency.
• Sample every This specifies the sampling interval. The appropriate units of the interval are
automatically written next to the data box according to the choice of the re-sampling axis.
There is no restriction on the size of the sampling interval, it may freely be defined to be
smaller or larger than the interval which was used for the original calculation of the curves.
However, it is worth thinking about the implications on the different graphical displays; for
example a coarse spacing in phase velocity will result in very poor representation of the

Disperse version 2.0.20a


6.7 Input and Output of Information 80

phase velocity-frequency plot of the modes which are close to parallel to the frequency axis
(eg high frequency a0 or s0 in a plate).

Having made the selection of the re-sampling parameters, click OK to perform the calculations. All
of the curves are processed. The re-sampled curves replace the original curves so it is worth keeping
a copy of the original file before re-sampling. The re-sampling is performed by interpolation; no
new root-finding calculations are performed. For that reason, re-sampled points are not as accurate
as the original calculated points. However, provided sensibly small steps were taken in the original
calculations, the interpolation error is negligible.

6.7 Input and Output of Information

Disperse has been developed to be integrated within the Windows environment, with flexible ex-
porting and importing of data, copying and pasting, Object Linking and Embedding, and printing.

6.7.1 Saving

As soon as the dispersion curves for a structure have been generated, all of the information should
be saved. Disperse uses its own format file to save all of the information for a particular structure,
including the geometry, the solution parameters, the dispersion curves and any labels. Thus if
you save a file, then exit the program, then restart and open the file again, you should get back to
the same displays and settings. However Disperse does not save the post-processed information
such as mode shapes, simulated signals or calculated functions. This is because these are readily
calculated again from the dispersion curves.

To save, simply choose from the menu File : Save. Alternatively, choose File : Save as if you
would like to change the name of the file. A .ds2 extension will automatically be appended to
the filename and should always be retained for Disperse files. Note that this is different from the
version 1.0 extension .dsp, see Section 1.6.

To open an existing file, use File : Open. When opening files, Disperse version 2.0 uses a filter
for version 2.0 files (.ds2) by default. However it will open the older format (.dsp) too. Subsequent
saves of the older format files will automatically be in the new format, so updating any of the old
files.

6.7.2 Exporting

The dispersion curve or mode shape results can be exported to a file which can be read by other
programs, for example for further processing or document preparation. This is done be choosing
the menu item Tools : Export.

The graphical image which is exported using these tools is that which is in the current display.
Thus select the dispersion curves window first if you want to export the dispersion curves, or the
mode shapes window if you want to export the mode shapes.

Several export file formats are available, as follows:

Disperse version 2.0.20a


6.7 Input and Output of Information 81

• Postscript File The postscript option creates a postscript file (.ps) that can be imported
into many vector drawing packages. This is a good way to transfer the results if high
resolution graphics are required.
• Windows 3.11 Metafile This creates the standard Windows data transfer file (.wmf) which
can be read in by many Windows programs. This is quick but the graphical resolution of this
format is not particularly good. This may be an important consideration if the document is
to be printed.
• Enhanced Metafile This creates the enhanced version of the Windows file (.emf) which
offers improved resolution. However this can only be read by the later versions of Windows
applications.

• Text (As displayed) This writes a simple tab-separated ASCII text file (.txt) which can
be edited by a text editor or read in by a wide range of other programs such as spreadsheet
or graph plotting programs. Only the information which is shown in the current plot is
exported. The dispersion curves or mode shapes are exported one at a time.
For dispersion curves, each mode is described by three columns of information: (1) X axis
values of the points on the curve, (2) corresponding Y axis values, (3) corresponding “Colour”
axis values (see Section 5.3.2) - the colour axis values will all be zero if no colour axis is
defined. If minima or zeroes (the small circles) are currently displayed then they are exported
too, each being treated as a single-point dispersion curve. Further options relating to the
arrangement of the information in the file is also defined by a dialog box which appears when
this file format is chosen. The information in this dialog box, Title header, Point header,
Layout and Curves, is described in the Preferences Section (5.5). By default the choices
here take the current selections defined in the Preferences.
For mode shapes, the first column of the file contains the distance positions through the
thickness of the structure, then there are two columns for each curve, corresponding to the
real and imaginary values of the mode shape.

• Text (Stnd variables) This applies only to dispersion curves. This is similar to the previous
item except that all of the “standard” variables which define a curve are written, rather
than just those currently on display. There are 5 standard variables and so 5 columns of
information for each curve. The variables are, in order: (1) frequency, (2) real wavenumber,
(3) attenuation, (4) phase velocity, and (5) group velocity.

• Text (All variables) This is similar to the previous item except that it additionally includes
any other information relating to the curves, for example the order number for cylindrical
solutions.

Note that, where relevant, the settings which are selected in the Preferences Section (5.5) will
apply to the exported data.

Finally, it is possible to export movie files of animated mode shapes; see Section 6.1.2.

6.7.3 Import Text File

This option is included to enable Disperse to read in information from a text file, in a format
such as is created using the text export options (although in fact it is quite flexible in what it can
interpret). For example this would enable Disperse to display curves which had been created by
another program or which had been exported from Disperse, then processed by another program,
then returned to Disperse. Start this option by selecting the menu item File : Import text file.

Disperse version 2.0.20a


6.7 Input and Output of Information 82

After identifying the file to be opened, a dialog box will appear, requesting information about the
file. The items in this box are:

• Num header lines This specifies the number of lines of information to ignore at the begin-
ning of the file before the start of the data. Thus header information can easily be skipped.
• Column order This specifies the arrangement of the columns which are read in. Up to
5 columns of information can be read in for each mode. xyz.... specifies that all of the
columns (up to 5) belong to the same mode. Thus “x”, “y” and “z” are different pieces of
information for the same mode. y y y y specifies that each column refers to a separate mode;
the same parameter (whatever the “y” information is set to be, see the column specification
below) is to be read for all of the columns across the file, and each is to be allocated to be
the property of a separate mode. xy xy xy specifies, in a similar way, that each pair of
columns refers to a separate mode; thus is the file has six columns then it will be assumed
that the first 3 refer to one mode and the second 3 to another. xyz xyz specifies, in the
same way, each set of 3 columns refers to a separate mode.
• Delimiter This specifies how the columns are separated - by tab, comma, or in fixed column
format.

• Colour This sets the colour to be used for the plotted curves.
• Col 1: This and the following rows specify how the information is to be interpreted by
Disperse. Thus column 1 of the data needs to be equated to the type of data which is being
read from column 1, and so on for the other columns. Each type of data is given a type
and also, in a data box to the right, a unit maybe specified. Thus, for example “Frequency”
could be selected for column 1, together with the units “(MHz)”. Note that the list of
possible column types is taken from the current possibilities for the Disperse display. Thus
it is possible that a desired column type is not offered. For example, if the current Disperse
solution on display does not have attenuation (elastic materials, no leakage), then attenuation
will not be offered as a column type. To get attenuation as a type, display an attenuative
solution in Disperse first before opening the import function.

6.7.4 Copy and Paste

Disperse offers flexible options for copying and pasting, with full OLE (Object linking and Em-
bedding) support. We consider here, separately, the copy and the paste functions.

Copying may be achieved using the menu functions, Edit : Copy and Edit : Copy all, and
the right mouse button functions (listed under Actions) Copy text, Copy picture and Copy
Metafile. These can be used with the data in the dispersion curve display, or indeed with other
displayed data such as slowness curves (Section 3.3.3), simulated signals (Section 6.2), or mode
shapes (Section 6.1). Starting with the right mouse button options, the copy options are as follows:

• Copy text This copies the data of the current curves onto the clipboard as columns of
ASCII text.
• Copy Metafile This copies using the standard Windows data transfer format which can
be pasted into many Windows programs. This is quick but the graphical resolution of this
format is not particularly good. This may be an important consideration if the document is
to be printed

Disperse version 2.0.20a


6.7 Input and Output of Information 83

• Copy Picture This uses the enhanced version of the Windows Metafile which offers im-
proved resolution. However this can only be read by the later versions of Windows applica-
tions. It should also be remembered that this takes more memory and so may take a long
time if there are many curves and points to copy.
• Copy This copies the currently selected mode. The information is copied simultaneously in
three formats: (1) as text, (2) as a standard Metafile, and (3) as a Disperse object which
can be pasted and linked in another application. When pasting into a destination document
the appropriate choices of these three formats should therefore all be offered. Note that this
necessarily takes quite a lot of memory - if there is a problem with the time taken then
a preferable alternative is to copy just the text or just the Metafile using the right mouse
button options. The right mouse button also offers the improved resolution picture with the
enhanced Metafile format.
• Copy all This does the same as the previous item except that it copies all of the modes
rather than just the currently selected mode. See important note in previous item relating
to memory demands.

Pasting of Disperse results within Disperse can be used to compare the results for different struc-
tures. First generate separately the results for the two different structures, keeping both dispersion
curves open. Then copy one or more modes from one set and paste them onto the other window.
The curves will be superimposed, and will in fact be added to that set, enabling direct comparison
on the screen, or by printing them, exporting them etc.. However the pasted modes will retain the
property of being “Foreign” modes, and will clearly not be connected to the structural properties
of the host window. Foreign modes can be selectively hidden from the display or deleted, see
Section 5.3.3. Since it is easy to confuse the two sets of dispersion curves, it is recommended that
users avoid using the sweep and trace options in a window that contains Foreign modes.

Pasting of Disperse results outside Disperse is also possible, provided that the destination program
is suitably configured. See the Copy options above to determine the appropriate copying mode
to use. If the destination program supports OLE (Object Linking and Embedding), then Disperse
results can be pasted as a linked object. Then subsequently if the user double clicks on the linked
object, Disperse will be automatically opened up with the curves for further calculation or editing.
Alternatively the Disperse results can simply be pasted as text or a graphical image, according to
the copying choice, in the destination program.

In the contrary sense, objects from other programs can be pasted into Disperse. This is done
using the menu command Edit : Insert new object. This brings up a dialog box with a list of
available programs from which the object could be taken. If you select one of these, then create (or
open/find) the desired object, then return to Disperse, the object will be pasted into the dispersion
curve display. Here it can be moved by clicking and dragging with the left mouse button, or re-
sized by dragging the control points in the usual way. This option is useful for example to add an
equation or an explanatory diagram to the display. Later, you can return to the donor program
with the object by double clicking on the object in the Disperse display. This needs to be done
carefully otherwise you will find that you just select the nearest mode rather then the object. To
select the object, position the mouse pointer over any part of the object whose colour is different
from the background colour of the page. The object can also be deleted by selecting it, then
pressing the Delete key.

Disperse version 2.0.20a


6.8 Labels 84

6.7.5 Printing

The dispersion curve graphical display, the mode shapes displays and the simulated signal displays
can each be printed directly on a Windows supported printer. This is done by choosing File :
Print. The right mouse button also offers the print instruction when the mouse is in the main
dispersion curve display. The display which is printed is the currently active window - this is
important to note because there are often several windows on the screen. If in doubt, click in the
top band of the window which is to be printed; this will select it as the active window.

The printer options are set in the usual way with the printer driver; this can be accessed by
the menu command File : Print setup. Additionally there are some other print options, for
example whether or not notes about the geometry should be included, which can be set using the
Preferences menu; see Section 5.5.

There is also a File : Print preview option which gives a preview on the screen of what would
be printed using the currently selected options.

6.8 Labels

Disperse provides a labelling facility which can be used to identify the different modes or mark
important parts of the dispersion curves. The labelling facility works by attaching each label to
a specific point on a dispersion curve. When the display type changes or the user zooms in on a
part of the dispersion curve, the label remains attached to the same point on the dispersion curve.
Note that the label of a mode is not necessarily the name of the mode. The name of a mode is
the name written at the start of the text at the top of the display when a mode is selected by
the pointer. Disperse gives names to all of the modes automatically as it calculates them, the
conventional names are given where possible, otherwise it just uses sequential numbers for the
names, see more about naming of modes in Section 4.2.

The label functions are available through the Tools : Labels submenu. These allow you to add
new labels, edit current labels, move labels to different locations, or delete labels. The functions
are operated as follows:

• New label This option changes the action of the mouse to the creation of new labels. A
small image of a label appears as the pointer and can be used to locate the position of a new
label. Just click with the left button on or near a curve to identify the position for a new
label. Disperse attaches the label to the nearest point on the nearest curve. After clicking,
Disperse immediately shows the Edit label dialog box for input of the label information.
This is described in the next item. Note that the pointer now retains the New label action
until you change it to another action (see explanation of this concept in Section 5.4.2). The
New label function can also be activated by clicking the leftmost of the four label icons on
the toolbar.
• Edit label This option can be used to change the properties of an existing label, by
clicking on or near the label (Disperse finds the nearest label). It opens a dialog showing
the properties of the label. The same dialog is shown when creating a label using the New
label command (previous item). The Edit label function can also be activated by clicking
the second from the left of the four label icons on the toolbar (in the shape of an I). Note
that the dialog box is rather general and not all of its options are useful here. The commonly
useful options in the dialog are as follows:

Disperse version 2.0.20a


6.8 Labels 85

– Label text Enter the text for the label directly into the text box at the top of the
dialog. This is the text which will appear as the label in the display.
– Attachment This defines the way in which the label is associated with the curve.
Point on a mode is the default: this simply attaches the label to a particular point
on the mode, as described above. Name of a mode does the same thing, but it also
automatically sets the label text to be the name of the mode. Furthermore, if the
Label text is now changed, this will change the name of the mode. Position on X
or Position on Y place the label on the X or Y axis respectively; the position on the
axis is defined using the X or Y data boxes, see below. Cursor places the label at the
XY position defined by the X and Y data boxes.
– Marker This sets the way in which the point on the mode is identified. The default,
Arrow, draws a line with an arrow head from the text to the point on the mode. Other
possibilities include putting a Cross or a Circle symbol on the point on the mode.
– Colour By default the label and its marker are black. This option allows the choice of
other colours for them.
– Length This sets the size of the marker; this is the length of the arrow or the size of
the symbol.
– Angle This sets the angle of the arrow marker. The angle is measured in degrees from
the horizontal, and is 30 degrees by default.
– X, Y these two data boxes define the location on the X or Y axis respectively if the
Attachment was set to one of the axes or the cursor. The units for the location are
the current units of the axis.

• Move label This allows a label to be moved to a different point on the same mode or indeed
to a point on a different mode. Just click with the left button on the current position, hold
the button down, drag to the new position, then release the button. The Move label
function can also be activated by clicking the third from the left of the four label icons on
the toolbar (in the shape of two labels with an arrow going from one to the other). If the
Attachment of the label has been set to Name of a mode and the label is moved to
another mode, then the name is automatically updated to that of the new mode.

• Delete label This deletes the label that is closest to the point where the left mouse button
is clicked. This function can also be activated by clicking the right-most of the four label
icons on the toolbar (in the shape of a labels with a crossed red circle) Remember that the
mouse function will remain as Delete label until reset, so it is advisable to change it back
to pointer as soon as all chosen labels have been deleted (see explanation of this concept in
Section 5.4.2). If you wish to delete all labels, there is a quick way: select Edit : Delete :
All labels.

Disperse version 2.0.20a


Chapter 7

Wave Propagation Model

This chapter describes the development of a general purpose model for wave propagation in Carte-
sian and cylindrical systems. When isotropic materials are used, the model can account for elastic
and visco-elastic isotropic materials, single or multi-layered structures, and free or leaky systems.
The model can also calculate dispersion curves for multi-layered anisotropic systems, provided
that the system is elastic, non-leaky, and Cartesian.

Based on the geometry and material properties of the system, the model determines what res-
onances can exist in order to satisfy the boundary conditions and the bulk wave propagation
characteristics in each of the layers. These resonances control how ultrasonic waves will be guided
in the system and what properties each of these waves will have. The solutions to the guided wave
problem lie on continuous lines called dispersion curves that must be found iteratively in frequency,
wave number, attenuation space. Once the solutions have been found and the characteristics of
the waves have been determined, they can be used to design an effective non-destructive testing
system.

The earliest version of Disperse was a DOS program and was limited to flat plate geometries and
isotropic materials [51, 53]. The program was then ported to X-Windows and then Microsoft Win-
dows and developed to include cylindrical systems, anisotropic materials, and analytical solutions
for simple geometries [68, 69].

In order to construct the wave propagation model, we first determine the nature of the bulk waves
which can exist in elastic and visco-elastic isotropic materials. Once the nature of the bulk waves
is known, the stresses and displacements in a layer can be expressed in terms of the amplitudes of
all of the bulk waves that can exist in that layer. The stresses and displacements at the boundaries
of each layer can be combined with the boundary conditions to describe the entire system in one
large global matrix that relates the bulk wave amplitudes to the physical constraints. Figure 7.1
demonstrates the construction of a five layer flat plate system and Figure 7.2 shows that for a five
layer cylindrical system. The partial (or bulk) waves labeled L+-, SV+-, and SH+- are assembled
by matching the boundary conditions at each of the interfaces. At certain frequency and wave
number and attenuation combinations, these partial waves combine to form a guided wave which
propagates down the axis of the infinitely long cylinder. These valid combinations may be found
by solving the global matrix equation for its modal response. For a given system, the global matrix
equation is a function of frequency (time varying component), real wave number (spatially varying
component), and attenuation (spatial decay rate). Solutions must be found iteratively by varying
these three parameters until a valid root is converged upon. Once an initial root has been found,
roots that lie on the same line of solutions, or dispersion curve, can be traced. This process can

Disperse version 2.0.20a


7.1 Wave Equations in Bulk Isotropic Media - General Theory 87

Figure 7.1: Sample geometry of a five layer flat plate system showing the partial waves in each
layer (L+-, SV+-, and SH+-) that combine to produce a guided wave. The SH+- partial waves
are omitted in the case of Lamb waves in an isotropic plate, leaving just 4 partial waves in each
layer. The L+- and SV+- partial waves are omitted in the case of Love waves in an isotropic plate,
leaving just two partial waves in each layer.

then be repeated to find other dispersion curves that exist.

Once all of the dispersion curves have been traced, we can extract information about the guided
waves and determine which are promising for characterising the system ultrasonically.

The first part of this chapter develops the appropriate wave propagation equations for each of
the types of layers that the model can simulate. The later parts describe how each of these wave
propagation equations are combined to describe the entire system.

7.1 Wave Equations in Bulk Isotropic Media - General The-


ory

Before deriving equations that describe the behaviour of layered media, it is important to under-
stand the principles of wave propagation in unbounded media. In other words, we need to know
the bulk behaviour of a material. This information can be readily found in most text books on
the subject[59, 4, 42, 9, 10, 40], but is repeated here in its general form for completeness; specific
derivations are also presented for the flat plate and cylindrical geometries in following sections.

Beginning with Newton’s second law applied to an infinitesimal element of the material, and
the principal of conservation of mass within an arbitrary volume within an elastic solid, Euler’s
equation of motion can be derived. Euler’s equation of motion relates the particle displacement
field u(r, t), which is function of the position r, and time, t, to the stress dyadic, ~~σ (a two
dimensional vector field), in the following way,

ρ(∂ 2 u/∂t2 ) = ∇ · ~~σ (7.1)

in which the mass density of the layer, ρ, is assumed to be constant, the material is assumed to be

Disperse version 2.0.20a


7.1 Wave Equations in Bulk Isotropic Media - General Theory 88

Figure 7.2: Sample geometry of a five layer cylindrical system showing the partial waves in each
layer (L+-, SV+-, and SH+-) that combine to produce a guided wave.

linearly elastic, and the body forces (i.e. gravity) are neglected. The generalised Hook’s law then
relates the stress dyadic, ~
~σ , to the elastic constants of the material. The theory of elasticity shows
that for a homogeneous, isotropic material, the 21 possible components of the elastic stiffness
tensor reduce to two material constants, λ and µ which are called the Lamé constants (See, for
example, reference [4]). In this case, Hooke’s law simplifies to

~
~σ = λI∇ · u + µ(∇u + u∇) (7.2)
where I is the identity matrix.

Combining the two previous equations leads to Navier’s displacement equation of motion for an
isotropic, elastic, medium which is, in invariant form,

(λ + µ)∇∇ · u + µ∇2 u = ρ(∂ 2 u/∂t2 ). (7.3)

This equation may also be written as,

dilatational rotational
(7.4)
d · u)
(λ + 2µ)∇(∇ + µ∇ ×d(∇ × u) = ρ(∂ 2 u/∂t2 )

where u is the displacement vector, ρ is the density, λ and µ are the Lamé constants, and ∇2 is
the three-dimensional Laplace operator. In the second form, (λ + 2µ)∇(∇ · u) accounts for the
dilatation (compressional) portion and µ∇ × (∇ × u) accounts for the rotational (equivoluminal)
portion of the solution.

Material damping may be introduced into the system in a number of ways. For typical applications
in NDE, a damping model which describes a constant loss of displacement amplitude per wave-
length travelled provides an adequate representation of the material behaviour. This description
has often been used in ultrasonic models [91, 90, 13, 35, 84, 85, 86, 74, 52, 103, 14, 23, 34, 56, 51].

Disperse version 2.0.20a


7.1 Wave Equations in Bulk Isotropic Media - General Theory 89

The introduction of material damping replaces the Lamé constants λ and µ by operators such
that,
λ0 ∂ µ0 ∂
λ becomes λ + and µ becomes µ + (7.5)
ω ∂t ω ∂t
where the constants λ0 and µ0 are the viscoelastic material constants and ω is the frequency. The
viscoelastic model clearly reduces to the elastic model if the viscoelastic constants are zero. Thus,
the displacement equation of motion (7.3) can be expressed,
 0
λ + µ0
  0
∂2u
 
2 ∂u µ ∂u
(λ + µ)∇(∇ · u) + µ∇ u + ∇ ∇· + ∇2 =ρ 2 (7.6)
ω ∂t ω ∂t ∂t

Using Helmholtz decomposition (See reference [59] pages 52-53 for more information), the three
dimensional displacement vector in equation 7.3, u, which is finite, uniform, and continuous, and
vanishes at infinity, can be expressed as a sum of a compressional scalar potential, φ, and a
equivoluminal vector potential, H,
u = ∇φ + ∇ × H (7.7)
with,
∇ · H = F (r, t)
where F is a function of the coordinate vector, r, and the time, t. Gazis[29] points out that the
function F can be chosen arbitrarily due to the gauge invariance of the field transformations (See
reference [59] pages 207-211 for more information). In other words, the relation between the fields
and the potentials is not unique. The scalar potential φ can be replaced with φ0 + (1/c)(∂χ/∂t)
and the vector potential, H by H0 − ∇χ and provided that χ satisfies the appropriate boundary
conditions, the new potentials represent the same field. In this way, functions can be combined
to represent the entire field. We choose F (r, t) to be identically zero, making the equivoluminal
vector potential a zero-divergence vector, which implies that the field is solenoidal, i.e. there are
no sources or sinks of energy within the region. Specifying the divergence provides the necessary
additional condition to uniquely determine the three components of u from the four components
of the two potentials, which was introduced by the Helmholtz decomposition.

Substituting the expression for u into Navier’s equation of motion (equation 7.3), and using the
identities,
∇ · ∇φ = ∇2 φ ∇2 (∇φ) = ∇(∇2 φ) ∇ · ∇ × H = 0,
the resulting equation is
∇ (λ + 2µ)∇2 φ − ρ(∂ 2 φ/∂t2 ) + ∇ × µ∇2 H − ρ(∂ 2 H/∂t2 ) = 0.
  
(7.8)

This equation is satisfied if either of the two terms vanishes, which leads to the standard equations,
c21 ∇2 φ = ∂ 2 φ/∂t2
c22 ∇2 H = ∂ 2 H/∂t2 (7.9)
where 1/2 1/2
λ + 2µ − i(λ0 + 2µ0 ) µ − iµ0
 
c1 = c2 = , (7.10)
ρ ρ
or if the material is elastic,
 1/2  1/2
λ + 2µ µ
c1 = c2 = . (7.11)
ρ ρ

If the material is elastic then the constants c1 and c2 are real valued and are equal to the lon-
gitudinal and shear bulk wave velocities respectively. If there is material damping then they are
complex, but their real parts still approximate to the bulk wave velocities if the damping is light.
See discussion in following section.

Disperse version 2.0.20a


7.1 Wave Equations in Bulk Isotropic Media - General Theory 90

7.1.1 Properties of the Wave Equation with Attenuation

We are interested in including a model for attenuating waves for two kinds of phenomenon (which
may exist together), (1) structures whose materials have damping (as defined by the complex Lamé
constants in the previous section) such that the energy of the waves is absorbed as they travel, (2)
structures which are immersed in a fluid or embedded in a solid such that energy radiates (leaks)
into the half spaces. In either case the attenuation of the guided waves is one of their properties,
just as are velocity or wavenumber. Both kinds of attenuation can be described by an exponential
reduction with distance along the plate, which can be expressed by a complex wavenumber. Thus
the real part of the wavenumber, kreal , describes the propagation of the wave and the imaginary
part, kimag , describes the attenuation of the wave in space. A general form of the wave equation
will therefore be (using displacement u for example:

u = Aei(k·x−ωt) (7.12)

Separating the real and imaginary parts the equation may be rewritten as,

u = Aei(kreal ·x−ωt) e−kimag ·x (7.13)

where the first exponential term, which is wholly imaginary, describes the harmonic propagation
of the wave in the direction of the vector kreal and the second, real term describes the exponential
decay of the wave with distance in the direction of the vector kimag . In general both kreal and
kimag are vectors and can be arbitrarily oriented in relation to each other.

The attenuation per wavelength is constant in this model, or equivalently, the attenuation per unit
distance increases linearly with frequency. If the real part and imaginary part of the wavenumber
are parallel, as would normally be the case when measuring the bulk ultrasonic properties of a
material, then the material constants can be related to the measurable wave speed and attenuation:

ω cph cph
c1 or c2 = = |ki |
= (7.14)
|kr | + i|ki | 1 + |kr | 1 + iκ/(2π)

where cph is the phase velocity of the wave (NB this is always a real quantity) and κ is the atten-
uation in Nepers per wavelength, such that a wave of unit amplitude is reduced to an amplitude
of e−κ after travelling one wavelength.

Thus if the Lamé constants are known, then the bulk wave phase velocities (which are real quan-
tities) and the attenuations are given by:

c1,im c2,im
cph,1 = c1,re − κ1 , cph,2 = c2,re − κ2
2π 2π
c1,im c2,im
κ1 = −2π , κ2 = −2π (7.15)
c1,re c2,re

Or if the bulk phase wave velocities and attenuations are known, then the Lamé constants are
given by:

Disperse version 2.0.20a


7.1 Wave Equations in Bulk Isotropic Media - General Theory 91

2   2 
ρ (cph,2 ) 4π 2 4π 2 − κ22 0 ρ (cph,2 ) 4π 2 (4κ2 π)
µ = 2 , µ = 2
(κ22 + 4π 2 ) (κ22 + 4π 2 )
2   2 
ρ (cph,1 ) 4π 2 4π 2 − κ21 0 ρ (cph,1 ) 4π 2 (4κ1 π)
λ = 2 − 2µ, λ = 2 − 2µ0 (7.16)
(κ21 + 4π 2 ) (κ21 + 4π 2 )

Note that attenuation can also be modelled in the wave equation by assuming the frequency, rather
than the wavenumber, to be complex. Both approaches are valid, yet give different results, because
they represent different physical problems; see for example [79, 8] and references to discussion on
this topic in [51]. However with the focus on guided waves which may propagate very significant
distances, the complex wavenumber approach is the more appropriate here and is the default in
Disperse. Further information on attenuation models may be found in [82, 73, 72, 30, 19, 20, 1,
47, 101, 3, 86, 74, 52, 103, 17, 23, 64, 62, 16].

7.1.2 Treatment for Fluid Materials

The approach which Disperse takes to model fluids is to represent them, with appropriate param-
eters, by the material models which have been derived for solids. Theoretical details of such an
approach may be found in [60, 74].

If the fluid has zero bulk attenuation then it does not support shear waves. There is thus only one
solution to the wave equation, yielding the bulk compression velocity

 1/2
λf
c1 = (7.17)
ρ

in which λf is the bulk stiffness of the fluid. Clearly it is straightforward to substitute this for the
solid expression for compression waves whenever a fluid is defined. The absence of the shear wave
is conceptually simple - it can simply be ignored; however it does necessitate some housekeeping
when assembling the matrix equations in order to omit rows and columns which relate to fluid
layers.

If the fluid attenuates waves without viscous loss, for example by scattering from suspended
particles, then the longitudinal wave expression may be modified to account for the loss in the
same way as was introduced for solid materials, that is to say an imaginary stiffness constant is
added. The material is thus defined by a density, a bulk compression wave velocity, and a bulk
compression wave attenuation.

If the fluid attenuates waves by viscosity then it is necessary to include a bulk shear wave and also
to modify the bulk compression wave. The need for the shear wave is obvious - it provides the
shear coupling of a solid layer to a viscous fluid; the attenuation of the shear wave is always very
large so the shear wave does not travel far, it simply takes energy out of the solid and dissipates
it within a thin skin depth. The compression wave becomes attenuative because the longitudinal
wave motion involves distortion (shear deformation) as well as volume change.

For the viscous fluid model in Disperse, we consider, and limit the scope to, a compressible
Stokesian fluid with Newtonian viscosity ([55]). This is sometimes called a linearly viscous fluid.

Disperse version 2.0.20a


7.1 Wave Equations in Bulk Isotropic Media - General Theory 92

Stokes’ expression for the state of stress in a deforming fluid is

σ = −p + F (D) (7.18)

where σ is the stress tensor, p is the equilibrium pressure (that is the pressure if the fluid was at
rest), F is an arbitrary function, and D is the tensorial rate of deformation (the time derivative of
the strain). If the fluid is non-viscous then this equation reduces simply to give the state of stress
equal to the equilibrium pressure (σ = -p). The change to the stress due to the viscosity (F(D))
is known as the viscous stress.

If the function F is linear then the fluid is said to be a Newtonian fluid. Then this equation can
be written as the Navier-Poisson law of a Newtonian fluid:

σij = −pδij + λ0 Dkk δij + 2µ0 Dij (7.19)

Where λ’ and µ’ are viscosity constants. If the fluid is subjected to a bulk pressure (σxx = σyy =
σzz ), then this equation gives the bulk viscosity, κ, of the fluid:

2
κ = λ0 + µ0 (7.20)
3

A fluid for which κ is zero, such that the viscous losses are all due to shearing deformations, is
said to be a Stokes fluid. Then the viscosity of the fluid can be defined by the single constant
µ’. In this case it also follows that the mean pressure of the moving fluid is equal to the (static)
equilibrium pressure p.

Thus for κ = 0, the Navier-Poisson law of a Newtonian fluid with no bulk viscosity is:

2
σij = −pδij − µ0 Dkk δij + 2µ0 Dij (7.21)
3

We are interested in deriving stress-strain relationships for the fluid in order to equate it to a
continuum solid model. Consider first the pressure contribution. For the Stokes fluid, a static
consideration will suffice, so that we have simply,

σxx = σyy = σzz = −p = λf (xx + yy + zz ) (7.22)

Now the viscous contribution requires the relationship between the deformations and the defor-
mation rates. Since in our case the fluid motion is harmonic in time, any deformation rate can
always be stated as -iω times the deformation. The stress-strain relationship is thus:

Disperse version 2.0.20a


7.1 Wave Equations in Bulk Isotropic Media - General Theory 93

(λf − 43 µ0 iω) (λf + 23 µ0 iω) (λf + 32 µ0 iω)


    
 σxx   xx 
(λf + 23 µ0 iω) (λf − 43 µ0 iω) (λf + 32 µ0 iω)
   
σyy yy 

 
   

 
   
(λf + 23 µ0 iω) (λf + 23 µ0 iω) (λf − 34 µ0 iω)
 
σzz zz
    
=
−µ0 iω


 σ xy 



 xy 
−µ0 iω
 
σ yz 
   
yz

 
  
 
 
−µ0 iω
 
σxz xz
   
(7.23)

We equate this to the conventional Cij notation for the material stiffness of an isotropic solid,

    

 σxx 
 C11 C12 C12 
 xx 

σyy C12 C11 C12 yy

 
   


 
   

σzz C12 C12 C11 zz
    
=  (7.24)

 σxy 


 C44 
 xy 

σ C44 yz
   
yz

  
 
 

   
σxz C44 xz
   

in which there are actually only two independent constants because C11 = C12 +2C44 . Thus we
find

C44 = −µ0 iω
4 4
C11 = λf − µ0 iω = λf + C44 (7.25)
3 3

The constant C44 defines the properties of the shear waves and thus primarily models the viscous
losses where a solid is in contact with a viscous fluid. However it can be seen that the C44 constant
also appears in the constant C11 which describes the compression wave. There are thus also viscous
losses in a bulk compression wave in a viscous fluid. The viscosity constant µ0 is called the dynamic
viscosity of the fluid, often denoted η. It has dimensions of stress multiplied by time (M/(LT)).
The unit Poise is used, defined by 1 Poise = 0.1N s/m2 . An alternative parameter which is often
used for viscosity is the kinematic viscosity, ν, which has dimensions L2 /T , and is defined by:

η µ0
ν= = (7.26)
ρ ρ

Thus

C44 = −iωη = −iωρν (7.27)

Equivalent “Solid” Shear Wave Properties

It is the shear wave velocity and attenuation which create the vorticity mode in a viscous fluid. The
derivation here results in the same expression which was presented in [74]. Assume a bulk wave
travelling in an infinite medium with its attenuation vector parallel to its propagation vector. The
latter condition is convenient for the equations here, because it allows the wavenumber, velocity

Disperse version 2.0.20a


7.1 Wave Equations in Bulk Isotropic Media - General Theory 94

and attenuation to be treated simply as scalar values but this will not be a limitation of the model.
Then, from equations 7.10 and 7.14,

 1/2
ω C44 1/2
= = (−iων) (7.28)
k ρ

Rearranging,

−iω
k2 = (7.29)
ν

Expressing the complex wavenumber by k = kr + iki , where the first term is the real (propagation)
vector and the second term is the imaginary (attenuation) vector, and both are assumed here to
be parallel, and squaring, we have a second expression for k 2 :

k 2 = kr2 + 2ikr ki − ki2 (7.30)

Equating 7.29 and 7.30, and noting that the former is an entirely imaginary number,

kr2 = ki2
−ω
kr ki = (7.31)

Taking the positive root of these coupled equations,

 ω 1/2
kr = ki = (7.32)

The bulk shear wave velocity and attenuation (for definition see 7.14) are therefore given by:

ω 1/2
Cs = = (2ων)
kr

κs = ki = 2π (7.33)
kr

Equivalent “Solid” Longitudinal Wave Properties

The viscosity of the fluid causes the longitudinal bulk wave to become attenuative, simply because
longitudinal waves comprise some distortion of the material in addition to the change of volume;
only a hydrostatic compression is solely compressive. The derivation here follows that of [60],
giving the same result for the approximate solution. The approach is similar to that of the shear
wave, except that the C11 constant is used in place of the C44 constant, with the substitution of
the appropriate fluid constants:

Disperse version 2.0.20a


7.1 Wave Equations in Bulk Isotropic Media - General Theory 95

 1/2  4 1/2
ω C11 λf C44
= = + 3 (7.34)
k ρ ρ ρ

Rearranging, and substituting kr + iki for k,

 
2 λf 4iων
kr2 ki2

ω = + 2ikr ki − − (7.35)
ρ 3

The real part of this equation is

 λf 8
ω 2 = kr2 − ki2 + kr ki ων (7.36)
ρ 3

Substituting the bulk longitudinal wave attenuation κl = 2πki /kr (equation 7.14), the longitudinal
bulk velocity is:

1/2
κ2l
 
ω λf 4
Cl = = 1− + κl ων (7.37)
kr (2π)2 ρ 3

But in fact the longitudinal bulk velocity is one of the knowns supplied by the user. To be strict we
would need to incorporate the change to the bulk velocity caused by the viscosity, using the above
equation. However we can safely take this to be small enough to neglect for typical applications.
For example, the change to the bulk velocity is about 0.002 % at 1 MHz for a fluid with wave speed
and density of water and dynamic viscosity of 1 N s/m2 . This viscosity is roughly representative
of glycerine at ambient temperature. For an error of 1 % at this frequency, the dynamic viscosity
would have to be about 100 N s/m2 . So the approximate longitudinal velocity is:

 1/2
λf
Cl ≈ (7.38)
ρ

Now returning to the imaginary part of equation 7.35, equating it to zero (for real frequencies),
and rearranging,

   κ 2  κ
4ωνρ l l
1− = (7.39)
3λf 2π π

This is a quadratic equation in κl /(2π). The positive root gives

s 
 2
3πλf  4ωνρ 
κl = 1+ −1 (7.40)
2ωνρ  3λf 

 
If the factor 4ωνρ
3λf is small, then a much simpler expression is possible. Let this factor equal A,
then the approximate solution for the bulk longitudinal wave attenuation is

Disperse version 2.0.20a


7.2 Wave Propagation in Isotropic Flat Plate Structures 96

A2
    
2π 4πων
κl ≈ 1+ − 1 ≈ πA ≈ (7.41)
A 2 3Cl2

Again, for typical applications, the error associated with this approximation is small. Taking
the two examples used earlier, the errors for the cases of dynamic viscosity of 1 N s/m2 and 100
N s/m2 are 0.01 % and 3.3 % respectively.

7.2 Wave Propagation in Isotropic Flat Plate Structures

Much of the information in this section of the manual has been published in reference [51].

The field equations for the displacements and stresses in a flat isotropic elastic solid layer may
be expressed as the superposition of the fields of four bulk waves within the layer. The approach
therefore is to derive the field equations for bulk waves, which are solutions to the wave equation
in an infinite medium, and then to introduce the boundary conditions at an interface between two
layers (Snell’s law), so defining the rules for coupling between layers and for the superposition of
the bulk waves. The analysis of the layers is restricted to two dimensions, with the imposition of
plane strain and motion in the plane only. The “SH” modes, which are decoupled for this type of
system, are treated separately.

7.2.1 Historical Background

The earliest theory for wave propagation in multilayered media was Lord Rayleigh’s derivation
[81] in 1885 for waves travelling along the free surface of a semi-infinite elastic half-space. The
derivation yields a third order expression whose roots determine the velocity of the propagating
surface wave. A generalisation of the single interface problem was developed by Stoneley [96] in
1924 to describe waves travelling along the interface between two different elastic solids. Later
studies addressed the conditions when these waves may travel without leaking into either of the
solids (true modes) [88] and the existence of leaky modes [75]. In 1917 Lamb [46] added another
interface to introduce the notion of a flat layer of finite thickness. His derivation was for plates in
vacuum and the roots of his two equations, one for symmetric modes and one for antisymmetric
modes, yield the well known Lamb wave dispersion curves. Love [50] showed that transverse
modes, involving shearing motion in the plane of the layer, were also possible in layers of finite
thickness. Discussions of these and other specific layer geometries may be found in references [25]
[100] [26] [9] [10].

7.2.2 Plane Waves in an Infinite Elastic Solid

The development of the equations of motion for an infinite elastic solid has been covered in many
texts [9][10][40]. The usual approach, summarised here, is to start with an infinitesimal cubic
element in an infinite elastic isotropic solid of density ρ. A Cartesian system is adopted, with
displacements u (u1 , u2 and u3 ) in the coordinate system x (x1 ,x2 and x3 ). By application of
Newton’s second law, equilibrium requires that:

Disperse version 2.0.20a


7.2 Wave Propagation in Isotropic Flat Plate Structures 97

∂ 2 u1 ∂σ11 ∂σ12 ∂σ13


ρ = + +
∂t2 ∂x1 ∂x2 ∂x3
∂ 2 u1 ∂σ21 ∂σ22 ∂σ23
ρ 2 = + +
∂t ∂x1 ∂x2 ∂x3
∂ 2 u1 ∂σ31 ∂σ32 ∂σ33
ρ 2 = + + (7.42)
∂t ∂x1 ∂x2 ∂x3

where σ11 ,σ12 etc. are the stress components acting on the faces of the cube and t is time.

These are the fundamental stress equations of motion for the medium. It will be more convenient
if they are expressed in terms of displacements, by introducing the stress- strain and the strain-
displacement equations:

σ11 = λ∆ + 2µ11 σ22 = λ∆ + 2µ22 σ33 = λ∆ + 2µ33


σ12 = µ12 σ23 = µ23 σ13 = µ13
∂u1 ∂u2 ∂u3
11 = 22 = 33 =
∂x1 ∂x2 ∂x3
∂u1 ∂u2 ∂u2 ∂u3 ∂u1 ∂u3
12 = + 23 = + 13 = + (7.43)
∂x2 ∂x1 ∂x3 ∂x2 ∂x3 ∂x1

where λ and µ are Lamé’s elastic stiffness constants and ∆ = 11 + 22 + 33 is the change in
volume (dilatation) of the element.

Substitution yields the displacement equations of motion

∂ 2 u1
 
∂ ∂u1 ∂u2 ∂u3
ρ 2 = (λ + µ) + + + µ∇2 ∂u1
∂t ∂x1 ∂x1 ∂x2 ∂x3
∂ 2 u2
 
∂ ∂u1 ∂u2 ∂u3
ρ 2 = (λ + µ) + + + µ∇2 ∂u2
∂t ∂x2 ∂x1 ∂x2 ∂x3
∂ 2 u3
 
∂ ∂u1 ∂u2 ∂u3
ρ 2 = (λ + µ) + + + µ∇2 ∂u3 (7.44)
∂t ∂x3 ∂x1 ∂x2 ∂x3

which may be expressed in the vector from:

∂2u
ρ = (λ + µ)∇(∇ · u) + µ∇2 u (7.45)
∂t2
   2 
∂ ∂ ∂ 2 ∂ ∂2 ∂2
where ∇ is the vector operator xˆ1 ∂x 1
, x
ˆ 2 ∂x2 , x
ˆ 3 ∂x3 and ∇ is the scalar operator ∂x2 , ∂x 2 , ∂x 2 .
1 2 3

This equation cannot be integrated directly; a form of solution must be assumed and checked
for suitability by differentiation and substitution. Here it is assumed that the wavefront is an
infinite plane which is normal to the direction of propagation. It is also assumed that, at any
position in the propagation direction and at any instant in time, all displacements are uniform
over the plane of the wavefront. This defines a homogeneous plane wave. For such assumptions
there are two solutions, one for ’longitudinal’ waves and the other for ’shear’ waves (the ’bulk
waves’). The particle motion in longitudinal waves is entirely in the direction of propagation and
the wave motion consists of change of volume (dilatation) only. The particle motion in shear
waves is normal to the direction of propagation and the motion consists of rotation of the medium

Disperse version 2.0.20a


7.2 Wave Propagation in Isotropic Flat Plate Structures 98

without change of volume. A convenient way of presenting the solutions in vector form is by the
Helmholtz method [55], in which longitudinal waves (1 ) are described by a scalar function φ and
shear waves (2 ) by a vector function H whose direction is normal to both the direction of wave
propagation and the direction of particle motion:

φ = AL ei(k·x−ωt)
|H| = AS ei(k·x−ωt) (7.46)

Here AL and AS are the longitudinal and shear wave amplitudes, k is the wavenumber vector and
ω is the angular frequency. The wavenumber vector is in the direction of propagation of the wave
and describes its wavelength and speed.

2π ω
Wavelength = Speed = c = (7.47)
|k| |k|

To generalise the expressions the amplitudes may be taken as complex quantities AL eiψ ,AS eiψ ,
where ψ is the phase of the wave at the spatial and temporal origin x = 0, t = 0.

The displacement field is given by the operations:

u = uL ∇φ + uS ∇ × H (7.48)

where × denotes the vector cross product, ∇φ corresponds to the dilitational (longitudinal) motion,
and ∇ × H corresponds to the equivoluminal (shear) motion. Substitution into the equation of
motion (7.45) gives the wave speeds, c1,2 , in terms of the material properties:

 1   12
λ + 2µ 2 E(1 − ν)
c1 = = (7.49)
ρ ρ(1 + ν)(1 − 2ν)
  12   12
µ E
c2 = = (7.50)
ρ 2ρ(1 + ν)

where for completeness the expressions in terms of Young’s modulus (E) and Poisson’s ratio (ν)
are included.

7.2.3 Plane waves in a Two-Dimensional Space

It is usually assumed in multilayered plate models that the wavelengths are significantly smaller
than the width of the plate and of the wave fields and therefore that a plane strain analysis is valid.
The coordinate system may then be reduced to the plane defined by the direction of propagation of
the waves and the normal to the plate. For convenience here, the plane is defined by x3 parallel to
the plate and x1 normal to the plate. Figure 7.1 illustrates the geometry of the plate system, using
the coordinate system X, Y, Z which is used by Disperse. Thus x1 , x2 , x3 correspond to X, Y, Z
respectively. Note that the layer interfaces and plate boundaries which are shown in the figure are

Disperse version 2.0.20a


7.2 Wave Propagation in Isotropic Flat Plate Structures 99

not yet considered at this stage. For plane strain there is no variation of any quantity in the x2
direction (∂/∂x2 = 0). Furthermore, for simplicity, the derivation of the model is restricted here
to waves whose particle motion is entirely in the plane (u2 = 0), thus excluding Love [50] modes.
The modelling of Love modes requires only minor additions to the derivations, and these are of
course included in the implementation in Disperse. The Love type modes are entirely uncoupled
from the Lamb type modes and so it is permissible to study them separately as is done here. The
addition of the equations for the Love type modes increases the order of the matrices from 4 to
6 (for the general solution there are thus 6 partial waves rather than 4 as shown here). Further
information about the Love type modes may be found in the seismological literature where they
have received considerable attention [26] [91] [90] [89] [87].

From equation 7.48 the displacements of longitudinal and shear waves are now given by:


k1
uL = ∇φ =  0  AL ei(k·x−ωt) (7.51)
k3
 ∂     
∂x1 0 k3
∂  ×  H3  =  0  AS ei(k·x−ωt)
uS = ∇ × H =  ∂x 2
(7.52)
∂ 0 k1
∂x 3

in which the vector potential H points in the x2 direction so that all particle motion is in the x1 x3
plane.

7.2.4 The Superposition of Plane Waves in a Layered Plate

The development of a model for wave motion in multilayered plates is achieved by the super-
position of longitudinal and shear bulk waves and the imposition of boundary conditions at the
interfaces between the layers. At each interface, it is sufficient to assume eight waves: longitudinal
(dilatational) and shear (equivoluminal) waves arriving from ’above’ the interface and leaving ’be-
low’ the interface (L+, S+) and, similarly, longitudinal and shear waves arriving from below the
interface and leaving above the interface (L-, S-). There are thus four waves in each layer of the
multilayered plate (Figure 7.1). Snell’s law requires that for interaction of the waves they must
all share the same frequency and spatial properties in the x3 direction at each interface. It follows
that all displacement and stress equations have the same ω and the same k3 (= ξ) component of
wavenumber (the ’plate wavenumber’), being the projection of the wavenumber of the bulk wave
onto the interface. All field equations for all locations in all layers therefore contain the following
factor, F , which is an invariant of the system:

F = eξx3 −ωt (7.53)

This constrains the angles of incidence, transmission and reflection of homogeneous bulk waves in
the layers according to their bulk wave velocities, by the relationship:

ξ 1 sin(θL ) sin(θS )
= = = (7.54)
ω cph c1 c2

where θL and θS are the angles at which longitudinal and shear bulk waves propagate with respect
to the normal to the layers (x1 direction). The invariant cph is the projection of the bulk wave

Disperse version 2.0.20a


7.2 Wave Propagation in Isotropic Flat Plate Structures 100

velocities in the x3 direction and will be the phase velocity of the propagating waves described
by modal solutions. The k1 components of the bulk waves in each layer may also be expressed in
terms of the plate wavenumber (ξ = k3 ) and the bulk wave velocities for the material, c1 and c2 :

1/2
ω2

k1L± = ± − ξ2
c21
1/2
ω2

k1S± = ± − ξ2 (7.55)
c22

Again the + and − signs denote waves travelling in directions with positive (’downward’) and
negative (’upward’) x1 components respectively. If ω 2 /c21 is greater than ξ 2 , then k1 is real and
the wave is homogeneous and travels at some non-zero angle with respect to the x3 direction. If
ω 2 /c21 is less than ξ 2 , then k1 is entirely imaginary and the wave is inhomogeneous, or ’evanescent’,
propagating in the x3 direction and decaying in amplitude in the x1 direction.

Thus the displacements and stresses at any location in a layer may be found from the amplitudes
of the bulk waves, using the field equations:

For longitudinal bulk waves:

u1 = ±ζ1 AL± F e±iζ1 x


u3 = ξAL± F e±iζ1 x
σ11 = iρ(ω 2 − 2c22 ξ 2 )AL± F e±iζ1 x
σ22 = iρω 2 (1 − 2c22 /c21 )AL± F e±iζ1 x
σ33 = iρ(ω 2 − 2c22 ζ1 2 )AL± F e±iζ1 x
σ13 = 2iρc22 ξζ1 AL± F e±iζ1 x
σ12 = σ23 = 0

1/2
where ζ1 = ω 2 /c21 − ξ 2

For shear bulk waves:

u1 = −ξAS± F e±iζ2 x
u3 = ±ζ2 AS± F e±iζ2 x
σ11 = ∓2iρc22 ξζ2 AS± F e±iζ2 x
σ22 = 0
σ33 = −σ11
σ13 = iρ(ω 2 − 2c22 ξ)AS± F e±iζ2 x
σ12 = σ23 = 0 (7.56)

1/2
where ζ2 = ω 2 /c22 − ξ 2

The displacements and stresses at any location in a layer may therefore be found by summing the
contributions due to the four wave components in the layer. For a multilayered system, the field

Disperse version 2.0.20a


7.3 Cylindrical Wave Propagation in Isotropic Materials 101

quantities of interest are those which must be continuous at the interfaces: the two displacement
components u1 and u3 , the normal stress σ11 and the shear stress σ13 . Making the substitutions
for convenience,

gζ1 = eiζ1 x1 gζ2 = eiζ2 x1 (7.57)

and omitting the common factor, F , the field quantities in a layer are thus expressed by the matrix
equation:

 
  ζ1 gζ1 − gζζ1 −ξgζ2 − gξζ  
u1 ξ
1 2  AL+
− gζζ2

 u3   ξgζ1 ζ2 gζ2
gζ1   AL−
 
2
 σ11  = 
   iρ(ω 2 −2c22 ξ 2 ) 2iρξc22 ζ2

 iρ(ω − 2c22 ξ 2 )gζ1
2
−2iρξc22 ζ2 gζ2  AS+
 
gζ1 gζ2
σ13 AS−
 
−2iρξc22 ζ1 iρ(ω 2 −2c22 ξ 2 )
2iρξc22 ζ1 gζ1 gζ1 iρ(ω 2 − 2c22 ξ 2 )gζ2 gζ2
(7.58)

In summary, the matrix in equation (7.58) is the field matrix, describing the relationship between
the wave amplitudes and the displacements and stresses at any location in any layer. Its coefficients
depend on the through-thickness position in the plate (x1 ), the material properties of the layer
at this position (ρ, c1 and c2 ), the frequency (ω), and the invariant plate wavenumber (ξ = k3 ).
The origin of the x1 coordinate may be placed arbitrarily and may even be different for each layer
because phase differences between layers can be accounted for by the phase of the complex wave
amplitudes. The field matrix will be abbreviated here to [D].

7.3 Cylindrical Wave Propagation in Isotropic Materials

7.3.1 Historical Background of Cylindrical Wave Propagation

Cylindrical wave propagation problems were first studied numerically in the late nineteenth cen-
tury. Pochhammer [78] and Chree [18] were the first researchers to mathematically investigate the
propagation of guided waves in a free bar and their names are still associated with the equation
that describes the modes of a solid cylinder.

However, most of the applications of cylindrical wave propagation have occurred much more
recently. In the mid twentieth century, a good deal of research was performed on solid bars
[22, 36, 57, 65, 66, 67, 98]. Much of this work concentrated on the use of rods as acoustic waveguides
for use in delay lines. The early work frequently used shell approximations or limited the scope
of the investigation to axi-symmetric modes. However, the later work fully developed all of the
branches of the complete three dimensional problem. The dispersion curves for a hollow cylinder
were definitively treated by Gazis in 1959 [29, 28]. The derivation for a general case that follows
is largely based on his work.

Leaky cylindrical systems have been much more difficult to model than their free counterparts.
Much of the difficulty comes from the need to calculate complex Bessel functions, which until
recently was almost impossible. Early work concentrated only on non-leaky modes that could
exist in a system that would otherwise be leaky [98]. However, more recent work has been able to
model true leaky modes [43, 44, 83, 94, 99]. In addition, recent work [61, 6, 5] has looked at wave

Disperse version 2.0.20a


7.3 Cylindrical Wave Propagation in Isotropic Materials 102

propagation in transversely isotropic rods that are immersed in a fluid. This work builds upon
the solutions for a free transversely isotropic rod first developed by Mirsky [58].

All of these previous solutions to wave propagation in leaky cylindrical systems solve a specific
elastic problem. The derivation which is implemented in Disperse is based on the approach taken
by Gazis [29] expanding the solution to work for a general system. This can consist of an arbi-
trary number of layers, elastic or visco-elastic materials, axisymmetric or non-axisymmetric wave
propagation and the possibility of leakage into a surrounding solid medium.

7.3.2 Assumptions and Limitations

In developing wave propagation models, certain assumptions will need to be made that limit the
scope of systems that the model can handle. These assumptions are explained below.

This model assumes that the cylindrical system is axi-symmetric and infinitely long. The material
properties may only vary in the radial direction and all of the variations must occur as instanta-
neous changes at the boundaries of discrete layers. The boundaries between these layers will be
assumed to be perfect. These assumptions imply that local phenomena, such as defects, cannot
be modelled. It should be noted that although the geometry of the system must be axisymmetric,
non-axisymmetric wave propagation will be permitted.

All of the materials are assumed to be isotropic for the derivation presented here (NB Disperse can
model the special case of transversely isotropic materials with the cylindrical axis parallel to the
axis of isotropy; for details of the theory of that model see [61, 6, 5]). Both elastic and visco-elastic
isotropic materials may be modelled, according to the description given in section7.1.

The waves will be assumed to be continuous, the frequency real, and the energy finite. The
assumptions of a continuous wave and real frequency indicate that transitory effects cannot be
directly incorporated into the model. The assumption that the system contains finite energy
implies that energy is not being added to the system from external sources, although energy may
be lost due to visco-elastic effects and waves leaking into an infinitely large surrounding medium.
A wave is continually created at one end of the cylindrical structure and allowed to propagate.
Solutions will only be sought for guided waves which propagate axially.

7.3.3 Boundary Conditions

This model assumes that the system is comprised of infinitely long, concentric cylinders of homo-
geneous, isotropic materials. Therefore, the boundaries between layers are defined by a surface
with a constant radius, extending infinitely in the ẑ direction. Three different types of interfaces
are considered, solid-solid, solid-vacuum, and solid-liquid.

It is also assumed that a perfect boundary exists between consecutive layers. This assumption
implies that the the displacements and the stresses are continuous at the interface. All three
displacements, r̂, θ̂, and ẑ, must be continuous or there would have to be void at the boundary. In
addition, the normal stress, σrr , and the two tangential stress along the interface, σrz and σrθ will
be transmitted across the boundary and must be continuous. The in-plane stresses, such as σzz ,
do not need to be continuous across the interface because they are not coupled from one medium
to the next. These boundary conditions can be summarised as,

ur + = ur −

r=a r=a

Disperse version 2.0.20a


7.3 Cylindrical Wave Propagation in Isotropic Materials 103


uθ = uθ

+ −
r=a r=a
uz = uz

+ r=a−
r=a
σrr = σrr

+ −
r=a r=a
σrθ = σrθ

+ −
r=a r=a
σrz + = σrz − (7.59)

r=a r=a

for a boundary of radius, r, equal to a, where the + superscript indicates that interface is ap-
proached from the inner material (with respect to r) and the − indicates that it is approached
from the outer material. It is also possible to define other types of solid-solid interfaces to model
the behaviour of non-perfect boundaries. These are not included in this model but in some cases
may be simulated by adding thin interface layers with reduced stiffnesses.

The boundary conditions change when a layer is bordered by a vacuum. In this case, the displace-
ments are no longer constrained at the surface. In addition, since there is no medium on the other
side of the interface to support a stress field, the normal and two tangential stresses that used to
be continuous across the interface must now equal zero. For a boundary at radius, r, equal to a,
these conditions may be written,

σrr = 0

r=a
σrθ = 0

r=a
σrz = 0. (7.60)

r=a

The third type of interface that is important for our modelling is the boundary between a liquid
and solid layer. In this model, we assume perfect liquids that do not support shear. Therefore,
instead of the tangential stresses being transferred across the interface as they are for solid-solid
boundaries, the tangential stresses must be zero at the interface. In addition, there will no longer
be constraints on the displacements which are tangential to the interface because the two materials
are free to slip over each other. However, the displacement and compressional stress normal to
the boundary must still be continuous across the interface to avoid the formation of voids. These
conditions are summarised as,

ur = ur

r=asolid r=aliquid

σrr = σrr

r=asolid r=aliquid

σrθ = 0

r=a
solid
σrz = 0. (7.61)

r=asolid

7.3.4 Waves in Finite Layers

Equation 7.9 describes how a wave will propagate in an infinitely large medium. This section
expands on this knowledge to solve these equations for a specific coordinate system and to inves-
tigate how waves will propagate when boundaries are added. The first part of the section explores
possible solutions of equation 7.9 in cylindrical coordinates. These solution types are then used to

Disperse version 2.0.20a


7.3 Cylindrical Wave Propagation in Isotropic Materials 104

determine the displacement stress fields that arise because of the propagation of the wave. Finally,
these fields are combined with the boundary conditions to create a layer matrix that may be used
to describe a complete system.

Proposed Solution for Cylindrical Geometry

Since the Helmholtz equation (7.7) is separable, the solution may be divided into the product of
functions of each of the spatial dimensions. If we assume a harmonically oscillating source, the
solutions to equation 7.9 will each be of the form,

φ, H = Ξ1 (r)Ξ2 (θ)Ξ3 (z)ei(k·r−ωt) (7.62)

where k is the complex wavenumber vector, r is the position vector, and ω the circular frequency.
Assuming that the wave does not propagate in the r̂ direction and assuming that the displacement
field does not vary in the ẑ or θ̂ direction except for the harmonic oscillation described by the
real part of the vector wavenumber and the attenuation described by the imaginary part , we can
rewrite this equation as,
φ, H = Ξ2 (r)ei(kθ θ+ξz−ωt) (7.63)
where kθ is the angular wavenumber component, ξ is the component of the complex vector
wavenumber in the ẑ direction, which will henceforth be referred to simply as the wavenum-
ber. The angular wavenumber component, kθ must be an integer, n. Otherwise the solution
would be different at θ and θ + 2π and the solution would not be unique. Phase angles may be
arbitrarily added to each of the functions, since the added phases will be compensated for by the
coefficients of the functions themselves. Choosing the phase angles to match those used by Gazis
yields the following expressions for the components of the Helmholtz potentials in equation 7.7,
for cylindrical waves propagating in the ẑ direction,

φ = f (r)ei(nθ+ξz−ωt)
Hr = −gr (r)ei(nθ+ξz−ωt)
Hθ = −igθ (r)ei(nθ+ξz−ωt)
Hz = −ig3 (r)ei(nθ+ξz−ωt) (7.64)

Another possible solution results in sines and cosines for the θ dependence. This formulation
more closely follows Gazis’ formulation and more clearly demonstrates the wave’s behaviour for
torsional and longitudinal modes, when n = 0. The choice of the phase angles and the sine or
cosine function is constrained by the expressions for the displacements, which are developed below.

φ = f (r) cos(nθ)ei(ξz−ωt)
Hr = −igr (r) sin(nθ)ei(ξz−ωt)
Hθ = −igθ (r) cos(nθ)ei(ξz−ωt)
Hz = g3 (r) sin(nθ)ei(ξz−ωt) (7.65)

Substituting the equation for φ from equation 7.64 into equation 7.9

c21 ∇2 φ = ∂ 2 φ/∂t2

yields,
d2 f n2 ω2
  
1 df
+ − − − ξ2 f = 0. (7.66)
dr2 r dr r2 c1 2

Disperse version 2.0.20a


7.3 Cylindrical Wave Propagation in Isotropic Materials 105

The series solution of this differential equation is a Bessel function, whose differential operator is
defined as,  2  2 
∂ 1 ∂ n
Bn,r = + − −1 . (7.67)
∂r2 r ∂r r2
Therefore,
Bn,ζ1 r [f ] = 0 (7.68)
is a valid solution of for the potential function φ when ζ1 2 = ω 2 /c21 − ξ 2 . In equation 7.68, ζ1
corresponds to the radial component of the vector wave number and helps describe the behaviour
of the wave as it radiates.

Substitution of the vector potential, H from 7.64 into equation 7.9 produces three coupled scalar
equations, as revealed by the Laplacian,

c22 ∇2 H = ∇(∇ · H) − ∇ × (∇ × H) = ∂ 2 H/∂t2


   
Hr 2 ∂Hθ Hθ 2 ∂Hr
r̂ ∇2 Hr − 2 − 2 + θ̂ ∇2 Hθ − 2 + 2 + ẑ ∇2 Hz (7.69)

=
r r ∂θ r r ∂θ

Separating equation 7.69, results in the following equations,

d2 g3
 2  2 
1 dg3 n ω 2
+ − − −ξ g3 = 0 (7.70)
dr2 r dr r2 c2 2
d2 gr 1 dgr 1 2 2 ω2
+ + (−n g r + 2ngθ − g r ) − ξ gr + gθ = 0 (7.71)
dr2 r dr r2 c2 2
d2 gθ 1 dgθ 1 2 2 ω2
+ + (−n gθ + 2ngr − gθ ) − ξ gθ + gr = 0 (7.72)
dr2 r dr r2 c2 2

Equation 7.70 is uncoupled and has a solution of the form,

Bn,ζ2 r [g3 ] = 0 (7.73)

where ζ2 2 = ω 2 /c22 − ξ 2

Adding and subtracting equations 7.71 and 7.72 leads to the following solutions of the wave
potential functions,

Bn+1,ζ2 r [gr − gθ ] = 0
Bn+1,ζ2 r [gr + gθ ] = 0 (7.74)

Gauge invariance tells us that one of the potentials can be set to zero without loss of generality.
In other words the displacement field corresponding to an equivoluminal potential can be derived
by a combination of the other two equivoluminal potentials. So we set g2 = 0 such that

gr = −gθ = g1 (7.75)

and we are left with three independent potential functions, f (r), g1 (r), and g3 (r), which each
satisfy the Bessel differential operator.

Choice of Bessel functions Each of the three of independent potential functions, f (r), g1 (r),
and g3 (r), satisfy the Bessel differential operator. However, there are several different combinations
of Bessel functions that may be used to satisfy this differential operator. In order to fully satisfy
the operator, a pair of linearly independent Bessel functions is required, which correspond to

Disperse version 2.0.20a


7.3 Cylindrical Wave Propagation in Isotropic Materials 106

Table 7.1: Substitutions that should be made and criteria that should be used for the selection of
the type of Bessel functions to be used for the cylindrical layer matrix.
when Vp ph > c1 when c1 > pVph > c2 when c1 > pc2 > Vph
ζ1 = pζ1 2 ζ1 = p−ζ1 2 ζ1 = p−ζ1 2
ζ2 = ζ2 2 ζ2 = ζ2 2 ζ2 = −ζ2 2
γ1 = 1 γ1 = −1 γ1 = −1
γ2 = 1 γ2 = 1 γ2 = −1
Zn (ζ1 r) = Jn (ζ1 r) Zn (ζ1 r) = In (ζ1 r) Zn (ζ1 r) = In (ζ1 r)
Wn (ζ1 r) = Yn (ζ1 r) Wn (ζ1 r) = Kn (ζ1 r) Wn (ζ1 r) = Kn (ζ1 r)
Zn (ζ2 r) = Jn (ζ2 r) Zn (ζ2 r) = Jn (ζ2 r) Zn (ζ2 r) = In (ζ2 r)
Wn (ζ2 r) = Yn (ζ2 r) Wn (ζ2 r) = Yn (ζ2 r) Wn (ζ2 r) = Kn (ζ2 r)

inward and outward propagating waves. The ’standard’ pair of functions that satisfy the Bessel
differential operator are the Bessel function of the first kind (Jn ) and the second kind (Yn ).
These functions represent a pair of standing waves. The modified Bessel functions, In and Kn ,
which represent decaying waves, may also be used. The third valid combination of functions are
(1) (1)
the Hankel functions of the first and second kind (Hn and Hn ), which represent waves that
propagate toward or away from the origin.

Since this wave propagation model studies guided waves, which can be considered through thick-
ness (radial) resonances that propagate in the axial direction, standing or decaying waves best
represent the solution for internal layers. The choice between the unmodified or modified Bessel
functions depends on the argument of the Bessel functions. For a non-attenuating system, the
arguments of the Bessel functions, ζ1 r and ζ2 r, will either be purely real or purely imaginary.
When using Bessel functions that take complex arguments, the Bessel functions Jn and Yn could
be used for both the real and imaginary cases, however it is more stable (and elegant) to use Jn
and Yn when the arguments are real and use the modified Bessel functions, In and Kn , when
the arguments would be imaginary. Since both of the unmodified Bessel functions (Jn and Yn )
increase exponentially along the imaginary axis, the effect of the inward and outward propagating
waves cannot be clearly separated and the solution becomes unstable as the arguments become
large. This problem is analogous to the “large fd” problem that is discussed for plates [51] and is
especially important for large radius pipes. However the modified Bessel functions separate with
(1)
In (z) = (−i)n Jn (iz) increasing and Kn (z) = (1/2)πi(n+1) Hn (iz) decreasing as the argument
grows. This leads to a more stable solution. Since the modified Bessel functions solve a slightly
different equation and since they have different reoccurance relationships than the unmodified
Bessel functions, some changes need to be made to the definition of ζ1 , ζ2 . In addition, two pa-
rameters, γ1 , and γ2 , have been defined to account for the differences in reoccurance relationships
that are employed when derivatives of the functions are taken. The first parameter, γ1 , relates to
reoccurance relationships that involve ζ1 and γ2 relates to those that involve ζ2 . The substitutions
that need to be made, as well as the criteria for when these substitutions should be made, are
collected in Table 7.1. This table introduces the functions Zn and Wn which will be used in the
rest of the derivation to represent inward (Zn ) and outward (Wn ) oriented Bessel functions. The
(1)
function Zn can be substituted by Jn , In , or Hn and the function Wn takes the place of Yn ,
(2)
Kn , or Hn .

Leakage into a surrounding medium For systems which leak energy into a surrounding
medium, such as a bar embedded in an infinitely large medium, an appropriate Hankel function
must be chosen for the outward propagating wave in the layer that represents the infinite sur-
rounding medium. The Y or K Bessel functions, which represent outgoing waves in a layer that
has finite size, portray standing or decaying waves. These solutions will not correctly describe an

Disperse version 2.0.20a


7.3 Cylindrical Wave Propagation in Isotropic Materials 107

infinitely thick layer. For our choice of the direction of the wave, ei(ξz−ωt) , we choose the Hankel
(1)
function of the first kind Hn , which decays at infinity, to describe the outward propagating wave
(2)
in the infinitely thick surrounding medium. The Hankel function of the second kind, Hn , which
represents the incoming waves, increases as its argument approaches infinity. If this function is
included in the solution, the assumption that the energy is finite would be violated because there
would be constant energy input along the entire infinite length of the system. Instead, the three
incoming partial wave amplitudes (L− , SV− , and SH− ) are set to zero. The Hankel functions
follow the same reccurance relationships as the unmodified Bessel functionsp of the first and second
kind
p (J and Y). Therefore, γ 1 and γ 2 must always be 1, ζ1 always equals ζ1 2 , ζ2 always equals
2
ζ2 .

The cylindrical core of the system may also be modelled as a semi-infinite half-space. The core
only has three partial waves (all of which propagate inward), as opposed to the interior layers
that contain six partial waves each (three inward and three outward). For the core, the outward
travelling waves are represented by the inward travelling waves from the other side of the bar.
These waves are normally represented by the J or I Bessel functions, which represent standing
resonances through the thickness of the core. However, some cases arise when it is advantageous to
eliminate the transmission of the inward propagating waves from one side of the core to the other.
An example, which is explained in Section 9.2.1, is the modelling of a large radius fluid filled pipe.
If the inward travelling waves are not transmitted across the fluid core, the attenuation values may
better relate to the decrease of the peak amplitude of a wave packet travelling in the pipe. When
a sink is not present, modes of the fluid are excited and the attenuation of the mode in the pipe
is reduced because of retransmission of energy from the water back into the pipe. However, this
retransmission may occur after a delay in time, which would cause the transmitted wave packet
to appear to be dispersive and the maximum amplitude will decay more rapidly than the total
energy in the mode. Some other cases when the transmission of bulk waves through a liquid core
may be considered undesirable include viscous fluids and partially filled pipes.

In order to eliminate the transmission of the partial waves through the liquid core, the normal
wave solution that uses the J or I Bessel function can be replaced by a solution that incorporates
(2)
the Hankel function of the second kind, Hn . The Hankel function of the second kind has a sink
at the origin which absorbs the incoming bulk waves. When using the Hankel function of the
second kind, the parameters γ1 and γ2 should be set to one and the value and ζ1 and ζ2 must be
defined in the same manner as for Hankel functions of the first kind.

Choice of the phase for ζ1 and ζ2 The phase of the arguments to the Bessel functions, ζ1 r
and ζ2 r, strongly affects the behaviour of the solution and must therefore be chosen with care.
Since ζ1 and ζ2 are defined by the value of their squares, they may take on two different values
that are 180 degrees out of phase. The choice of roots depends on the type of Bessel function and
whether the partial wave is homogeneous or inhomogeneous. Homogeneous waves propagate both
along the direction of propagation of the guided wave and normal to the interface between layers.
On the other hand, imhomogeneous waves propagate along the direction of propagation of the
guided wave, but die away exponentially in the direction normal to the interface between layers.
When there is no attenuation, it is easy to determine which waves are homogeneous and which
are inhomogeneous. If the argument to the Bessel function is real then the wave is homogeneous,
but if the argument to the Bessel function is imaginary, the partial wave is inhomogeneous. When
these conditions apply may be extracted from the definition of ζ1 ,
s 
2  2
ω ω
ζ1 = − (7.76)
v1 vph

Disperse version 2.0.20a


7.3 Cylindrical Wave Propagation in Isotropic Materials 108

Table 7.2: Criteria for the the choice of phase for the arguments of the Bessel functions depending
on the type of Bessel function and the type of wave (homogeneous or inhomogeneous).
Bessel Function Phase (φ) for homogeneous Phase (φ) for inhomogeneous
wave wave
Jn , Yn , and Hn e−iπ/2 < φ < eiπ/2 ei0 < φ < eiπ
In and Kn ei0 < φ < eiπ e−iπ/2 < φ < eiπ/2

where the expression ω/vph has been substituted for the axial wavenumber, ξ. Whenever the
phase velocity of the guided wave, vph , is greater than the bulk velocity, v1 , ζ1 2 is positive and
ζ1 is real. Conversely, when the phase velocity is below the bulk velocity, ζ1 2 is negative and ζ1
is imaginary. With attenuation present, the arguments of the Bessel functions will be complex.
In this case, a homogeneous wave can be said to exist when the real part of ω 2 /cbulk is greater
than the real part of ξ 2 . This condition is also usually true when the phase velocity of the wave
is greater than the appropriate bulk velocity (c1 for ζ1 and c2 for ζ2 ). Otherwise the partial wave
will be inhomogeneous. Since ζ1 and ζ2 represent the wave number in the direction normal to
the interface between the layers, it is necessary to choose their phase so that the behaviour is
consistent with the nature of the homogeneous or inhomogeneous waves. Both the propagation
of homogeneous waves and the decay of inhomogeneous waves should be away from the interface.
Tabulating this information for the different types of Bessel functions leads to the selection criteria
found in Table 7.2.

Displacement Field

We are now ready to apply the information that we know about the solution in order to calculate
the displacements at an arbitrary location in the medium. Recalling from equation 7.7 that
u = ∇φ + ∇ × H and the vector operations in cylindrical coordinates given in appendix A.8,

∂φ 1 ∂Hz ∂Hθ
ur = + −
∂r r ∂θ ∂z
1 ∂φ ∂Hr ∂Hz
uθ = + −
r ∂θ ∂z ∂r
∂φ 1 ∂ 1 ∂Hr
uz = + (rHθ ) − (7.77)
∂z r ∂r r ∂θ

leads to the displacement field in terms of the potential functions of equation 7.64,

 n 
ur = f 0 + ξg1 + g3 ei(nθ+ξz−ωt)
n r 
uθ = i f − ξg1 + g30 ei(nθ+ξz−ωt)
 r 
(n + 1) 0
uz = i ξf + g1 + g1 + 0 ei(nθ+ξz−ωt) (7.78)
r

where the prime notation indicates the derivative with respect to the radius, r. The expressions
for the displacements may also be expressed in sine and cosine notation,

Disperse version 2.0.20a


7.3 Cylindrical Wave Propagation in Isotropic Materials 109

 n 
ur = f 0 + ξg1 + g3 cos(nθ)ei(ξz−ωt)
n r 
uθ = − f − ξg1 + g30 sin(nθ)ei(ξz−ωt)
 r 
(n + 1) 0
uz = i ξf + g1 + g1 cos(nθ)ei(ξz−ωt) . (7.79)
r

Equation 7.79 clearly shows that longitudinal modes (n = 0) have no θ̂ displacement and have a
constant displacement profile around the circumference.

Stress Field

Knowing the expressions for the displacements and the relations between the displacement and
the stresses, we can now determine expressions for the stresses. Hooke’s Law defines the stress as
τii = λ∆δij + 2µeij , where in cylindrical coordinates,

∂ur 1 ∂uθ ur ∂uz


rr = θθ = + zz =
∂r
 r ∂θ r ∂z
1 ∂ur ∂uz
rz = +
2 ∂z ∂r
 
1 ∂ uθ  1 ∂ur

rθ = r +
2 ∂r r r ∂θ
 
1 ∂uθ 1 ∂uz
θz = + (7.80)
2 ∂z r ∂θ

and

σrr = λ∆ + 2µrr
σrz = 2µrz
σrθ = 2µrθ (7.81)

where ∆ is the dilatation given by

∆ = ∇2 φ = −(ω 2 /c21 )f ei(nθ+ξz−ωt) = −(ζ1 2 + ξ 2 )f ei(nθ+ξz−ωt)

Applying these expressions to the displacements in equation 7.78, the stresses that are specified
in the boundary conditions become

 
λ 2 2 00 0 n 2n 0
σrr = µ − (ζ1 + ξ )f + 2f + 2ξg1 − 2 2 g3 + g
µ r r 3
 
2n 2n 0 (n + 1)ξ 0 00 2
σrθ = iµ − 2 f + f + g1 − ξg1 + 2g3 + ζ2 g3
r r r
   
nn+1 n n
σrz = iµ 2ξf 0 + + ξ 2 − ζ2 2 g1 + g10 + ξg3 (7.82)
r r r r

Disperse version 2.0.20a


7.3 Cylindrical Wave Propagation in Isotropic Materials 110

In calculating σrθ and σrz the substitutions, taken from the Bessel equation,

1  n2 
− g30 = g300 + ζ2 2 − 2
r  r
(n + 1)2

1 0
g100 = − g1 + − ζ2 2
(7.83)
r r2

were used to match the form of the expressions given by Gazis[29]. Please note that the g1 term
of σrz does not match Gazis; he also incorrectly multiplies ξ 2 − ζ2 2 by n/r. This error is not
propagated through to the other results.

Material Layer Matrix

The expressions for the displacement and stress fields in equations 7.78 and 7.82 are expressed in
terms of the potential functions. The solutions for these potential functions that are given in Sec-
tion 7.3.4 must be substituted into these equations before the actual stress field can be calculated.
These substitutions provide six equations in terms of six unknown partial wave amplitudes (three
for incoming waves and three for outgoing waves) that correspond to the coefficients of the Bessel
functions that are used in the solution. The use of these partial waves is shown in Figure 7.2,
where six bulk waves are present in each interior layer. The partial waves amplitudes in Figure
7.2 are labelled by their Cartesian equivalents, L, SH, and SV, which stand for Longitudinal waves
(equivalent to the solutions of f (r)), Shear Vertical waves (g1 ), and Shear Horizontal waves(g3 ).

We can assemble a 6 by 6 matrix from the six equations for the displacements (eqn. 7.78) and
pertinent stresses (eqn. 7.82) in terms of the amplitudes of the 6 partial waves. The resulting
matrix is called a material layer matrix and expresses the equation
   
 ur 
  
 L+  
 uθ   L− 

 
 
 

   
uz SV+
   
= [D] (7.84)

 σrr 
 
 SV−  
σrθ  SH+ 

  
 

 
 
 

σrz SH−
   

where [D] is the material layer matrix. This matrix expresses all of the displacements and stresses
that are required to evaluate the boundary conditions described in Section 7.3.3 and will be used in
following sections to assemble an equation that represents the entire system. The layer matrix has
been organised so that the columns are ordered, outward propagating f , inward propagating f ,
outward propagating g1 , inward propagating g1 , outward propagating g3 , and inward propagating
g3 , and the rows are ordered, in-plane displacement, uz , normal displacement, ur , shear displace-
ment, uθ , normal stress, σrr , shear stress (in direction of propagation), σrz , and shear stress
perpendicular to the direction of propagation, σrθ . This order was chosen to match a pre-existing
layer matrix for flat plates. It is shown below in graphical form, where the + subscripts denote
outward propagating waves (downward direction for the plate case), the − subscripts denote the
inward propagating waves (upward direction for the plate case) and the partial wave amplitudes
are denoted L, SV , and SH.

Disperse version 2.0.20a


7.3 Cylindrical Wave Propagation in Isotropic Materials 111

L+ L− SV + SV − SH + SH −
uz D(1, 1) D(1, 2) ··· D(1, 6)
ur D(2, 1)
uθ (7.85)
.. ..
σrr . .
σrz
σrθ D(6, 1) ··· D(6, 6)

For a cylindrical isotropic material layer, the material layer matrix is given by,

uz σrr
D(1, 2) = i(ξr2 Zn (ζ1 r)) D(4, 2) = ((ξ 2 − β 2 )r2 + 2n(n − 1))Zn (ζ1 r) + 2γ1 αrZn+1 (ζ1 r))
D(1, 4) = i(βr2 Zn (ζ2 r)) D(4, 4) = 2ξβr2 Zn (ζ2 r) − 2ξr(n + 1)Zn+1 (ζ2 r)
D(1, 6) = 0 D(4, 6) = 2n(n − 1)Zn (ζ2 r) − 2γ2 nβrZn+1 (ζ2 r)
2
D(1, 1) = i(ξr Wn (ζ1 r)) D(4, 1) = ((ξ 2 − β 2 )r2 + 2n(n − 1))Wn (ζ1 r) + 2αrWn+1 (ζ1 r)
2
D(1, 3) = i(γ2 βr Wn (ζ2 r)) D(4, 3) = 2γ2 ξβr2 Wn (ζ2 r) − 2(n + 1)ξrWn+1 (ζ2 r)
D(1, 5) = 0 D(4, 5) = 2n(n − 1)Wn (ζ2 r) − 2nβrWn+1 (ζ2 r)

ur σrz
2
D(2, 2) = nrZn (ζ1 r) − γ1 αr Zn+1 (ζ1 r) D(5, 2) = i(2nξrZn (ζ1 r) − 2γ1 ξαr2 Zn+1 (ζ1 r))
2
D(2, 4) = ξr Zn+1 (ζ2 r) D(5, 4) = i(nβrZn (ζ2 r) + (ξ 2 − β 2 )r2 Zn+1 (ζ2 r))
D(2, 6) = nrZn (ζ2 r) D(5, 6) = i(nξrZn (ζ2 r))
D(2, 1) = nrWn (ζ1 r) − αr2 Wn+1 (ζ1 r) D(5, 1) = i(2nξrWn (ζ1 r) − 2ξαr2 Wn+1 (ζ1 r))
2
D(2, 3) = ξr Wn+1 (ζ2 r) D(5, 3) = i(γ2 nβrWn (ζ2 r) + (ξ 2 − β 2 )r2 Wn+1 (ζ2 r))
D(2, 5) = nrWn (ζ2 r) D(5, 5) = i(nξrWn (ζ2 r))

uθ σrθ
D(3, 2) = i(nrZn (ζ1 r)) D(6, 2) = i(2n(n − 1)Zn (ζ1 r) − 2γ1 nαrZn+1 (ζ1 r))
2
D(3, 4) = i(−ξr Zn+1 (ζ2 r)) D(6, 4) = i(−ξβr2 Zn (ζ2 r) + 2ξr(n + 1)Zn+1 (ζ2 r))
2
D(3, 6) = i(nrZn (ζ2 r) − βγ2 r Zn+1 (ζ2 r))D(6, 6) = i((2n(n − 1) − β 2 r2 )Zn (ζ2 r) + 2γ2 βrZn+1 (ζ2 r))
D(3, 1) = i(nrWn (ζ1 r)) D(6, 1) = i(2n(n − 1)Wn (ζ1 r) − 2nαrWn+1 (ζ1 r))
D(3, 3) = i(−ξr2 Wn+1 (ζ2 r)) D(6, 3) = i(−ξγ2 βr2 Wn (ζ2 r) + 2ξr(n + 1)Wn+1 (ζ2 r))
D(3, 5) = i(nrWn (ζ2 r) − βr2 Wn+1 (ζ2 r)) D(6, 5) = i((2n(n − 1) − β 2 r2 )Wn (ζ2 r) + 2βrWn+1 (ζ2 r))

(7.86)

where Zn () and Wn () represent incoming and outgoing Bessel functions, n is the circumferential
order, r is the radius, ξ is the axial wavenumber (the wavenumber in the direction of propagation),
µ is one of Lamé’s constants, ζ1 2 is (ω/v1 )2 − ξ 2 , ζ2 2 is (ω/v2 )2 − ξ 2 , v1 and v2 are the longitudinal

Disperse version 2.0.20a


7.3 Cylindrical Wave Propagation in Isotropic Materials 112

and shear bulk velocities, and γ1 and γ2 account for differences in the recurrence relationships
of different Bessel functions as specified in Table 7.1. In the formation of the above matrix,
the displacements have been multiplied by r2 and the stresses by r2 /µ to simplify the resulting
expressions and avoid dividing by the radius. For single layered problems the scaling factors
can be omitted during the formation of the global matrix (see Section 7.5) and the solution is
not affected. However, for multilayered problems, Lamé’s constant, µ, must be included in the
expressions for the stresses. In addition, the phase factors, −i, are normally included so that the
relative phases of the various components is expressly shown. With the exception of the scaling
factors, the expressions for the stresses that appear in the above equation closely match those that
Gazis derives[29]. However, it should be noted that the ζ1 1 in the first part of Gazis expression
for the inward longitudinal component of σrz (c31 in his derivation) should actually be an a.

In the above equation, Zn represents the inward directed waves, which can be expressed by the
(2)
Bessel functions, Jn , In , or Hn , depending on the solution type and the phase of the argument.
(1)
Wn represents the outward directed wave and can be expressed by Yn , Kn , or Hn . Since the
(1) (2)
reoccurance relations for In , and Kn , are different from the relations for Jn , Yn , Hn ,or Hn , the
parameters γ1 and γ2 have been introduced into the equations. These parameters are −1 when
the Bessel functions In and Kn are used and are 1 when any of the other Bessel functions are
used. The parameters have been divided into two parts, γ1 and γ2 , the first of which relates to
the longitudinal component and the second of which relates to the shear component. Section 7.3.4
discusses under what conditions the different types of Bessel functions should be used.

Liquids

We treat liquids quite differently, according to whether or not they include viscous losses.

If the liquid is viscous, then we model it using longitudinal and shear bulk velocities, just as for
solids, as explained in Section 7.1.2.

If the liquid has no viscosity, then it is often possible to obtain accurate results for flat plate
models by representing the liquid as a solid with a very low shear velocity (see [51]). However this
can never be done for cylindrical cases because a small shear velocity forces the arguments of the
Bessel functions to be too large for the numerical routines to be able to calculate them. However
the material matrix can easily be simplified for the case of an ideal liquid layer by eliminating the
shear components. To model an ideal inviscid liquid, only the scalar potential φ needs to be used
since the stress now becomes a single dimensional, scalar field. In addition, the material property,
µ, which most strongly influences the shear action of the material, is now set to be zero. Thus the
displacements and stresses become,
ur = f 0 ei(nθ+ξz−ωt)
n
uθ = i f ei(nθ+ξz−ωt)
r
uz = iξf ei(nθ+ξz−ωt)
σrr = −λ(ζ1 2 + ξ 2 )f ei(nθ+ξz−ωt)
σrθ = 0
σrz = 0 (7.87)

Since an inviscid liquid does not support shear, the boundary conditions for a solid-liquid interface
are different from those for a solid-solid interface. As expressed in equation 7.61, only the normal
(perpendicular to the axis of the cylinder) displacement and the normal, compressional stress need
to be continuous at the boundary between a liquid and a solid.

Disperse version 2.0.20a


7.4 Cartesian Wave Propagation in Orthotropic Media 113

From these conditions and equation 7.87, we can assemble a material matrix for a cylindrical
inviscid liquid layer that is equivalent to the material layer matrix for a solid layer in equation
7.86. All of the elements of the liquid material matrix are zero, except for the following,

ur
D(2, 1) = nrWn (ζ1 r) − ζ1 r2 Wn+1 (ζ1 r)
D(2, 2) = nrZn (ζ1 r) − ζ1 r2 Zn+1 (ζ1 r)

σrr
D(4, 1) = −ρω 2 r2 Wn (ζ1 r)
D(4, 2) = −ρω 2 r2 Zn (ζ1 r) (7.88)

where both the displacements and the stresses are scaled by 1/r2 .

7.4 Cartesian Wave Propagation in Orthotropic Media

The following section presents a simplified derivation of the wave propagation equations for an
orthotropic material with wave propagation along a principal axis. For this case the “Lamb” type
modes are decoupled from the “SH” type modes. Therefore, the solutions are more simple than
the general anisotropic case for which the waves are all coupled.

For the following derivation, the abbreviated subscript notation used for the stiffness constants [4,
p. 65] is
1 xx
2 yy
3 zz
(7.89)
4 yz, zy
5 xz, zx
6 xy, yx
This notation agrees with that used be Jones (p. 40), but not with that used by Borese (p.250).

7.4.1 Bulk Waves in Anisotropic Media

In order to determine the wavenumbers at which bulk waves will propagate in anisotropic material
we need to use the Christoffel equation [4, page 211], which, for a Cartesian system, is,

     
Γ11 Γ12 Γ12 vx vx
k 2  Γ21 Γ22 Γ22  =  vy  = ρω 2  vy  (7.90)
Γ31 Γ32 Γ32 vx vx
where  
lx 0 0
   0 ly 0 
lx 0 0 0 lz ly  
 0 0 lz 
Γ= 0 ly 0 lz 0 lx [Cij ] 
  (7.91)
 0 lz ly 
0 0 lz ly lx 0  
 lz 0 lx 
ly lx 0

Disperse version 2.0.20a


7.4 Cartesian Wave Propagation in Orthotropic Media 114

and lx , ly , and lz are directional cosines, vx , vy , and vz are the particle velocities, ρ is the density,
ω is the circular frequency, k is the wave number, and Cij are the stiffness constants.

For an arbitrarily oriented orthotropic material (represented by triclinic in this case), in the X-Z
plane this simplifies to [4, p. 393]

C11 ξ 2 + C66 kx2 + 2C16 ξkx − ρω 2 C16 ξ 2 + c26 kx2 + (C12 + C66 )ξkx C15 ξ 2 + c46 kx2 + (C14 + C56 )ξkx
" #
Γ12 C66 ξ 2 + C22 kx2 + 2C26 ξkx − ρω 2 C56 ξ 2 + c24 kx2 + (C25 + C46 )ξkx =0
Γ13 Γ23 C55 ξ 2 + C44 kx2 + 2C45 ξkx − ρω 2
(7.92)

and for an orthotropic material whose axis corresponds to the global X-Y-Z axes (and is therefore
still orthotropic in the propagating coordinate axis Z),

C11 ξ 2 + C66 kx2 − ρω 2


 
(C12 + C66 )ξkx 0
 (C12 + C66 )ξkx C66 ξ 2 + C22 kx2 − ρω 2 0 =0 (7.93)
2 2 2
0 0 C55 ξ + C44 kx − ρω

where kx and ξ(= kz ) are the components of the wavenumber in the x and z directions. We want
to know the kx for any given ξ so that we can match the through-layer resonance with the wave
number of the propagating wave. In the isotropic case, we knew that the wavenumber of the
partial wave was equal to ω/vbulk , for any angle in the layer, so we could just resolve to take the
proper component for our solution. However, for the anisotropic case, the partial wave number
varies with the angle, so it needs to be calculated on the fly by solving this matrix. This matrix
decouples into two equations. Solving the upper sub-matrix leads to the two solutions,


2 −B − B 2 − 4AC
ζ1 =
√2A
−B + B 2 − 4AC
ζ2 2 = (7.94)
2A
where

A = C66 C22
B = −(C66 + C22 )ρω 2 − (C12
2
+ 2C12 C66 − C11 C22 )ξ 2
C = (C11 ξ 2 − ρω 2 )(C66 ξ 2 − ρω 2 ) (7.95)

which are the quasi-longitudinal and quasi-shear waves. The third solution is a pure shear wave
with particle motion normal to the XZ plane. This wave is equivalent to the SH mode in an
isotropic plate and is given by the expression,

ζ3 2 = kx2 = (1/C44 )(ρω 2 − C55 ξ 2 ) (7.96)

7.4.2 Lamb-Type Modes

For the moment we ignore the ζ3 root and concentrate on the “Lamb” type modes.

Disperse version 2.0.20a


7.4 Cartesian Wave Propagation in Orthotropic Media 115

The particle displacements in the medium are equal to the eigenvectors of the problem. Therefore,
substituting into Christoffel’s equation for an orthotropic medium (eqn. 7.93),

C11 ξ 2 + C66 kx2 − ρω 2


 
ux = − uz (7.97)
(C12 + C66 )ξkx

The eigenvectors may be scaled arbitrarily, since they indicate directions and solve a linear equa-
tion. We choose to scale the ζ1 eigenvector by ξ and ζ2 by ζ2 , so the that displacements from this
anisotropic derivation match the Cartesian isotropic case and the different types of materials can
be mixed. In addition, for the upward travelling waves, we reverse the x̂ axis, such that a deriva-
tive with respect to x will reverse sign for the upward propagating waves. Thus, the displacements
can be written,

Quasi-longitudinal (ζ1 ) Quasi-shear(ζ2 )


±iζ1 x
u1 = ±e(1, x)A1± F e u1 = e(2, x)A2± F e±iζ2 x
u3 = e(1, z)A1± F e±iζ1 x u3 = ±e(2, z)A2± F e±iζ2 x (7.98)

where F = ei(ξx−ωt) , e(a, b) stands for the component of the eigenvector from root a, in direction
b, and the ± signs indicate the differences between the downward (on top) and the upward (on
bottom) travelling waves.

For the isotropic case, this simplifies to

Quasi-longitudinal (ζ1 ) Quasi-shear (ζ2 )


u1 = ±ζ1 A1± F e±iζ1 x u1 = −ξA2± F e±iζ2 x (7.99)
±iζ1 x ±iζ2 x
u3 = ξA1± F e u3 = ±ζ2 A2± F e

which matches the previous derivation for isotropic materials shown in equation 7.56.

For a Cartesian coordinate system we know that the strain vector S is given by [4, p. 29],

 ∂
  ∂ux 
∂x 0 0 ∂x
 0 ∂ ∂uy
∂y 0    ∂y

ux
  
 0 ∂  ∂uz
0 ∂z

 uy  =  ∂z

S= (7.100)
  
∂ ∂  ∂uz ∂uy
 0 +
 
 ∂ ∂z ∂y  uz
 ∂y ∂z 
∂   ∂ux ∂uz

 ∂z 0 ∂x
  ∂z + ∂x

∂ ∂ ∂ux ∂uy
∂y ∂x 0 +
∂y ∂x

and that the stress vector, T, is given by,

Disperse version 2.0.20a


7.4 Cartesian Wave Propagation in Orthotropic Media 116

   
σxx C11 C12 C13 0 0 0

 σyy  
  C21 C22 C23 0 0 0 

 σzz   C31 C32 C33 0 0 0 
T= =  [S] (7.101)

 σyz  
  0 0 0 C44 0 0 

 σzx   0 0 0 0 C55 0 
σxy 0 0 0 0 0 C66


Therefore, for an orthotropic material aligned along an axis, such that ∂y = 0 and uy = 0, the
normal and tangential stresses are

∂ux ∂uz
σxx = C11 + C13 (7.102)
∂x ∂z
and  
∂ux ∂uz
σxz = C55 + (7.103)
∂z ∂x

Substituting into these equations leads to the following expressions for the stresses,
Quasi-longitudinal (ζ1 )
σ11 = i [C11 ξe(1, z) + C12 ζ1 e(1, x)] A1± F e±iζ1 x
σ13 = ±iC66 [ζ1 e(1, z) + ξe(1, x)] A1± F e±iζ1 x (7.104)
and
Quasi-shear (ζ2 )
σ11 = ±i [C11 ξe(2, z) + C12 ζ2 e(2, x)] A2± F e±iζ2 x
σ13 = iC66 [ζ2 e(2, z) + ξe(2, x)] A2± F e±iζ2 x (7.105)

7.4.3 Shear Horizontal Modes

For the ’SH’ type modes, the derivation is much simpler because the wave corresponding to ζ3
root is decoupled from the other bulk waves. As shown by Christoffel’s equation (7.93), all of the
displacement is in the ẑ direction. We choose to scale the eigenvector by (ξ − ζ3 ) in order to match
the isotropic derivation.

Substituting this value into the expression for the third interfacial stress yields,
 
∂uy
σ23 = C44
∂x
= ∓iC44 ζ3 (ξ ± ζ3 )A3± F e±iζ3 x (7.106)
where the ± terms represent downward and upward travelling waves and have been set by switching
ζ3 to −ζ3 .

7.4.4 The Layer Matrix

Combining equations 7.99, 7.104, and 7.105 to form a layer matrix, as was performed for the
isotropic Cartesian case yeilds the following layer matrix, D, whose members all zero, except for
the following,

Disperse version 2.0.20a


7.5 Global Matrix Method 117

u1
D(1, 1) = e(1, x)eiζ1 x
D(1, 2) = −e(1, x)eiζ1 (d−x)
D(1, 3) = e(2, x)eiζ2 x
D(1, 4) = e(2, x)eiζ2 (d−x)

u2
D(2, 5) = −(ξ + ζ3 )eiζ3 x
D(2, 6) = −(ξ − ζ3 )eiζ3 (d−x)

u3
D(3, 1) = e(1, z)eiζ1 x
D(3, 2) = e(1, z)eiζ1 (d−x)
D(3, 3) = e(2, z)eiζ2 x
D(3, 4) = −e(2, z)eiζ2 (d−x)

σ11
D(4, 1) = i(C13 e(1, z)ξ + C11 e(1, x)ζ1 )eiζ1 x
D(4, 2) = i(C13 e(1, z)ξ + C11 e(1, x)ζ1 )eiζ1 (d−x)
D(4, 3) = i(C13 e(2, z)ξ + C11 e(2, x)ζ2 )eiζ2 x
D(4, 4) = −i(C12 e(2, z)ξ + C11 e(2, x)ζ2 )eiζ2 (d−x)

σ13
D(5, 1) = iC55 (ζ1 e(1, z) + ξe(1, x))eiζ1 x
D(5, 2) = −iC55 (ζ1 e(1, z) + ξe(1, x))eiζ1 (d−x)
D(5, 3) = iC55 (ζ2 e(2, z) + ξe(2, x))eiζ2 x
D(5, 4) = iC55 (ζ2 e(2, z) + ξe(2, x))eiζ2 (d−x)

σ23
D(6, 5) = −iC44 (ξ + ζ3 )eiζ3 x
D(6, 6) = +iC44 (ξ − ζ3 )eiζ3 (d−x) (7.107)
where e((, a), b) are the eigenvectors as described above, ζ1 , ζ2 , and ζ3 are the through thickness
wave number components, and d is the thickness of the layer.

Before combining this material layer matrix with the isotropic one, the displacements must be
scaled by the inverse of the frequency and the stresses scaled by the inverse of the frequency
squared.

7.5 Global Matrix Method

Once we know how to describe each material layer and the boundary conditions for each interface,
we are ready to combine the layers to describe the entire system. We have settled on the global

Disperse version 2.0.20a


7.5 Global Matrix Method 118

matrix method, which was first proposed by Knopoff in 1964[39] and subsequently developed by
other researchers such as Randall[80] and Schmidt[85] for use in modelling multi-layered geolog-
ical systems. Compared to the Thomson-Haskell transfer matrix technique[97, 33], which is also
commonly used, this method has the advantage that it remains stable at high frequency-thickness
products. In addition, the global matrix method also allows the same intuitive base matrix to be
used for real or complex wavenumber, vacuum, liquid or solid half-spaces, and modal or response
solutions. A postulated disadvantage is that the global matrix may be large and the solution
therefore may be relatively slow when the systems involve many layers, although it has yet to be
demonstrated whether other techniques for stable solutions are any faster. In any case the speed
of modern computers reduces the effect of any such limitation. A comparison and description of
various matrix techniques can be found in [51].

In the global matrix method, a single matrix represents the complete system. The global matrix
in its general form consists of 6(n − 1) equations, where n is the number of layers (including each
of the semi-infinite half-spaces as a layer). The equations are based, in sets of six, on satisfying
the boundary conditions at each interface. Reduced forms may use a reduced number of sets of
equations: for example the Lamb wave solutions require sets of only four equations, so there are
4(n − 1) equations in total; the Love wave solutions require sets of only two equations, therefore
2(n − 1) equations in total. Returning to the general case, the columns of the global matrix
correspond to the amplitudes of the partial waves in each layer, six for each of the interior layers
and three for each of the exterior (semi-infinite) layers. With this assembly of the material layer
matrices, the continuity of the stresses and displacements is enforced (along the rows) and the
partial waves at the top and the bottom of the layer are consistent (provided by the row continuity).
Figure 7.3 demonstrates how the global matrix is assembled for a five layer system, similar to the
one displayed in Figures 7.1 and 7.2. Multiplying the global matrix by the partial wave amplitudes
simultaneously satisfies all of the boundary condition equations. This equation will always be zero
provided that no energy is being added to the system. It provides the characteristic equation for
our system and may be written,
[G] {A} = 0 (7.108)
where [G] is the global matrix, and {A} is a vector of the partial wave amplitudes. This equation
is satisfied if the determinant of the global matrix is zero. The frequency, wavenumber, and
attenuation values determine whether this condition is met, and need to be found via an iterative
procedure that is explained in the next section. It is worth noting here that an accurate algorithm
for calculating the determinant is essential otherwise a search algorithm will not easily converge
on a zero. This is especially tricky since the searches are necessarily close to the zeros of the
determinant. We use a Gaussian elimination scheme with partial pivoting which has proved to
work well. It is described in reference [53]

There are some conditions for which the global matrix will become singular other than valid wave
propagation solutions. The most common of these conditions occurs when the phase velocity
(ω/ξ) is equal to one of the bulk velocities. In this case, the effect of the upward and downward
travelling waves is indistinguishable and two of the columns become linearly dependent, causing
the global matrix to become singular.

If one or both of the half-spaces is either an inviscid liquid or a vacuum, the matrix can be easily
modified to remove the unnecessary boundary conditions and partial waves, as demonstrated for
the top half space (or centre core for a cylinder) in Figures 7.3(b) and 7.3(c). When the top
half space (or centre core) is vacuum, the first three columns and rows are removed by making
all of the elements zero except for the diagonal elements (1,1), (2,2), and (3,3), which are set to
one. This operation removes the constraints on the displacements at the inside of the system and
eliminates reference to the bulk waves that would exist in the centre core. If the center core is
an inviscid liquid, the constraints on the two tangential displacements are removed. In addition,
the shear bulk waves are removed. This operation leaves elements that relate the amplitude of

Disperse version 2.0.20a


7.6 Spring Interface 119

Figure 7.3: The structure of the global matrix for a (a) solid, (b) liquid, and (c) vacuum half-
space, where blank spaces are zeros, Dij is the layer matrix, and L+-, SV+-, and SH+- are the
partial waves amplitudes in the various layers.

the longitudinal bulk wave to the normal displacement and normal stress at the interface of the
centre core and the first layer. If the lower (or outer) half-space is a vacuum or a liquid, similar
modifications can be made to the lower right part of the matrix.

7.6 Spring Interface

It is sometimes useful to model layered structures in which the contact conditions between the
layers are not perfect. For example it may be required to model a free sliding condition in which
two layers may slide with respect to each other but may not separate or overlap - such a condition
would be described by continuity of the displacement in the normal direction but not in the
tangential direction. Or it may be required to represent an imperfect interface according to some
interface stiffness; this may be useful for example to represent the partial contact conditions of
two plates in contact. Some references where spring interfaces have been used are [2, 76, 77, 70].

Disperse has been developed to accommodate such possibilities in a general way. It allows the
user to control each of the degrees of freedom at the interface independently. For each degree of
freedom the user may choose the contact condition to be one of three possibilities: (1) Rigid; this
is the default “welded” condition which is assumed in the absence of selecting the interface option,
corresponding to the continuity of the displacement in the chosen degree of freedom; (2) Free; in
this condition the surfaces are deemed to be totally disconnected in the chosen degree of freedom
and the corresponding stress is zero; (3) Spring; with this option the user must supply a spring
stiffness (per unit area of the interface) in the chosen degree of freedom. Clearly the spring must
always lie somewhere between the rigid and the free conditions: a spring with infinite stiffness
corresponds to option (1) whereas a spring with zero stiffness corresponds to option (2). Spring
interfaces may be defined separately for any of the interfaces between the layers or indeed for the

Disperse version 2.0.20a


7.6 Spring Interface 120

upper and/or lower half spaces.

In introducing the spring interface condition, the philosophy which was adopted was to incorporate
it with minimum disruption to the program structure, particularly the assembly of the global
matrix. The spring interface has therefore been introduced as a “material layer”. The layer has
material constants which govern the stiffness across it and has zero thickness. The layer can be
built in to the appropriate location in the global matrix without making any changes to the other
layers of the system.

General scheme for an Arbitrary Location between Two Layers

Consider a spring interface layer at an arbitrary location in a multilayered structure. Using the
example of the five layer structure, the global matrix is as follows:

A−
  

  1
[D1b ] [−D2t ]  

{A2 }

 

[D2b ] [−D3t ]
   
  {A3 } = {0} (7.109)
 [D3b ] [−D4t ] 
{A+
4}
 
+
 
[D4b ] [−D5t ] 
 

A5
 

The “[D]” entries in the matrix are the layer submatrices and the {A} entries are the partial wave
amplitudes, as discussed elsewhere. Their positions and constituent coefficients can be seen by
direct comparison with Figure 7.3, or by reference to the equations on page 538 of [51]. Here, and
throughout, blank coefficients in the matrix denote zeroes.

Take, for example, the third layer to be the spring layer. Then the matrices [D3t ] and [D3b ] are
the matrices which must contain the coefficients which give the spring behaviour. The first, [D3t ],
relates to the boundary conditions between the “top” of the spring layer and the real material
layer above. The second, [D3b ], relates to the boundary conditions between the “bottom” of the
spring layer and the real material layer below. Here for convenience of reference we use the names
“top” (t) and “bottom” (b) to describe the boundary functions of the spring layer, although it
actually has no thickness. As usual the rows of the matrices relate to the degrees of freedom, and
these are in the following order: {u1 , u2 , u3 , σ11 , σ12 , σ13 }, where direction 1 is the normal to the
interface and positive downwards, 3 is along the plate or pipe in the direction of propagation, and
2 is orthogonal to 1 and 3, as shown in Figures 7.1 and 7.2. Thus u2 and σ12 are the degrees
of freedom which are not needed for the common Lamb type modes in two dimensional plate
calculations. The “wave amplitude” vector {A3} is reserved for the amplitudes of partial waves
in the spring layer, their meanings will be introduced later.

Consider a spring stiffness in each of the relevant orientations: (1) k11 is a spring defining the
stiffness between the layers in the direction normal to the interface; (2) k12 is a spring defining the
shear stiffness between the layers in the 12 plane; (3) k13 is a spring defining the shear stiffness
between the layers in the 13 plane. The springs are uncoupled and the value of each k may
be defined independently. The spring constants relate the displacements and stresses across the
spring layer as follows:

−σ11
u1(t) − u1(b) =
k11
−σ12
u2(t) − u2(b) =
k12

Disperse version 2.0.20a


7.6 Spring Interface 121

−σ13
u3(t) − u3(b) = (7.110)
k13

Each displacement in these equations requires a subscript (t) or (b) to show whether it is at
the top or bottom of the spring layer; this is necessary because the springs allow differentials of
displacements between the two materials in contact. On the other hand the stresses do not require
any such subscripts because they are constant (and in equilibrium) across the layer; for example
σ11 at the top of the layer must be the same as σ11 at the bottom of the layer, etc.. The signs
of these equations need careful thought, particularly regarding the convention for shear stresses;
it turns out for our coordinate system that each equation has a negative sign before the stress,
as shown (all three equations would be positive if the coordinate direction was taken as positive
in the upwards direction). The value of each k is the stiffness (eg N/mm) per unit area of the
interface (/mm), thus units are N/mm3 .

Consider now the coefficients for the two matrices [D3t ] and [D3b ]. Starting with the bottom of
the spring layer, let the field matrix for this interface be a unit matrix:

     

 u1 
 1 
 AL+ 

u2 1 A
   
L−

 
   
 

     
u3 1 A
   
SV +
 
=  (7.111)

 σ 11 


 1 
 
 A SV − 

σ 1 ASH+
   
12

 
   
 

   
σ13 1 layer3,(b) ASH−
   
layer3,(b) layer3

Thus the matrix [D3b ] is defined:

 
1

 1 

 1 
[D3b ] =   (7.112)

 1 

 1 
1

The vector of “wave amplitudes” for the layer thus becomes the vector of the displacements at
the bottom of the layer and the stresses in (throughout) the layer. Now the field matrix equation
for the top layer becomes:

   

 u1 
 
 u1 

u2 u2

 
 
 


 
 
 

u3 u3
   
= [D3t ] {A3 } = [D3t ] (7.113)
 σ11
 
  σ11
 

 σ12  σ12

 
 
 

 
  

σ13 σ13
   
layer3,(t) layer3,(b)

The matrix [D3t ] now represents the transfer of the displacements across the spring layer, and so
is easily populated using equation 7.110:

Disperse version 2.0.20a


7.6 Spring Interface 122

 
1 −1/k11

 1 −1/k12 

 1 −1/k13 
[D3t ] =   (7.114)

 1 

 1 
1

Modification for a Zero Stiffness Spring

A zero stiffness spring corresponds to the disconnection of a degree of freedom. It is defined by


modifying appropriate coefficients of the [D3t ] and [D3b ] matrices of equations 7.114 and 7.112
respectively. Each degree of freedom can be considered separately so a mixed condition boundary
such as free sliding but stiff normal contact can be modelled. The modification for a chosen degree
of freedom is to set to zero the diagonal coefficient for the relevant stress in both matrices. The
relevant stresses are the obvious, i.e. σ11 for the k11 stiffness, σ12 for the k12 stiffness, σ13 for the
k13 stiffness.

With these two modifications, the meaning of the “wave amplitudes” relating to this degree of
freedom are also affected. There are two “wave amplitudes” for each direction: the displacement
and the stress. Normally these are equal to the displacement and stress values at the bottom of
the spring layer, as explained earlier. Now the displacement wave amplitude remains that at the
bottom of the layer, but the stress wave amplitude becomes the displacement at the top of the
layer. Thus the two wave amplitudes are the displacements at the bottom and top of the spring
layer, which of course are now disconnected and can be different. At the same time the two stresses
are enforced to be zero, as must be required by such a boundary condition.

Modification for an Infinite Stiffness Spring

An infinite stiffness in a degree of freedom corresponds to a rigid connection. This is identical


to the coupling condition normally assumed between layers when interface spring layers are not
used. Thus it only makes sense to define a rigid condition in one or two of the degrees of freedom;
if all three degrees of freedom were defined to be rigid then there would be no purpose in using a
spring interface layer.

To model the rigid condition, it is only necessary to modify the [D3t ] matrix. The off-diagonal
coefficient (containing the “k” coefficient) in the relevant displacement row of the matrix is removed
(replaced with zero).

Scheme for a Spring Interface at the Top or Bottom Boundary of the Model

If the spring layer is to be the top half space of the model, then the matrix containing the spring
properties is [D−1b ]. This matrix has 6 rows but only 3 columns.

The spring constants now determine the absolute displacements and stresses at the top of the
model, that is to say across the spring layer and to ground. The equations for the bottom of the
spring layer where it connects to the first real material layer are as follows:

Disperse version 2.0.20a


7.7 Analytical Solutions for Leaky Lamb Case 123

σ11
u1(b) =
k11
σ12
u2(b) =
k12
σ13
u3(b) = (7.115)
k13

The matrix equation for [D−


1b ] is :

 
1/k11

 1/k12 

 −  1/k13 
D1b =   (7.116)

 1 

 1 
1

Similarly if the spring interface is the bottom half space of the model, then the bottom right
submatrix of the global matrix must be modified. In the example here, this is the matrix [D5t+ ].
The equations for the top of the spring layer where it connects to the lowest real material layer
are as follows:

−σ11
u1(b) =
k11
−σ12
u2(b) =
k12
−σ13
u3(b) = (7.117)
k13

The matrix equation for [D+


5t ] is thus:

 
−1/k11

 −1/k12 

 +  −1/k13 
D5t =   (7.118)

 1 

 1 
1

It follows from the unit diagonal of the lower half of these matrices that the wave amplitudes are
the stresses, σ11 , σ12 and σ13 . Modifications of these top and bottom matrices for zero stiffness or
infinite stiffness are exactly as described earlier for the general spring interface.

7.7 Analytical Solutions for Leaky Lamb Case

The global matrix method allows a wide number of different geometries to be analysed using the
same configuration. However, generalised matrix methods are not particularly fast. For simple

Disperse version 2.0.20a


7.7 Analytical Solutions for Leaky Lamb Case 124

systems that are commonly analysed, it is beneficial to evaluate an analytical form of characteristic
equation. Disperse uses this technique for single-layered, elastic, isotropic, plates in vacuum or
immersed in inviscid liquids. These cases are true Lamb waves and leaky Lamb waves. A summary
of the derivation follows.

The same derivation is used for both the Lamb cases and the leaky Lamb cases, except that the
loss due to the waves leaking into the surrounding fluid is eliminated when the plate is in vacuum.
The derivation follows the same general outline as the derivation for the layer matrix for a flat
isotropic plate. However, the boundary conditions and equations for the surrounding media are
inserted directly into the characteristic equation.

At the interface between the solid layer and the surrounding liquid, certain boundary conditions
will be enforced. The normal displacement, u1 , will need to be continuous to avoid the formation
of voids. In addition, the normal stress, σ11 should be continuous to balance the pressure. The
shear stress, σ13, must be zero at the interface, because the liquid cannot couple the shear forces.
The rest of the displacements and stresses are not constrained.

The derivation for a Leaky Lamb wave begins in the same fashion as the derivation for an isotropic
layer. We assume that the propagation of the guided wave is in the x3 direction, the x1 direction
is normal to the layer, and that the system is in a state of plane strain in the x2 plane. The
shear horizontal mode, which is decoupled, is ignored and only the motion in the X-Z (1-3) plane
is considered. The displacement can be expressed as two potentials, one scalar and one vector,
u = ∇φ + ∇ × H, where the scalar vector, φ, is of the form,

φ = [AL1 cos(k1 x1 ) + AL2 sin(k1 x1 )] ei(k3 x3 −ωt) (7.119)

and the vector potential, H, is of the form,

H = [AS1 cos(k1 x1 ) + AS2 sin(k1 x1 )] ei(k3 x3 −ωt) . (7.120)

The through thickness field has been described as the sum of sines and cosines so that the solution
can later be separated into symmetric and anti-symmetric portions. If the through thickness field
is described as exponentials, as done for the isotropic layer case, the symmetric and anti-symmetric
portions will be coupled.

Since the amplitude of the wavenumber is known to be ω/v1 for the scalar potential and ω/v2 for
the vector potential, the two components of the wavenumber can be specified in relation to each
other. Therefore, if the wavenumber component in the direction of propagation, k3 , is represented
as ξ, the square of the wavenumber component normal to the direction of propagation can be
specified as either,

ω2
ζ1 2 = − ξ2
v12
ω2
ζ2 2 = − ξ2 (7.121)
v22

where ζ1 is the normal wavenumber component for the scalar potential, φ, and ζ2 is the normal
wavenumber component for the vector potential, H.

Specifying the displacement in terms of the two different potentials leads to the expressions for the
displacement due to a compressional wave (uL ) and those due to a rotational wave (uS ). Thus,

Disperse version 2.0.20a


7.7 Analytical Solutions for Leaky Lamb Case 125

the displacement in the plate can be written,


 ∂φ 
∂x1
∂φ
uL = ∇φ = 
 
∂x2 
∂φ
∂x3
   
−ζ1 sin(ζ1 x2 ) ζ1 cos(ζ1 x2 )
= AL1  0  ei(ξx3 −ωt) + AL2  0  ei(ξx3 −ωt) (7.122)
iξ cos(ζ1 x2 ) iξ sin(ζ1 x2 )

 ∂  
∂x1 0

uS = ∇×H=  ×  H3 
∂x2
∂ 0
∂x3
   
−iξ cos(ζ2 x2 ) −iξ sin(ζ2 x2 )
= AS1  0  ei(ξx3 −ωt) + AS2  0  ei(ξx3 −ωt) (7.123)
−ζ2 sin(ζ2 x2 ) ζ2 cos(ζ2 x2 )

in which the vector potential H points in the x2 direction so that all particle motion is in the x1 x3
plane.

As specified in equation 7.43, the normal compressional stress and the tangential shear stress can
be expressed as,

σ11 = λ∆ + 2µ11
σ13 = µ13 (7.124)

where
∂u1 ∂u2 ∂u3
11 = 22 = 33 =
∂x1 ∂x2 ∂x3
∂u1 ∂u3
13 = + ∆ = 11 + 22 + 33 (7.125)
∂x3 ∂x1
and where λ and µ are Lamé’s elastic stiffness constants and ∆ is the change in volume (dilatation)
of the element. Inserting the expressions for the displacements into these equations yields the
following expressions for the normal and shear stresses.

Compressional Rotational
2 2
σ11+ = −µ(ζ2 − ξ ) cos(ζ1 x1 ) σ11+ = 2µξζ2 sin(ζ2 x1 )
2 2
σ11− = −µ(ζ2 − ξ ) sin(ζ1 x1 ) σ11− = −2µξζ2 cos(ζ2 x1 )
σ13+ = −2iµζ1 ξ sin(ζ2 x1 ) σ13+ = −µ(ζ2 2 − ξ 2 ) cos(ζ2 x1 )
σ13− = 2iµζ1 ξ cos(ζ2 x1 ) σ13− = −µ(ζ2 2 − ξ 2 ) sin(ζ2 x1 )(7.126)

where the term ei(ξx3 −ωt) has been omitted from all of the expressions. In the calculation of the
expression for the normal stress, σ11 , the substitution, λ(ζ1 2 + ξ 2 ) + 2µζ1 2 = µ(ζ2 2 − ξ 2 ), was used.

The surrounding inviscid liquid can be modelled using a single scalar potential since it does not
support any shear stresses. The two potentials corresponding to the upper and lower half-spaces

Disperse version 2.0.20a


7.7 Analytical Solutions for Leaky Lamb Case 126

can be expressed as,


h i
ΦL1 = Aliq1 e−i(γx1 ) ei(ξz−ωt) for x1 ≤ −d/2
h i
ΦL2 = Aliq2 ei(γx1 ) ei(ξz−ωt) for x1 ≥ d/2 (7.127)

For the liquid, the displacement and stresses are therefore expressed as the
∂ΦL
u1 =
∂x1
= −iγAliq1 e−i(γx1 ) for x1 ≤ −d/2
i(γx1 )
or iγAliq2 e for x1 ≥ d/2 (7.128)
2
σ11 = ω ρL ΦL (7.129)
p
where γ = ω 2 /c2L − ξ 2 and cL is the bulk wave velocity in the surrounding liquid.

Evaluating the expressions for the normal displacement, the normal compressional stress and the
tangential shear stress at the top surface (−d/2) and the bottom surface (d/2) and setting the
values for the solid plate equal to the values for the liquid loading, leads to a matrix A, which
solves the equation,  
u


 1 



 x1 =d/2 


   
u

A
 
 L+  
 1 

   x 1 =−d/2 
A
    
L−
  
  σ11

 
 
 

AS+
 
x1 =d/2
[A] = (7.130)

 AS−    σ11 
  
Aliq1  x1 =−d/2 

  
 


  
   σ 
Aliq2
 

 13 



 x1 =d/2 


 
 σ13
 

x1 =−d/2

where A is,
−ζ1 sin(ζ1 d/2) ζ1 cos(ζ1 d/2) −iξcos(ζ2 d/2) −iξsin(ζ2 d/2) 0 −iγeiγd/2
−A11 A12 A13 −A14 iγeiγd/2 0
−µ(ζ2 2 − ξ2 )cos(ζ1 d/2) −µ(ζ2 2 − ξ2 )sin(ζ1 d/2) 2iµξζ2 sin(ζ2 d/2) −2iµξζ2 cos(ζ2 d/2) 0 ω 2 ρL eiγd/2
A31 −A32 −A33 A34 ω 2 ρL eiγd/2 0
−2iµξζ1 sin(ζ1 d/2) 2iµξζ1 cos(ζ1 d/2) −µ(ζ2 2 − ξ2 )cos(ζ2 d/2) −µ(ζ2 2 − ξ2 )sin(ζ2 d/2) 0 0
−A51 A52 A53 −A54 0 0
(7.131)

Equation 7.130 is satisfied when the determinant of matrix A is zero. Before solving for the
determinant the matrix A can be separated into two submatrices in order to accelerate the calcu-
lation process. Adding and subtracting various rows and columns and dividing the partial wave
amplitudes in the liquid half-spaces by the common factor eiγd/2 yields the following submatrices

ζ1 sin(ζ1 d/2) iξsin(ζ2 d/2) iγ
µ(ζ2 2 − ξ 2 )cos(ζ1 d/2)
2

2iµξζ 2 cos(ζ 2 d/2) −ω ρ L =0
(7.132)
2iµξζ1 sin(ζ1 d/2) µ(ζ2 2 − ξ 2 )sin(ζ2 d/2) 0

and
ζ1 cos(ζ1 d/2) iξcos(ζ2 d/2) −iγ
µ(ζ2 2 − ξ 2 )sin(ζ1 d/2)

2iµξζ 2 sin(ζ2 d/2) −ω 2 ρL =0
(7.133)
2iµξζ1 cos(ζ1 d/2) µ(ζ2 2 − ξ 2 )cos(ζ2 d/2) 0

which account for the symmetric and anti-symmetric solutions to the characteristic equations.

Disperse version 2.0.20a


7.8 Finding a Root 127

These 3 by 3 matrices can be solved and we are left with the standard lamb wave equations with
an extra term appended. This solution for symmetric modes is,
2
ζ2 2 − ξ 2 cos(ζ1 d/2)sin(ζ2 d/2) + 4ξ 2 ζ1 ζ2 sin(ζ1 d/2)cos(ζ2 d/2)
ρL ζ1 kS4
− i sin(ζ1 d/2)sin(ζ2 d/2) (7.134)
ρS ktliq
and for the anti-symmetric modes is,
2
ζ2 2 − ξ 2 sin(ζ1 d/2)cos(ζ2 d/2) + 4ξ 2 ζ1 ζ2 cos(ζ1 d/2)sin(ζ2 d/2)
ρL ζ1 ω 4
+ i cos(ζ1 d/2)cos(ζ2 d/2) (7.135)
ρS γc42

These expressions can be used in place of the global matrix formulation to calculate the dispersion
curves for a single layered plate surrounded by a liquid or vacuum. Since the symmetric and anti-
symmetric modes are decoupled, the solution is more stable (and quicker) than the general solution
technique. Although this technique provides an analytical form for the characteristic equation,
valid frequency - wave number - attenuation combinations must still be found iteratively using the
same techniques as are used when the characteristic equation is calculated by the global matrix.
The root finding technique that is used is outlined in the following section.

7.8 Finding a Root

In order to find a point on a dispersion curve, we need to find a root of the characteristic equation
(equation 7.108), which in most cases corresponds to a point where the determinant of the global
matrix, a complex value, is zero. The coefficients of the global matrix are dependent on the
geometry of the system, the material properties, the frequency, the real wave number, and the
attenuation. For a given problem, the latter three properties, the frequency, real wave number, and
attenuation, are varied to find valid roots. The frequency describes the time varying harmonic
characteristics of the mode in question, the real wave number the spatially varying harmonic
characteristics, and the attenuation the decay of the wave in space or time. The first step to find a
root involves performing a coarse sweep, for which two of the variables are held constant while the
third is varied. The routine then searches for minima in the absolute value of the determinant to
use as starting points for a more detailed search. The absolute value of the determinant is used for
searches because at this point the routine is only looking for local minima. Zeros of the function
only occur when all three search variables (frequency, real wave number, and attenuation) take
on appropriate values. If only two of the variables are correct, the zero will appear as a local
minimum of the determinant. If a minimum of the characteristic equation is found, the routine
begins a fine search, which then tries to converge onto a valid root.

If all of the materials are elastic and the half-spaces are both vacuum, then there is no way
for energy to leave the system and so there will be no attenuation. In this case the fine search
routine can be relatively simple. Since the attenuation of propagating modes will always be zero,
a single one dimensional minimum search routine suffices to find the root. However, the process is
more complicated when attenuation exists and a two dimensional search routine is required. The
attenuation is an important property of the system, describing the rate at which a guided mode
decays due to damping or leakage as it travels. It is unknown and has to be found as part of the
solution.

Figure 7.4 illustrates the process of finding a valid root of the characteristic equations when
attenuation is present in the system. The surface corresponds to the log of the absolute value

Disperse version 2.0.20a


7.9 Tracing a Dispersion Curve 128

Figure 7.4: The process of finding a root involves a coarse sweep (thick line) to find an initial
minimum and a fine search (dashed lines) to narrow down on a root (solid circle). The example
above shows the absolute value of the determinant of the global matrix for a 1 mm steel plate
immersed in water at a wave number of 4 rad/mm.

of the determinant of the global matrix when the real wave number is held at 4 rad/mm for a
1 mm steel plate immersed in water. To find a root, a coarse frequency sweep is made at zero
attenuation, as shown by the thick line at the front of the surface. The two minima are identified
as possible roots and are used as starting points for a fine search. The left minimum is less well
defined than the right minimum because the attenuation value of the sweep is very different than
the actual attenuation of the root.

Next, starting at a minimum located by the coarse sweep, the fine search uses a two dimensional
iterative bisection routine (dashed line with open circles) to converge onto the root (solid circle).
The routine starts by taking small steps in frequency (dashed line) until a tighter minimum (open
circle) is found. The segments that bracket the minimum are each split in two and the function
is calculated at these points, allowing a new minimum to be chosen from the middle three points.
The bracketing continues until the desired tolerance is achieved, at which point the routine uses
the same logic to converge in attenuation. The routine repeats the search for minima in frequency
and then attenuation until it determines that it has found a valid root by examining nearby phase
changes. A more complete description of this robust search routine can be found in [53].

7.9 Tracing a Dispersion Curve

Although it is possible to find a lot of roots of the dispersion equation and connect the dots, there
is a much more efficient and robust method for tracing dispersion curves. The tracing routine
employed for this model uses extrapolation to provide predictions of where subsequent points on
a dispersion curves will fall. Therefore, once a single point on a mode has been found, the rest of
the mode may be traced much more quickly than if each point needed to be converged upon from

Disperse version 2.0.20a


7.9 Tracing a Dispersion Curve 129

Figure 7.5: Using an extrapolating routine to trace modes dramatically improves the program’s
speed and reliability.

an inaccurate starting value.

As shown in Figure 7.5, the tracing routine starts from a root that was found from a coarse sweep,
labelled the initial root. Next the routine takes a very small step in one of the three independent
variables, in this case wave-number, and converges on a second root. These first two points are
used to linearly extrapolate to and converge upon a third root, which is then used to linearly
extrapolate to a fourth point. After the fourth root has been found, the routine predicts the next
root by quadratic extrapolation from the three previous roots. After the seventh root, every other
root is used for the quadratic extrapolation to reduce the effect of a single solution ’jumping’
onto another curve as two modes pass near to each other[51]. For the case shown in Figure
7.5, the wavenumber is stepped and the frequency and attenuation are extrapolated. It is also
possible to choose a different variable for the steps. For example steps may be taken in frequency
together with iterations in attenuation and wavenumber. Indeed Disperse uses, as its default
scheme, steps which are a combination of frequency and wavenumber, named the “frequency-
wavenumber combo”. With this scheme, each step is taken in a combination of frequency and real
wavenumber such that the step is tangential to the frequency-real-wavenumber dispersion curve;
the iterations of the other real variable are then made using another combination of frequency
and wavenumber such that the iteration direction is normal to the dispersion curve. An example
which well illustrates the utility of such a scheme is the calculation of the dispersion curves for a
plate with high material damping, for which the frequency-wavenumber curves form “e” shaped
loops, see example of this in Section 8.4.

Once a mode has been traced completely, another starting point is used to trace out the next
mode, until the entire set of dispersion curves has been identified.

Disperse version 2.0.20a


7.10 Mode Shapes 130

7.10 Mode Shapes

One method of examining the solutions of the dispersion equations is to look at the mode shape
at each of the points on the mode. The mode shape displays how the displacements, stresses, or
energy varies through the thickness of the system. The variation in the direction of propagation
is known to be sinusoidal and is therefore not of interest.

To calculate the mode shapes, we arbitrarily assume the amplitude of one of the bulk waves.
This assumption allows the amplitudes of the other bulk waves to be found by solving a modified
version of equation 7.108, which is the characteristic equation for this problem. Once all of the
bulk wave amplitudes are known, the stresses and displacements for an arbitrary point in a layer
can be calculated by multiplying the wave amplitude vector for that layer by the layer matrix
given in equation 7.86.

In addition to the stresses and displacements, several other quantities may be calculated and their
variation across the thickness of the structure plotted. Details of the derivations of these quantities
can be found in for example [4].

Strain energy density is given by the following summations of the stress-strain products. Note
that these expressions assume complex values of the quantities ([4]); if real values are used
then these expressions should be multiplied by two.
(a)for flat plate systems:
      
1 ∂ux ∂uy ∂uz
SED = σxx + σyy + σzz
4 ∂x ∂y ∂z
      
1 ∂ux ∂uy ∂uy ∂uz ∂ux ∂uz
+ σxy + + σyz + + σxz +
8 ∂y ∂x ∂z ∂y ∂z ∂x
(7.136)

(b)for cylindrical systems:


( )
∂   
1 ∂u r 1 ∂uθ ur ∂uz
SED = σrr ∂ + σθθ + + σzz
4 ∂r
r ∂θ r ∂z
( !)
    ∂ ∂
1 ∂ur ∂uz ∂  uθ  1 ∂ur ∂u θ 1 ∂u z
+ σrz + + σrθ r + + σθz ∂
+ ∂
8 ∂z ∂r ∂r r r ∂θ ∂z
r ∂θ
(7.137)

Kinetic energy density is given by a quarter of the sum of the products of mass with velocity-
squared (again, this is for complex quantities; multiply this expression by 2 if using real
values), which is:

( 2  2  2 )
ρ ∂u1 ∂u2 ∂u3
KED = + + (7.138)
4 ∂t ∂t ∂t

where 1, 2, 3 correspond to X,Y,Z or R, θ, Z for flat plate or cylindrical systems respectively.

Total energy density, NRG is the sum of the strain energy density and the kinetic energy
density.

Disperse version 2.0.20a


7.10 Mode Shapes 131

Power flow density is obtained from the product of the velocity vector and the stress tensor, and
is itself a vector (the Poynting vector),

∂u1 ∗ ∂u2 ∗ ∂u3 ∗


    

 σ11 ∂t + σ12 ∂t + σ13 ∂t 


 

 
∂u2 ∗ ∂u2 ∗ ∂u3 ∗ 
    
1  σ12 + σ22 + σ23
PWR = − ∂t ∂t ∂t (7.139)
2 
∂u3 ∗ ∂u2 ∗ ∂u3 ∗ 

    
 σ13 + σ23 + σ33

 

 ∂t ∂t ∂t 

in which * denotes the complex conjugate. As with the previous items, multiply this by 2 if
using real values.
The Lines plotting options in Disperse allow the user to plot the variation of any of the
three components (R, q, Z) of this vector across the thickness of the structure. The Power
option allows the vector itself to be plotted, thereby displaying the direction of the power
flow. The power flow density information is useful to show how the energy is flowing in the
structure - for example the depth at which the greatest transmission of energy takes place,
or the manner in which energy leaves the structure by leakage.

One final matter relating to the mode shapes is the power flow of a mode. This is a quantity for
the whole structure rather then at a particular location in the structure and it requires an integral
of the power flow density in the direction of propagation of the wave, over the thickness of the
structure. The power flow of a mode indicates the rate at which energy is transmitted along the
structure, that is to say Joules per second passing a particular point. The power flow of a mode
is thus given by the integral

Z   ∗  ∗  ∗ 
1 ∂u1 ∂u2 ∂u3
Pm = − σ13 + σ23 + σ33 dS (7.140)
2 S ∂t ∂t ∂t

in which the integration is performed over the cross section S of the plate or cylinder. Since the
amplitudes of the mode shapes which Disperse predicts are by definition arbitrary (see explanation
of this in Section 6.1), Disperse has to choose some scale when plotting them. A convenient scale
has been chosen to be the power normalised scale. Thus Disperse plots the mode shapes with the
amplitudes corresponding to unit power flow in the direction of the wave propagation. So, for
example, when the mode shape of a displacement component is plotted, the value of displacement
shown for any point on the curve is that which would be present if the mode had unit power flow.
Disperse performs the integration across all of the finite layers of the structure. The power flow
normalisation is correct for both elastic and lossy materials. However, note that if the structure is
embedded or immersed then the power normalisation is not strictly correct because the components
of the power flow in the half spaces are neglected. When performing the normalisation for plate
structures, Disperse assumes a dimension of 1 unit (m) in the direction normal to the plane (Y).
When performing the normalisation for waves propagating along cylindrical structures, Disperse
works on the basis of unit power flow for the whole circumference. For circumferential waves it is
per unit (m) distance in the direction normal to the plane of calculation (the Z direction).

The Energy velocity can be calculated from the above quantities by integrating over both a
whole cycle and the whole cross section. It is given by

R R
(P W RZ ) dT dS
Ve = RS RT (7.141)
S T
(N RG) dT dS

Disperse version 2.0.20a


7.10 Mode Shapes 132

in which S is the cross section of the plate or cylinder, T is the time period of the wave, P W RZ
is the component of the power flow density (Poynting vector) in the direction along the plate or
cylinder and NRG is the total energy density. The energy velocity is the velocity at which a wave
packet travels. It is equal to the group velocity if the mode has zero attenuation. However if the
mode loses energy through leakage or material damping then the group velocity is not strictly
correct whereas the energy velocity calculation remains strictly true ([7]). For low losses the group
velocity is usually quite accurate, and it is quick to calculate so is to be preferred; if the losses are
significant, then consider calculating the energy velocity instead, see Section 5.3.2.

Further information about how to plot mode shapes is given in Section 6.1; see also the examples
in Chapters 8 and 9.

Disperse version 2.0.20a


Chapter 8

Cartesian Examples

This chapter discusses the tracing and interpretation of the dispersion curves for some sample
Cartesian systems. These examples are included to illustrate how Disperse can be used to model
a wide range of systems and to demonstrate what type of dispersion curves should be expected
for simple systems. New users are advised to try to duplicate some of these results for practice.

In general, the examples begin with simple, common cases and become increasingly more complex.

8.1 Lamb Waves

In this section Lamb waves are defined as guided waves that propagate in a single isotropic elastic
layer that is surrounded by vacuum. This case corresponds to the system that Horace Lamb first
studied in 1917[46].

Since this is such a common case to study, Disperse uses the analytical versions of the characteristic
equation, which are developed in Section 7.7, instead of using the general global matrix method.
The analytical functions calculate the symmetric and the anti-symmetric solutions separately,
which adds extra stability to the solution. In addition, the analytical routines that are specific to
a single geometry are faster than the global matrix formulation of the same problem.

The dispersion curves for Lamb waves are relatively simple to calculate. Because the system is
elastic and does not leaky energy into a surrounding medium, there is no attenuation. Therefore,
only the wavenumber and frequency have to be specified to define a valid root, which makes the
routines much more efficient than if roots have to be found in three dimensions. In addition,
the stability of the analytical solution helps. For these reasons, the automatic tracing (Section
4.2)can usually find all of the dispersion curves without much or any extra input from the user. A
description of how to calculate and interpret a simple set of Lamb wave dispersion curves follows.

8.1.1 Tracing the Curves

Specifying the geometry is the first step that is required to calculate the Lamb wave dispersion
curves. In order to do this,

Disperse version 2.0.20a


8.1 Lamb Waves 134

• Choose File : New from the menu.


• Ensure that “Plate” is selected as the geometry.
• Ensure that “Structure is in vacuum” is selected for the semi-infinite half spaces.
• You will see three entries in the window, defining three layers. The first and the third
say “Free boundary”, the middle says “Undefined”. Click on the middle with the left mouse
button and select a material from the pull-down menu which appears. If you want a material
other than those on this predefined list, then you can define one, see Section 3.3).
• Change the thickness (in mm) by editing the text field to the right of the material name.
• Press the OK button

Once the geometry is defined, a dispersion curve window appears. A description of the system
is given in a tree format on the left side of the window. Click the “+” symbol to expand the
information about each layer. The description includes information about the geometry as well as
the material properties of the individual layers. Please refer to Section 3.2 for more details.

The second step involves actually tracing the dispersion curves. For the Lamb wave case, the
automatic tracing routine will almost always find the dispersion curves quickly without any manual
assistance from the user. In order to run the automatic tracing routine (Section 4.2), select Trace
: Automatic tracing from the menu. The default settings for the automatic tracing should
not need to be changed, although you may want to modify the frequency to which the dispersion
curves should be traced, which is specified in the second box from the top of the window. You may
also increase or decrease the number of points used to describe each curve by changing the step
size, known as the Increment. But in most cases the default value is probably sensible. Disperse
will now search for the fundamental symmetric and anti-symmetric modes and trace these two
modes. It will then perform a sweep at a high phase velocity and trace out each mode that it
finds.

As the modes are traced, Disperse attempts to name each of the modes. The modes are named
“S” for symmetric modes (which are coloured red) and “A” for anti-symmetric modes (which are
coloured blue). An index follows the letter marking to indicate the order of the mode, starting
with the fundamental modes, which are labelled “0”. In the case of Lamb modes, the names which
Disperse ascribes are intended to follow the usual convention, and indeed in most cases Disperse
will get these right. However note that the names which are found automatically are not always
correct and should only be used as a guide. The names of the curves are shown whenever a point
on a dispersion curve is selected - to do this click the right mouse button, then select Pointer,
then click anywhere on a curve. Methods of attaching labels to each of the curves and of changing
the names of the curves are given in Section 6.8.

If the automatic tracing routine misses a mode, the missing modes can be traced by using the
Trace from selected point action (Section 4.3.2).

Pressing the stop button on the tool bar causes Disperse to abort the automatic tracing procedure
which it is currently executing. All information which has been completed at this time is retained.

8.1.2 Interpreting the Results

Once the dispersion curves have been traced, information can easily be extracted. By default,
Disperse displays the dispersion curves projected into frequency - phase velocity space. Sample

Disperse version 2.0.20a


8.1 Lamb Waves 135

curves for a 1 mm steel plate are shown in Figure 8.1. By selecting options from the Display
menu, other common projections such as real wavenumber(5.2.4), group velocity (5.2.2) and angle
of incidence (5.3.2)can be shown. These projections are displayed in Figures 8.1 and 8.2.

Labels (6.8) have been attached to the various curves in Figure 8.1, so that they can be identified.
These mode names were assigned by the automatic tracing routine. However, the user should keep
in mind that the naming routine is not very sophisticated and frequently makes mistakes. For
example, if a mode is divided into two parts, the two parts will be labelled as different modes.
The mode names can be reassigned using different conventions by using the Tools : Labels :
Edit labels option, see Section 6.8 for more details.

The phase velocity projection with overlaid mode shapes shown in Figure 8.3 displays some char-
acteristics of Lamb wave dispersion curves. The two fundamental modes, A0 and S0, propagate
from zero frequency. The fundamental p symmetric mode, S0, propagates with a phase and group
velocity equal to Young’s velocity ( E/(ρ(1 − ν 2 ) at zero frequency. In this case, the mode is
purely extensional, the motion is parallel to the direction of propagation, and the energy is evenly
distributed across the thickness of the plate. As the frequency increases, the S0 mode begins to
take on more complex characteristics. In addition to the in-plane displacements that are parallel
to the direction of propagation, the mode experiences many “normal” displacements that are nor-
mal to the plate (perpendicular to the direction of propagation). As the frequency continues to
increase, the S0 mode begins to act more and more like a pair of Rayleigh waves propagating on
the top and bottom surfaces of the plate. The energy and displacement of the Rayleigh mode is
concentrated in the outer portions of the plate, primarily in the first wavelength in depth. Rayleigh
waves are discussed in more detail in Section 8.3.

The fundamental anti-symmetric mode, A0, also displays some characteristic behaviours. The
mode begins at zero frequency and zero phase velocity as a pure bending mode but the phase and
group velocity quickly increase. Because of the sudden change in group velocity at low frequencies,
the mode is very dispersive in this region, indicating that wave packets will tend to spread out as
different frequency components travel at different speeds. Like the phase velocity of the S0 mode,
the phase velocity of the A0 mode tends to the Rayleigh velocity as the frequency increases.

The higher order modes are the modes that do not propagate at zero frequency. The modes
have been divided into two categories, symmetric (noted by “S”) and anti-symmetric (noted by
“A”), which describe the nature of the in-plane displacements of the mode. The numerical index
that follows the S or A notation indicates the order of the modes in increasing frequency of their
“cut-off” frequencies. The cut-off frequency of a mode is its low-frequency limit where the phase
velocity tends to infinity; this relates to through-thickness resonance of the plate. The higher
the index of the mode, the more maxima and minima will be present in the profile (mode shape)
of the displacements of the mode. In fact by examining the analytical solution, expressions can
be derived to specify the cut-off frequencies in terms of the bulk wave velocities. The cut-off
frequencies for the symmetric modes are given by

fcut = mv2 , m = 1, 2, 3 . . .
2n − 1
or v1 , n = 1, 2, 3 . . . (8.1)
2
and for the anti-symmetric modes by

fcut = mv1 , m = 1, 2, 3 . . .
2n − 1
or v2 , n = 1, 2, 3 . . . (8.2)
2
where the cut-off frequencies, fcut are in MHz-mm, the bulk velocities, v1 and v2 , are in mm/µs,
and m and n are integer counter variables.

Disperse version 2.0.20a


8.1 Lamb Waves 136

(a)

(b)

Figure 8.1: (a) Phase velocity and (b) group velocity dispersion curves for a typical Lamb wave
problem, a 1 mm thick steel plate in vacuum

Disperse version 2.0.20a


8.1 Lamb Waves 137

(a)

(b)

Figure 8.2: (a) Wave number and (b) angle of incidence dispersion curves for a typical Lamb wave
problem, a 1 mm steel plate in vacuum

Disperse version 2.0.20a


8.2 Leaky Lamb Waves 138

Figure 8.3: Mode shapes for the two fundamental Lamb wave modes overlaid on the phase
velocity dispersion curves.

The Shear Horizontal (“SH”) modes have not been included in the dispersion curves shown in Fig-
ure 8.1. The SH modes are decoupled from the Lamb wave modes and have all of their displacement
in the direction that is perpendicular to the direction of propagation and also perpendicular to
the through thickness direction. For a single isotropic layer in vacuum, the dispersion curves for
these modes can be expressed explicitly as shown in reference [4]. The dispersion curves for the
SH modes are shown in Figure 8.4. In order to trace these modes, the toggle that is originally set
to “Decouple SH modes” was changed to “Trace only SH modes” in the Trace : Solution type
dialog. Manual sweeps and traces were performed to trace the modes, see Section 4.3.

The fundamental SH mode propagates at a constant phase velocity equal to the bulk shear velocity
in the material. The higher order SH modes show some dispersion. They begin to propagate at
regularly spaced intervals equal to multiples of half of the bulk shear velocity. As the frequency
increases, the phase velocity of each of the higher order SH modes tends to the bulk shear velocity
in the layer.

8.2 Leaky Lamb Waves

For the purposes of this section, Leaky Lamb Waves refer to guided waves that exist in a single
elastic, isotropic layer that is immersed in a surrounding liquid. An example of common practical
use that will be developed in this section is a steel plate in water.

As explained in Section 7.7, the leaky Lamb case uses the same analytical functions as the Lamb
wave case. However, an extra term appears in the characteristic equation that accounts for the

Disperse version 2.0.20a


8.2 Leaky Lamb Waves 139

Figure 8.4: Phase velocity dispersion curves for the “SH” modes for a 1 mm thick steel plate in
vacuum

liquid loading on the surface of the plate. As the wave propagates down the isotropic layer, bulk
waves are generated in the surrounding fluid layers. These bulk waves carry energy away from
the system and cause attenuation. Therefore, in order to find a valid wave propagation solution,
the attenuation value must be found in addition to the frequency and real wave number; thus the
search for the roots of the dispersion curves is a search in three variables. Accordingly the solution
takes somewhat longer than the search in two variables for the strict Lamb Wave case.

In order to define the geometry for a Leaky Lamb system, you need to specify the following
parameters in the new materials dialog.

• Choose “Plate” as the geometry.


• Choose “Structure is embedded” in the second box.
• Specify the surrounding liquid in the first and third layers.
• Specify the material that makes up the plate and its thickness as the second layer, as in the
previous example.

If the surrounding material (defined in the first and third layers) are both the same non-viscous liq-
uid, an analytical function will be used to calculate the characteristic equation. If the surrounding
materials are different or they are not non-viscous liquids, the system will be treated as a general
multi-layered problem and the global matrix solution technique will be used. After accepting the
geometry, the automatic tracing routine can be used, as before, to find the dispersion curves.

Disperse version 2.0.20a


8.3 Surface and Interface Waves 140

The phase velocity curves for a steel plate in water are shown in Figure 8.5. In general, these
curves are very similar to the phase velocity curves for a steel plate in vacuum (Lamb wave
case) which were shown in Figure 8.1. For a material such as steel, which has a high acoustic
impedance, the water loading (with a much lower acoustic impedance) does not have much effect
on the phase velocity of the modes. However, there is a much greater effect for system such as a
graphite reinforced epoxy plate in water because the impedances are much more closely matched.
Although the water loading does not greatly affect the phase velocity curves for a steel plate,
one difference is immediately evident. For the water loaded case, there is a second low frequency,
anti-symmetric mode that does not appear in the Lamb case. This mode, called the Scholte wave,
is a wave that propagates at the interface of the steel and the water. Its phase velocity is always
less that the bulk velocity of the fluid and it experiences no attenuation. Its phase velocity tends
towards the bulk velocity in the water as the frequency is increased. The automatic tracing routine
may arbitrarily truncate the calculation of this mode when the velocity is very close to that of the
fluid. The grid mode shapes for this mode and also the A0 mode illustrate clearly the physical
nature of the leaky modes, see Section 6.1.2 for specific instruction with this example.

Although the phase velocity dispersion curves are very similar between a steel plate in vacuum and
one in water, the attenuation which is caused by the leakage into the water provides a significant
difference between the two cases. The attenuation (in decibels per meter) is shown in Figure 8.5.
The attenuation characteristics of the two fundamental modes provides a great deal of insight into
the behaviour of the leaky waves. The fundamental symmetric mode, S0, behaves entirely as an
extensional mode at zero frequency. Since practically all of the displacement is in the direction
along the interface between the steel and the water, and since this tangential motion does not
couple between the two materials, the wave in the plate takes no notice of the liquid loading
and the attenuation remains effectively zero. However, as the frequency increases and the mode
adopts also some lateral motion, the attenuation increases dramatically. At high frequencies, the
attenuation of the fundamental mode becomes asymptotic to the attenuation value for a leaky
Rayleigh wave. The attenuation of a leaky Rayleigh wave, if expressed per unit distance travelled,
is linear with frequency; if expressed as attenuation per wavelength it is constant. The fundamental
anti-symmetric mode, A0, behaves in a different fashion. This mode has significant bending motion
over its entire frequency range, which means that it always has significant lateral motion and so
we should expect that it should always transmit energy into the surrounding liquid. Indeed this
is true for most of its range. However, this mode does not behave this way at low frequencies. As
the phase velocity of the A0 mode drops below the bulk velocity of the surrounding liquid, the
attenuation drops to zero. This phenomenon is caused by the matching of the wavenumbers at
the interface of the two materials. A more complete description is given in Section 9.2.2, which
describes the leakage of a guided wave from a steel bar into a surrounding fluid.

8.3 Surface and Interface Waves

Rayleigh waves describe the guided wave that can exist on the surface of a semi-infinite half-
space. This type of guided wave was first described in the late 19th century by Lord Rayleigh
[81]. It has many applications in modern non-destructive testing. The energy of the propagating
wave is concentrated in approximately the top wavelength of the material; the amplitudes of
displacement and energy decrease exponentially away from the surface of the semi-infinite half-
space. Furthermore, because the wave does not have a propagation component into the half-space,
it does not lose any energy as it propagates. The surface nature of this wave and its low loss
characteristics allows it to effectively interrogate a structure for surface breaking cracks. The
depth of interrogation is controlled by adjusting the frequency at which the wave is generated.

Disperse version 2.0.20a


8.3 Surface and Interface Waves 141

(a)

(b)

Figure 8.5: (a) Phase velocity and (b) attenuation dispersion curves for a typical leaky Lamb wave
problem, a 1 mm steel plate in water

Disperse version 2.0.20a


8.3 Surface and Interface Waves 142

8.3.1 True Rayleigh Waves

The first case that we shall examine using Disperse is the strict Rayleigh wave case, a semi-infinite
half-space in vacuum. For this case, the Rayleigh wave will always propagate at the same velocity,
regardless of frequency. This velocity is both the phase velocity and the group velocity. This
velocity can be found by solving for the roots of the third order equation,
T 3 − 8T 2 + (24 − 16R)T − 16(1 − R) = 0 (8.3)
where T = (cr /c2 )2 , R = (c2 /c1 )2 , c1 is the bulk longitudinal velocity, c2 is the bulk shear velocity,
and cr is the Rayleigh velocity. If exact precision is not required, the Rayleigh wave velocity can
be approximated by the function,
 
0.87 + 1.12ν
cr ≈ c2 (8.4)
1+ν
where ν is Poisson’s ratio. However, Disperse does not use the analytical functions to calculate
Rayleigh waves, but relies on the general purpose Global Matrix method. In order to define the
system to trace a dispersion curve for a Rayleigh wave, it is recommended that you follow the
steps outlined below.

• Select “Plate” geometry.


• Select “Structure is embedded”.
• Leave the first material as the default “Free boundary”.
• Set the second and third materials to the material that you want to use as the semi-infinite
material.

It is recommended that you set both the second and third materials to the semi-infinite material
in order to improve the stability of the tracing routines and to ease the display of the mode shapes.
However, Disperse will function if only two layers are defined.

The user can examine the mode shapes for a Rayleigh wave in the same manner as any other
system. However, if the system was defined as only two materials (vacuum and an isotropic
medium), it is essential that the “Display in half-spaces” option is selected from the Advanced
menu, see Section 6.1. This option will cause the mode shape to be shown a certain distance into
the bulk of the semi-infinite half space. The amount of the material that is shown is controlled by
specifying a value in that menu. As described, the value (“X”, say) controls how many times the
total thickness of the internal layers will be shown. Therefore, if the system was defined as three
layers (vacuum-solid-solid) it will show X number of times the thickness of the second layer. If
the system was only defined as two layers (vacuum-solid), a thickness of one millimetre is assumed
for the internal thickness and so the value controls how many millimetres into the semi-infinite
half space is plotted. Remember that since the Rayleigh wave analytically behaves in the same
fashion regardless of frequency, the mode shapes that are shown can readily be scaled to any other
frequencies. The mode shapes for a Rayleigh wave on an aluminum half-space, for which the
velocity is 2.9244 km/s, are shown in Figure 8.6.

8.3.2 Leaky Rayleigh Waves

Leaky Rayleigh waves are very similar to Rayleigh waves except that the semi-infinite half-space is
loaded by a liquid instead of being surrounded by vacuum. This liquid loading causes the velocity

Disperse version 2.0.20a


8.3 Surface and Interface Waves 143

(a) Grid display (b) Line display

Figure 8.6: The mode shape of a Rayleigh wave propagating on the surface of an aluminum
half-space, shown (a) as a deformed grid and (b) as lines representing the displacement profiles.

of the Rayleigh wave to increase slightly. In addition, attenuation is introduced into the system
because of the radiation (leakage) of bulk longitudinal waves into the fluid. The attenuation (per
unit distance) increases linearly with frequency.

The system can be defined and traced in the same fashion as for the true Rayleigh wave case
except that the fluid material is put in place of the vacuum half-space (the first layer). The same
restrictions also apply for the mode shapes. However, when the “Display in half-spaces” option
is selected, a portion of the liquid half-space will also be shown. By looking at the mode shapes
in this system it is possible to differentiate between the behaviour of the Rayleigh wave in the
solid, which dies away from the interface, and the bulk wave present in the liquid loading material,
which propagates as a bulk wave away from the interface. The mode shapes for a leaky Rayleigh
wave propagating on an aluminum half-space with water loading can be seen in Figure 8.7. The
leaky Rayleigh in this case propagates at 2.937 km/s and has an attenuation (per unit distance)
which is linear with frequency.

8.3.3 Stoneley Waves

Stoneley waves exist at the interface between two solid semi-infinite half-spaces. They can only
exist for a certain range of material properties that are given in reference [88]. One such material
combination, that will be explored below, is aluminum-steel.

A Stoneley wave system is defined in the same manner as a Rayleigh system, except that the
material of the upper semi-infinite half space should be defined as the first layer instead of the
vacuum or liquid layer that was used for the previous cases. The automatic routine may not
work well for this system, so the user may need to start tracing the curves by doing a phase
velocity sweep (Trace : Sweep, see Section 4.3). The sweep should reveal a root at the Stoneley
wave velocity. A true Stoneley wave has zero attenuation and propagates at a velocity which is
independent of frequency and is between the shear and Rayleigh velocities of the denser material.

Figure 8.8 shows the in-plane and normal displacement for a Stoneley wave at the interface between

Disperse version 2.0.20a


8.3 Surface and Interface Waves 144

(a) Grid display (b) Line display

Figure 8.7: The mode shape of a leaky Rayleigh wave propagating on the surface of an aluminum
half-space which is water loaded, shown (a) as a deformed grid and (b) as lines representing the
displacement profiles.

semi-infinite half spaces of aluminum and steel. In this case, the aluminium was modelled with a
density of 2700 kg/m3 , v1 = 6348 m/s, v2 = 3185 m/s and the steel was modelled with a density of
7932 kg/m3 , v1 = 5960 m/s, v2 = 3172 m/s. If a different type of steel which has a shear velocity
greater than the shear velocity in the aluminum is used, then the Stoneley wave will not exist.
The Stoneley wave in this system propagates at a constant phase and group velocity of 3.1299
km/s and has no attenuation. The displacements in Figure 8.8 reveal that the wave is trapped at
the interface of the metals, propagating along the interface and with displacements and stresses
which decay in the direction away from it.

8.3.4 Thin Layer on a Half-Space

A case that is similar in construction to a Rayleigh wave case is a thin layer of a solid material
on a semi-infinite half-space. The example which we will examine in this section is a thin epoxy
layer on a semi-infinite aluminium substrate.

Since a Rayleigh wave ’penetrates’ farther into a material the lower its frequency, the guided wave
that exists in this multi-layered system changes its behaviour depending on the frequency. At low
frequencies (when the wavelength is long in comparison to the thickness of the epoxy layer), the
wave penetrates deep into the aluminium substrate, this material then dominates the response.
However, as the frequency increases, the effect of the thin epoxy layer becomes increasingly strong.
The phase velocity of the guided wave becomes asymptotic to the Rayleigh velocity in the epoxy
layer as the frequency increases. In addition, because of the layering of this system, more than
just one mode is present.

In order to define this system,

• Select “Plate” geometry.

Disperse version 2.0.20a


8.3 Surface and Interface Waves 145

(a) In-plane displacement (b) Normal displacement

Figure 8.8: The (a) in-plane and (b) normal displacement components of a Stoneley wave prop-
agating at the interface of aluminum and steel half-spaces. The vertical axis corresponds to the
direction perpendicular to the interface and shows a depth of approximately 5 wavelengths of the
Stoneley wave.

• Select “Structure is embedded”.


• Set the first material to “Free boundary”.

• Set the second material to “Epoxy” and specify the desired thickness.
• Set the third material to “Aluminium”.

Once this system is defined, the modes will probably need to be traced manually. An effective
method to do this begins by performing a frequency sweep (Trace : Sweep) at a phase velocity
between the Rayleigh velocities of the epoxy and aluminium materials: for each material this can
be approximated to be 90 % of the bulk shear velocity. This sweep should find several valid roots.
The modes can then be traced from these points by the usual method, for example Trace from
swept point or click the button on the toolbar, or select Trace from on the right mouse menu,
see Section 4.3. It is suggested that you first trace the modes back to zero frequency. Once all of
the modes have been traced, then set the frequency limit in the toolbar to the upper limit that is of
interest, and click on the high frequency end of the partially traced dispersion curves to complete
them. In order to look for missing modes, Find minima can be selected from the right hand
mouse button menu and an area can be outlined on the graph by clicking and dragging with the
left hand mouse button, again see section on manual tracing (4.3). Minima of the characteristic
equation will be drawn as circles. Since there is attenuation in the system, the locations of the
minima will not exactly correspond to the location of the dispersion curves. However, it should
still be possible to identify modes that have not already been traced; such modes will appear as a
line of minima that do not have a traced dispersion curve anywhere near them. In order to trace
these modes, select Sweep and Search from the right hand mouse button and start a sweep
over the region where the mode is likely to fall by clicking and dragging with the left hand mouse
button. Note that the value of attenuation which is assumed when performing the sweep and
search is not necessarily zero, but is the value most recently picked out by the pointer.

The complete set of dispersion curves for this system is shown in Figure 8.9. Figure 8.10 shows
only the fundamental mode, which acts as a Rayleigh wave on an aluminum half-space at low
frequency and then changes gradually until it acts like a Rayleigh wave on an epoxy half-space at
high frequency. Each of these types of behaviour can be seen in Figure 8.11.

Disperse version 2.0.20a


8.3 Surface and Interface Waves 146

Figure 8.9: Phase velocity dispersion curves for a 0.1 mm thick epoxy layer on an aluminium half
space.

Figure 8.10: Phase velocity dispersion curve for the fundamental mode of a 0.1 mm thick epoxy
layer on an aluminium half space.

Disperse version 2.0.20a


8.4 Visco-Elastic Layers 147

(a) 2.0 MHz (b) 8.0 MHz

Figure 8.11: The grid mode shapes for the fundamental mode of a 0.1 mm thick epoxy layer on
an aluminium half space at (a) 2.0 MHz and (b) 8.0 MHz, showing the change in behaviour from
a mode that is dominated by the properties of the aluminium properties to one that is dominated
by those of the epoxy.

8.4 Visco-Elastic Layers

Disperse is capable of modeling both elastic and visco-elastic isotropic materials. This section
discusses some issues concerning the propagation of guided waves in a single layer of a visco-
elastic isotropic material.

The technique for specifying visco-elastic layers is the same as the technique for elastic layers.
However, when the material is defined, both the shear and longitudinal attenuation value must
be specified. Disperse uses a model which assumes that the attenuation per unit distance due
to visco-elastic losses is linear with frequency for the component bulk waves, that is to say there
is a constant loss per wavelength travelled. The attenuation is specified in units of nepers per
wavelength, indicating that the amplitude of a wave of unit amplitude will decay to e−α after
propagating one wavelength if the attenuation, α, is in nepers. One way of obtaining these
values involves examining the successive reflections of a bulk wave in two different thicknesses of a
material. Assuming that the boundary conditions for the two blocks of material are the same, the
difference in the decrease of successive reflections can be related to the attenuation. This simple
technique works relatively well for obtaining the longitudinal wave attenuation in a water bath,
but less effectively for the shear wave attenuation. The shear wave attenuation is often much
higher than the longitudinal attenuation and does not allow successive reflections. Therefore the
first reflection will need to be used. More sophisticated methods can also be employed that take
advantage of critical angles to generate shear waves in the material or that model the predicted
amplitude of various reflections.

Tracing the dispersion curves for visco-elastic layers is also very similar to the elastic case. For
lightly attenuating materials, for example a metal, the dispersion curves look very similar, except
that the visco-elastic dispersion curves will now have attenuation values associated with each point.
In addition, the S1 mode tends to divide into two separate modes, one of which corresponds to
the portion of the S1 mode for which the phase and group velocities are in the same direction,
and the other which corresponds to the portion for which the phase and group velocities are in
opposite directions. In contrast, these two portions are joined in the same mode, for the elastic
case.

As the attenuation becomes more significant, for example 0.1 nepers per wavelength, the shape

Disperse version 2.0.20a


8.4 Visco-Elastic Layers 148

(a) Phase velocity (b) Attenuation

Figure 8.12: The (a) phase velocity and (b) attenuation dispersion curves for a layer of high
performance polyethylene.

of the dispersion curves can change dramatically depending on whether the materials are treated
as elastic or visco-elastic. For example, the dispersion curves for a layer of high performance
polyethylene (HPPE), a relatively high loss plastic, which are shown in Figure 8.12, change dra-
matically. Because of the large difference in attenuation values for the longitudinal and shear
bulk waves, new mode behaviours develop. Certain “low attenuation” modes appear that are
dominated by longitudinal wave components and tend to the bulk longitudinal velocity instead of
the bulk shear velocity as the frequency increases. The modes are further discussed by Chan [15]
and some experimental measurements are also given in [7]. In order to trace these modes, it is
helpful to retain the default tracing step type “Freq-Wavn Combo” in the Trace : Convergence
parameters dialog. This takes steps along the tangent to the frequency-wavenumber curve and
iterations normal to the tangent, and so is better able to follow curves which take irregular paths
than the procedures which simply take steps in wavenumber or frequency; see Section 7.9. In this
case the frequency-wavenumber dispersion curves form some loops in the shape of the letter “e”
and so the solution really requires such a scheme.

With highly attenuating materials, the behaviour of the modes at their cut-off points changes.
For the elastic case, all of the higher order modes begin at zero real wavenumber at a certain cut-
off frequency. No propagating roots exist for a given mode below its cut-off frequency, although
some non-propagating roots do (roots that have zero real wavenumber but a non-zero imaginary
wavenumber, see [4]). However, when the attenuation is added to the system, the formerly non-
propagating roots now have a real wavenumber component (although it is small in comparison to
the imaginary wavenumber component). Since Disperse will trace any modes that have a positive
real wave number, these portions of the modes will be traced. They appear as the section of
the mode that extends for the high phase velocity point of the mode to a zero frequency - zero
phase velocity limit. Although these portions may have an interesting theoretical significance,
they have little practical significance for non-destructive testing and are normally deleted beyond
a reasonable attenuation value. In order to delete a portion of a curve, click the Split one mode
into two icon on the tool bar (this is on the top row, immediately to the right of the Delete
point button), then click on the desired location on the curve: this separates the mode into two
modes within the program. You can then simply delete the unwanted part using the Delete an
entire mode button on the toolbar. Further information about these tools is given in Section
5.4.3.

Disperse version 2.0.20a


8.5 Multi-Layered Structures 149

8.5 Multi-Layered Structures

Most of the examples that have been discussed in this chapter so far have dealt with systems that
do not have many layers. This section begins to look at systems that comprise several layers.

An adhesive joint is a common multi-layered structure that is analysed in non-destructive testing.


This section analyses a simple aluminum - epoxy - aluminum joint under four conditions, vacuum
loaded, water loaded, with weakened epoxy, and with a very weak epoxy. The next section
examines a related problem, the guided waves that can exist in an epoxy layer that is embedded
in two semi-infinite half-spaces of aluminum. Some further discussion about both of these types
of problems in the context of non-destructive testing is given in [54].

In order to define the layers of the adhesive joint in vacuum, selecting File : New as usual, then
“Plate” and “Structure is in vacuum” in the dialog box. The first item of the layers information
is the top half space, which in this case remains its default “Free boundary”. Enter the properties
of the top aluminium layer in the second item. Now click the “” button to the left of this item,
and select “Insert after” from the drop-down menu which appears. A new item in the layers list
will appear. Repeat this operation to get yet one more layer. You now have enough layer items
in the list to complete the definition of the three layers, the final layer being “Free boundary”.
When defining the epoxy for the middle layer we could simply choose epoxy from the pulldown
menu. However, since we will later want to simulate a weakened epoxy by reducing the stiffness
of the epoxy layer, it is helpful to describe the epoxy layer in terms of its Young’s Modulus and
Poisson’s ratio. To do this, click on the “” button to the left of the item and select “Edit” from
the drop-down menu which appears. Now click the “Isotropic” tab to indicate that you want to
define an isotropic material. Select “Solid defined by modulus” in the pull-down menu at the top
of the dialog, then enter values for Young’s modulus, Poisson’s ratio and density. If the properties
are not known in terms of Young’s modulus and Poisson’s ratio then they can be converted simply
from bulk velocity values: choose “Solid defined by velocities” from the top pull-down menu, then
enter the bulk velocity values, then select “Solid defined by modulus” and the properties will be
converted and will be retained by the program as modulus values. Then enter a name for the
material and click “Add to list” so as to save it for use later. Further description of the definition
and editing of materials is given in Section 3.2.

The automatic tracing routines will work sufficiently well for most adhesive joints. However,
problems will occur if the epoxy layer is thick in comparison to the thickness of the surround-
ing aluminium. In addition, the automatic routine can frequently experience problems tracing
irregular multi-layered structures. This is because it is difficult for the routine to predict approx-
imately where the roots are likely to be in a arbitrarily multilayered system. Furthermore, since
the general purpose global matrix system is used for calculating the characteristic equation, the
symmetric and anti-symmetric solutions can not be separated, so the tracing routines may have
problems when two modes (a symmetric and an anti-symmetric) cross or run very close together.
Therefore start with the automatic tracing to get as far as you can, but then be prepared to do
a lot more of the root searching yourself, using the manual searching procedures, see Section 4.3.
Use the “Sweep and search” function (right mouse button) to find roots in the various regions of
graph (we suggest you search in the phase velocity - frequency space by preference because this
allows you to see clearly the any spurious points which correspond to bulk velocities - see Sections
7.5, 4.2). If you find individual spurious points, then you can delete them using Edit : Delete :
Next point selected or the Delete button on the toolbar. If part of a curve is unsatisfactory
then you can retrace that part by clicking on the Trace from point icon on the toolbar and then
clicking somewhere near the end of the good region of the curve; Disperse will immediately erase
all points after that location and repeat the trace from there. Finally, the smoothness or lack of
smoothness of the solution is most evident when you plot the group velocity curves because these
are found by numerical calculation of the derivative of the wavenumber - frequency curves. This

Disperse version 2.0.20a


8.6 Embedded Layers 150

Figure 8.13: The phase velocity dispersion curves for a aluminum-epoxy-aluminum adhesive joint
when the epoxy is 0.1 mm thick and has a stiffness of 4 GPa and the plates are 1 mm thick.

can be used as a final check for any bad points.

The dispersion curves for a typical adhesive joint in vacuum are shown in Figure 8.13, and are
not particularly remarkable. They resemble the dispersion curves for a single aluminum layer.
The main noticeable difference is that the modes tend to clump in pairs. As the epoxy layer
becomes thicker and weaker, this clumping becomes even more pronounced. Because of the close
proximity of the modes, the dispersion curves for weakly bonded structures are more difficult to
trace than the strongly bonded counterparts. The dispersion curves for a liquid loaded adhesive
joint in Figure 8.14 resemble those for a liquid loaded plate in the same way that the dispersion
curves for a vacuum loaded adhesive joint resemble those of a free plate.

Many non-destructive testing researchers are interested in investigating the integrity of the adhe-
sive layer in an adhesive joint. Therefore, it is interesting to examine the effect that changing the
properties of the epoxy has on the dispersion curves of the entire system. Two such modifications
are shown in Figures 8.15(a) and (b) which show the dispersion curves for a system that has an
epoxy whose stiffness is reduced by 25 percent and the dispersion curves for a system that has an
epoxy whose density is reduced by 50 percent and whose stiffness is reduced by 75 percent.

8.6 Embedded Layers

In this section, embedded layers refer to layers that are embedded in an infinite extent of another
material. Usually this represents a thin layer which is embedded between relatively thick solids.
The example that is investigated is an epoxy layer that is embedded in two semi-infinite half-spaces
of aluminum. This model allows the possibility of examining the propagation of waves that are
trapped in an epoxy layer. Further discussion of this example may be found in [54].

Disperse version 2.0.20a


8.6 Embedded Layers 151

(a) Phase velocity (b) Attenuation

Figure 8.14: The (a) phase velocity and (b) attenuation dispersion curves for a aluminum-epoxy-
aluminum adhesive joint in water. The plates are 1 mm thick and the epoxy is 0.1 mm thick.

(a) Epoxy modulus 3 GPa (b) Epoxy modulus 1 GPa

Figure 8.15: The phase velocity dispersion curves for the example aluminum-epoxy-aluminum
adhesive joint, when the epoxy has been artificially weakened by (a) lowering the stiffness to 3
GPa and (b) lowering the stiffness to 1 GPa and reducing the density by half.

Disperse version 2.0.20a


8.6 Embedded Layers 152

The wave propagation roots fall into three categories, waves that are entirely trapped in the
epoxy, waves that leak shear waves into the surrounding aluminum, and waves that leak shear
and longitudinal waves into the surrounding medium. The non-leaky modes, whose attenuation
is zero, are relatively easy to trace over most of their length. The waves that leak only shear
waves are also reasonably easy to trace. However, the modes that leak energy into both shear and
longitudinal waves are generally quite difficult to trace because of their high attenuation values.

In order to define this embedded system, choose “Plate”, then “Structure is embedded” in the
new material dialog. Then simply enter the three layers in the three rows. The first and third will
show the word “Infinite” at the right hand side, indicating that these are the semi-infinite half
spaces. In this example the first and third layers are the aluminium and the middle layer is the
epoxy.

In general the modes for this kind of system are difficult to trace, so start with the automatic
routine but expect to have to do quite a lot of manual searching yourself.

It is best to trace the non-leaky modes first. These may be found by performing a frequency sweep
at a phase velocity equal to about two thirds of the bulk shear velocity of the surrounding material
and zero attenuation. Each of the roots that is found can be traced forward and backwards in the
usual way. Many of the modes will appear to stop at the bulk shear velocity of the surrounding
medium when they are traced back to zero frequency. These modes are actually continued above
the bulk shear velocity, however, the two portions of the modes are discontinuous in this solution
space (although they should be connected in the imaginary plane). An explanation why these
modes do not leak energy into the surrounding liquid is given in Section 9.2.2, which describes
how the wavenumbers have to be coupled at the interface of the layers and how this relates to
which leaky waves can be generated.

The second portion of the modes, those that leak only shear waves into the surrounding media,
are found between the shear and longitudinal bulk velocities of the surrounding medium. The
attenuation of these sections of modes varies greatly, which can make them difficult to find. One
method of finding a large number of these roots is to perform a frequency sweep at a phase velocity
that is half way between the bulk shear and longitudinal speeds of the surrounding aluminum and
with zero attenuation. Once these roots have been traced, a second frequency sweep can be
performed at a higher attenuation value to see if some more roots are identified. Because of the
transition between different types of leakage, many of these mode sections will appear to end as
they reach the shear or longitudinal bulk velocities of the surrounding aluminum. In addition, the
tracing routine may tend to jump to other modes. If this happens, the tracing can be stopped by
pressing the “Stop” button, and the bad points can be deleted. If you have trouble with modes
jumping try changing the values of the “Increment” and the two iteration “search widths” which
are set in the Trace : Convergence parameters dialog. See Section 4.3 for details.

The third region of roots, those that leak both longitudinal and shear waves, are very difficult to
trace. The attenuation values are high and the tracing routines tend to jump about. In order
to trace these portions of the dispersion curves, the user will need to conduct another frequency
sweep to find starting points, but this time the sweep needs to be conducted at a phase velocity
greater than the longitudinal bulk velocity. Beware of roots that have negative attenuation values;
they tend to correspond to portions of the modes that have negative group velocity. In this system,
these portions tend to run from the cutoff limit of a high order mode back to the origin. Their
meaning is currently not well understood.

Disperse version 2.0.20a


8.7 Anisotropic Layers 153

(a) Phase velocity (b) Attenuation

Figure 8.16: The (a) phase velocity and (b) attenuation dispersion curves for an epoxy layer
embedded in aluminum half-spaces.

8.7 Anisotropic Layers

All of the previous sections have dealt with isotropic materials. However, Disperse is capable
of handling anisotropic materials within certain limitations (1.2). The anisotropic solutions are
necessarily slower than their isotropic counterparts, however it is still possible to trace dispersion
curves for a wide range of systems and for multiple layers. Multi-layered structures are represented
by large global matrices so the solution is slowed down considerably by the addition of each extra
layer. Therefore it is advisable to make use of the “repeated layer” and “symmetry” facilities
which are defined in Section 3.2.

It is helpful to keep in mind two classes of problem for anisotropic structures. If the the material is
orthotropic and the propagation is along a principal axis, then the solution proceeds conceptually
in the same way as for an isotropic material. That is to say, the “Lamb” type modes can be
described by the superposition of two longitudinal and two shear bulk waves in the layer. How-
ever, if the propagation is along a non-principal axis then the partial waves are coupled and the
solution requires six partial waves in the layer, two “quasi-longitudinal” and four “quasi-shear”.
This requires an eigenvalue solution to find the partial waves and is therefore somewhat more
complicated and lengthy to solve. The angle of propagation is predetermined when the material
layers are defined: that is to say, the angles of the orientation of the layers are specified with
respect to the propagation direction (3.3.3). Note, however, that the propagation direction here
is the direction of the phase of the waves; a wave packet (the energy) does not in general travel in
the same direction as the phase.

Defining the materials is much more difficult for anisotropic materials than it is for isotropic ones.
Disperse provides a range of different formats for the specification of the anisotropic materials,
and each of these may readily be converted to any of the others by selecting from the menu at
the top of the anisotropic materials definition tab, see Section 3.3.3. In addition to specifying the
stiffnesses (or engineering constants), the orientation of the material must be specified in terms of
three rotation angles which are defined in Figure 3.1.

We will consider three examples of anisotropic plates here: (1) a unidirectional composite in which
the propagation direction is aligned with the fibres (zero degrees), (2) the same plate with the
propagation at 45 degrees to the fibre direction and (3) a multilayered composite plate.

Disperse version 2.0.20a


8.7 Anisotropic Layers 154

(1) Unidirectional composite with propagation along the fibre direction.

This material constitutes a degenerate case of the general anisotropic system and is often referred
to as transversely isotropic, reflecting the fact that it exhibits rotational symmetry about an axis
parallel to the fibre direction. This means that the material can be completely specified by only five
independent constants. In this example we shall assume that the material is both non-attenuative
and homogeneous. Such assumptions are valid in the low ultrasonic frequency range up to a few
Megahertz.

This example discusses guided wave propagation in a single unidirectional laminate composed of
the graphite-epoxy composite material designated T300/914 and studied by Simon ([95]). We
shall trace the dispersion curves for modes propagating in the direction of the fibre axis. The
laminate is made up of 32 parallel plies of T300 carbon fibre embedded in a 914 epoxy matrix.
It can thus be considered here to be a single layer. It is possible by various methods to measure
the properties of anisotropic materials such as this and thus find their elastic constants (Cij ) and
density (ρ). Literature on this subject is very extensive, but see for example Chapter 15 of [63].
The properties of T300/914, given in appendix E of Simon’s paper, are as follows:

C11 = 143.8GP a C12 = C13 = 6.2GP a C22 = C33 = 13.3GP a


C23 = 6.5GP a C44 = 3.6GP a C55 = C66 = 5.7GP a
ρ = 1560kg/m3 (8.5)

In this case the reference co-ordinates for these data for the material are orientated such that
the fibre axis lies parallel to the 1 direction, and the 3 direction is normal to the median plane
of the laminate. To begin with, the system is defined, as usual, by selecting File : New and
accepting the defaults of a “plate” and “structure is in vacuum”. Assuming that this material
has not previously been defined the “” button is clicked adjacent to the “Undefined” layer and
“Edit” is selected from the menu. Since, as observed above, this is a special anisotropic case, the
“Anisotropic” tab is selected. After typing in an appropriate name (eg “Graphite-Epoxy-T300-
194”) and abbreviation (eg “T300-194”) for the material, the appropriate material and parameter
types must be set in the left-hand list-edit box beneath. For this case “Trans. Isotropic (Cij’s -
along x)” is the correct choice. The phrase “along x” refers to the axis of symmetry of the material
as it is specified and has nothing to do with the propagation direction. If the material properties
are known only in terms of engineering constants then the appropriate “engineering constants”
option can be selected instead, ensuring, of course, that the correct symmetry axis is specified.
The given Cij’s can now be entered in the edit boxes representing one half of the elastic stiffness
matrix for the material, ensuring that the units are correctly specified. Note that some of the
elements of the stiffness matrix of a transversely isotropic material are interdependent. In this
case:

C22 = C33 C13 = C12 C55 = C66 C22 = C33 C44 = 0.5(C33 − C12 ) (8.6)

Finally the density must be entered.

The direction of propagation in Disperse is always in the Z or 3 direction, yet the fibres lie in the
X or 1 direction so it is now necessary to rotate the material to the correct orientation, so that the
propagation will be along the fibres, as required for this example. This will necessitate a rotation
of 90 degrees about the Y axis (see Figure 3.1) which is effected by typing 90 in the “theta”

Disperse version 2.0.20a


8.7 Anisotropic Layers 155

(a) Phase velocity (b) Group velocity

Figure 8.17: The (a) phase velocity and (b) group velocity dispersion curves for modes propagating
parallel to the fibre direction in a unidirectional composite laminate.

rotation edit box. As we require propagation in the fibre axis direction no further rotations are
necessary. For a more detailed explanation concerning material definition and these rotations see
Section 3.3.3. Having defined the material, its definition is saved by clicking on the “Add to list”
button before clicking the OK button.

Entering the thickness of the laminate completes the geometry definition and the automatic tracing
routine can be initiated as in previous examples. The phase velocity and group velocity spectra
for this example are shown in Figure 8.17.

This simplest of anisotropic cases poses little problem for the tracing routine, particularly since,
in this ideal case, no material attenuation was specified. As one might expect in a material such
as this, with a high fibre-volume-fraction, the phase velocities of the modes propagating along the
fibre direction are dominated by the bulk velocities along the axis of the fibres and have relatively
large frequency ranges with very little dispersion. Modes tend to follow the longitudinal bulk
wave velocity in the fibre wherever in-plane displacement dominates the mode shape, dropping
sharply to subtend to the Rayleigh wave velocity as out-of-plane displacement becomes dominant
at higher frequencies. In Figure 8.17, both the vertically polarised and horizontally polarized
modes have been traced. For propagation in the principal directions, parallel and normal to the
fibre axes, the vertically polarised and horizontally polarised partial waves are de-coupled and
there is consequently no displacement in the Y direction. In this case mode conversion between
vertically-polarised and horizontally-polarised (SH) modes is not possible and vertically-polarised
excitation will not generate SH modes. These restrictions will not apply to modes propagating in
non-principal directions as will be seen in the next section. Further discussion of these aspect can
be found in standard texts such as [4, 63].

Unlike those of the multilayered composite systems considered shortly, the mode shapes of modes
propagating in the fibre direction through a unidirectional laminate are smooth and similar to the
equivalent modes in isotropic material. The displacement and stress mode shape for the S0 mode
at 1MHz-mm is shown for example in Figure 8.18, in which the y, or perpendicular, displacement
and (xy) shear stress are zero.

(2) Unidirectional composite with propagation at 45 degrees to the fibre direction.

To illustrate the effect of changing the propagation direction, so that modes are no longer propa-
gating parallel to the fibre axis, we shall now consider the case of modes in the same composite
laminate defined above, propagating parallel to the median plane of the laminate, but at an angle

Disperse version 2.0.20a


8.7 Anisotropic Layers 156

Figure 8.18: (a) Displacement and (b) stress mode shapes for S0 mode parallel to fibres in
unidirectional composite at 1 MHz.

of 45 degrees with respect to the axis of the fibres. Disperse does not, of course, restrict us to
propagation in this plane and any arbitrary direction can be set using the three rotations as de-
scribed in Section 3.3.3. The material was defined and saved in the case described above, and so,
having selected File: New, we can simply select the material: “Graphite-Epoxy-T300-914”, from
the list-edit box. This material was saved with a 90-degree theta rotation necessary to orientate it
to produce propagation along the fibres. This can be confirmed by selecting “” and edit to view
the material specification. In order to re-orientate the laminate for propagation at 45 degrees to
the fibre axes, a further rotation of 45 degrees about the X axis is required and this is effected by
typing 45 into the “Psi” edit box. The material specification can now be accepted, and, having
entered the thickness of the laminate (1mm) in the geometry definition, the automatic tracing
routine can be started (accepting the defaults) as in the previous case. In this case, the option to
trace only the modes polarised in the sagital plane, found in the simple isotropic material cases,
is now no longer available. This is because the horizontally and vertically polarised partial waves
are not de-coupled, except when the propagation vector is either parallel or perpendicular to the
fibre axes.

Propagation along a non-principal axis presents considerably greater complexity for the tracing
routine, which is consequently slower. In this case, where modes of similar type have very close
proximity in the phase velocity spectrum, jumping may occur from the trace of one mode to
another. Where this occurs the invalid segment can be removed by means of the “Split one mode
into two” and “Delete an entire mode” toolbar buttons as described in Section 5.4.3. The iteration
step and tolerance in the convergence parameters can then be tightened (say to 3e-003 and 5e-0007
respectively) and the problem segment retraced. Finally, the mode segments are joined using the
“join segments” button. Often it is quicker simply to erase the entire mode and redraw it, having
tightened the iteration convergence parameters.

The phase velocity and group velocity spectra for this case, which are presented in Figure 8.19, are
significantly different from those of Figure 8.17 for propagation along the fibres. As the propagation
angle with respect to the fibre axis increases, the phase velocity of modes with predominantly in-
plane displacement decreases, so that the non-dispersive frequency ranges of these modes are seen
at a lower phase velocity. In general there is less phase velocity variation of the modes. Note that
the group velocity here is calculated simply by the ω-k derivative expression in the usual way; in

Disperse version 2.0.20a


8.7 Anisotropic Layers 157

(a) Phase velocity (b) Group velocity

Figure 8.19: The (a) phase velocity and (b) group velocity dispersion curves for modes propagating
at 45 degrees to the fibre direction in a unidirectional composite laminate.

Figure 8.20: (a) Displacement and (b) stress mode shapes for S0 mode at 45 degrees to fibres in
unidirectional composite at 1 MHz.

fact the propagation of the energy of wave packets in anisotropic materials is considerably more
complicated: it is not parallel to the phase direction except in the special cases of propagation
along the principal axes. Disperse currently does not address this (1.2).

Propagation in a non-principal direction with respect to the fibre axes means that the vertical
and horizontally polarised shear waves are now coupled and consequently, the dispersion curves
will always include SH modes. The symmetric SH modes are marked S’n in Figure 8.19, while
the antisymmetric ones are marked A’n. However, since the direction displacement vector changes
with frequency a particular curve cannot simply be regarded as either a horizontally or a vertically
polarised mode. For example the vertically polarised S1 mode becomes horizontally polarised
above about 2MHz-mm. Nevertheless in general the mode shape of the modes are similar to those
for propagation in the fibre direction and indeed those of the isotropic material case, so that the
perpendicular (y) displacement and the shear stress (xy) are now significant in the mode shapes.
This is seen in the S0 mode shape presented in Figure 8.20.

Disperse version 2.0.20a


8.7 Anisotropic Layers 158

Table 8.1: Stacking sequence of example 8-ply composite.

+45 degrees
-45 degrees
0 degrees
90 degrees
90 degrees
0 degrees
-45 degrees
+45 degrees

(3) Multi-layered composite, with unidirectional laminates, in a quasi-isotropic layup.

In order to illustrate the modelling of multiple laminate composite material, more commonly em-
ployed in structures, and to discuss typical aspects of guided waves in such material, we shall
examine a quasi-isotropic composite consisting of eight unidirectional plies. The material in ques-
tion, a graphite-epoxy T800/924, was studied by Percival and Birt ([71]), who give the elastic
parameters of a single layer as:

E11 = 161GP a E33 = 9.25GP a G13 = 6.0GP a


ν13 = 0.34 ν23 = 0.41 ρ = 1.5g/cm3 (8.7)

As in the previous cases discussed, the material of a single ply has been defined by the authors
such that the 3 direction is normal to the laminate plane and the 1 direction lies parallel to the
axes of the fibres. When the laminate was constructed the plies were stacked with the orientations
given in Table 8.1.

This material is often designated as: [+,-,0,90]4S (as in Percival’s paper), or [+45,-45,0,90] S,
where the subscript S indicates the symmetric orientation of laminates about the mid-plane.

After selecting File: New the first task is to enter the new material. Click the “” button and
select “Edit” followed by the “anisotropic” tab as in previous examples. Since we have been given
engineering parameters, select “Trans. Iso. (Engr. sym-axis along x)” and enter the parameters
given above, ensuring that the units are correct. Before saving, it is useful to rotate the material
by 90 degrees about the Y axis (a theta rotation) so that the fibres lie parallel to the Z axis,
thus the orientation of each layer will necessitate only a single rotation about the X axis (a Psi
rotation). Finally, give the material an appropriate name and abbreviation and click “Add to
List” for future use, before closing the form with OK.

Next we shall define the other layers in the Geometry Definition Form. Since the material is
symmetric about the median plane of the plate it is only necessary to explicitly define the first
four layers. These are added by clicking the “” button adjacent to the list edit box marked
“Free Boundary”. Select “Insert Before” from the menu and repeat this twice more so that we
have a total of four layers between the top and bottom free boundaries. Insert the newly defined
material in each layer by selecting from the menu in the list-edit box for each layer. Also type in
the thickness of each layer; 0.125mm in this case. Next we must correctly orientate each layer.
This is done for each layer in turn by, once again, clicking the “” button and then selecting
“Edit” from the menu. The appropriate rotations about the X axis, defined in Table 8.1, are
now entered in the “theta” rotation box before clicking OK. Having now fully defined the four

Disperse version 2.0.20a


8.7 Anisotropic Layers 159

(a) At 0 degrees (b) At 45 degrees

Figure 8.21: Phase velocity curves for propagation in an 8 ply quasi-isotropic laminate, (a) in
the direction of the fibres in layers 3 and 6 and (b) at 45 to that in (a).

layers, simply change the“Free boundary” at the bottom of the list to “Symmetric system”. This
will automatically define the lower four layers of the 8-layer stack to be the mirror image of the
upper four layers. This completes the material definition. Often laminar materials are composed
of several identical sub-systems of layers, usually referred to as “super-layers”. In such cases a
super-layer can be defined by clicking the “” button and selecting “Repeated Group” from the
menu, see section 3.2. The layer is then transformed to represent a group of two layers to which
further layers can be added as defined before. The edit box to the right of the Group layer now
defines the number of super-layers rather than the thickness.

The system has now been set up for propagation in the direction parallel to the fibres of the 3rd
and 6th layers, that is the zero degree direction. Tracing the dispersion curves is initiated in the
usual manner using Trace : Automatic Tracing or F7. Although the tracing routine proceeds
considerably slower than in previous cases, as one should expect given the increased complexity,
in this example all the modes apart from the two fundamental ones are traced by the automatic
routine. However, the routine has some difficulty at low phase velocities and some modes are
not fully traced to the upper frequency limit, when the matrix becomes singular (error code 0).
Sometimes this is resolved just by re-tracing from the truncated end of the mode, but usually it is
necessary first to decrease the iteration search width in the convergence parameters (F11) see also
Section 4.3.3. The most difficult curve to trace is that of A0. This requires tracing from a zero at
about 1.5 MHz with the convergence parameters being set to “Wavenumber-combo” step with a
width of 0.01 and an iteration search width and tolerance of 0.005 and 5e-006. To trace the lower
frequency segment the search width is reduced to 0.001. The resulting phase velocity curves are
shown in Figure 8.21 part (a) and those for propagation in the same material, rotated through 45
about the X axis (“Psi” rotations of all layers increased by 45) are presented in part (b).

Comparing the dispersion curves in this figure with the unidirectional case of Figure 8.17, it is
seen that the symmetric SH mode, S’0, has become more dispersive in its lower frequency region.
Owing to the layer with fibres at 45 the SH waves are always coupled and above about 1.1 MHz
it is difficult to define curves as either SH or Lamb modes, due to changes in the direction of
the displacement vector with frequency. The displacement and stress mode shapes of the S0
and A0 mode are shown in Figure 8.22. Whilst some discontinuities are seen in the in-plane
particle displacements (particularly in the case of A0), reflecting the laminar structure, the out-of
plane displacement is much smoother and closely follows that of its counterpart in the case of
a single-layer isotropic plate. Indeed in general as the number of layers and orientation angles
of a multilayered composite plate are increased, the particle distributions approach those of the
equivalent isotropic plate.

Disperse version 2.0.20a


8.7 Anisotropic Layers 160

Figure 8.22: Displacement and stress mode shapes for the (a) S0 mode and (b) A0 mode at 1
MHz in the example 8-ply quasi-isotropic composite plate.

Disperse version 2.0.20a


8.7 Anisotropic Layers 161

Figure 8.23: (a) Displacement and (b) stress mode shapes for S0 mode propagating at 0 degrees
in an 8-ply 0/90 composite at 1 MHz-mm.

The mode shapes of modes in laminar composites are in general quite different from the equivalent
mode shapes in isotropic material since at the layer boundaries sudden changes or sharp gradients
in the particle displacements and stresses can occur. The shear stress distribution of the S0 mode
in an 8 ply 0/90 (cross-ply) composite, shown in Figure 8.23, is particularly interesting and is
discussed by Guo ([32]). It is seen that the shear stress alternates between maximum and zero at
alternate layer boundaries. The implication for NDE detection of delaminations is that this mode
will not be sensitive to delaminations where the shear stress is zero. For all symmetrical composite
laminates including the quasi-isotropic case, followed here, symmetric modes exhibit zero shear
stress on the mid-plane. The discontinuity of the shear stress at the boundaries of the layers is
due to abrupt changes in the stiffness of adjacent layers. More generally the stress distribution in
laminar composites commonly features localised high-stress regions, which become less prominent
as the number of layers and their angular variation is increased and the plate behaviour approaches
that of an isotropic material.

Disperse version 2.0.20a


Chapter 9

Cylindrical Examples

This chapter discusses some sample cylindrical cases which have been analysed using the wave
propagation model developed in Chapter 7. These cases are included to elaborate on general
principles of cylindrical wave propagation, especially as they apply to leaky systems. In addition,
some of the cases which are included in this chapter have been used to validate the model. The
solution of cylindrical problems is considerably more complicated than the solution of the more
usual flat, isotropic plate structures. The new user is therefore strongly recommended to work
through the cases in the previous Chapter (8) before tackling the examples in this chapter. The
explanations in this chapter will assume a greater degree of expertise of the user than was assumed
in the earlier explanations.

The detailed examples begin with the simple case which Gazis described in 1959 [29]: a single-
layered, elastic, isotropic cylinder. This case is used to illustrate several principles of dispersion
curves. The first parts of the section explains different projections of the dispersion curves that
will be used to display the wave propagation solutions and the system that will be used to label
the curves. The following part discusses the properties of some of the modes in a cylindrical
system. The final part evaluates the conditions when a hollow cylinder may be modelled sufficiently
accurately as a flat plate for computational efficiency.

The later examples gradually become more complicated as surrounding liquids and solids and
additional visco-elastic layers are added. The first such example which discusses the significance
of leaky waves considers a pipe that is filled with and surrounded by water. Different approaches
for treating the fluid filled core are addressed as well as the conditions that are required for waves
to leak into the surrounding fluid. The next example expands this work to analyse cylinders that
are embedded in a solid medium. The final example treats an even more complicated system, a
multi-layered visco-elastic cylinder embedded in a solid. The dispersion curves are very difficult
to trace for this final case, however most of the low attenuation modes have been successfully
identified.

All of the material constants that were used in the modeling can be found in Table 9.1 at the end
of the chapter.

Disperse version 2.0.20a


9.1 Cylinders in Vacuum 163

9.1 Cylinders in Vacuum

This section uses two examples of cylinders in vacuum to discuss some general principles of cylin-
drical wave propagation. The first part of the section examines the dispersion curves for a 2 mm
diameter steel bar to demonstrate some of the different views that can be used to display the
dispersion curves and the system that will be used to name each of the modes. The later parts of
the section use both this bar and a hollow cylinder to describe the behaviour of some commonly
used modes. In addition, the hollow cylinder is used to evaluate when a simpler Cartesian model
may be substituted for the exact cylindrical one.

9.1.1 Tracing the Dispersion Curves

In order to trace the dispersion curves for the bar which is described in this example, take the
following steps:

• Select File : New from the menu.


• Select “Cylinder” as the geometry.

• Select “Structure is in vacuum”.


• Set the “Inside Radius” equal to zero.
• Select “Steel” from the pulldown menu for the first layer.
• Set the thickness of the first layer equal to the radius of the bar.

• Click the “” button to the left of the second layer and select “Delete” from the drop-down
menu which appears.
• Press “OK” to save the geometry.

The procedure for the definition of a hollow empty cylinder is as follows:

• Select File : New from the menu.


• Select “Cylinder” as the geometry.

• Select “Structure is in vacuum”.


• Set the “Inside Radius” equal to the inside radius of the pipe.
• Select “Steel” from the pulldown menu for the second layer (the first layer should show “Free
boundary”).

• Set the thickness of the steel equal to the wall thickness of the pipe.
• Press “OK” to save the geometry.

The procedure for a filled and/or immersed cylinder is similar to that for an empty cylinder, except
that “Structure is embedded” must be selected, then the first layer is used to define the material
(fluid or solid) within the pipe and the final layer is used to define the material surrounding the

Disperse version 2.0.20a


9.1 Cylinders in Vacuum 164

pipe. Mixed combinations may be achieved by choosing a material for one of these extreme layers
and “Free boundary” for the other.

If the radius of a pipe is large in comparison to the wall thickness, then Disperse may give a
warning that the system can alternatively be modelled accurately as a plate. The conditions when
this alternative is valid are discussed in Section 9.1.6.

Once the geometry of the system has been defined, the modes can be traced using the automatic
routine as usual. By default the automatic routine will attempt to trace all of the zero and
first circumferential order solutions up to the frequency that the user specified. Also, the zero
circumferential order torsional mode will be decoupled from the other solutions and will not be
traced; for a simple pipe or bar it is in any case just the shear bulk velocity. The choice of mode
type to be calculated can be modified by adjusting respectively the appropriate options in the
Trace : Solution type and the Trace : Automatic tracing menus. For simple systems, the
automatic routine is usually successful, however, cylindrical systems are less stable to calculate
than Cartesian systems, so more post-processing of the dispersion curves will often be necessary.

Whenever cylindrical systems are being traced, the circumferential order can be specified by
changing the text box in the toolbar, by changing the value in the Sweep dialog, or by selecting
a point on a mode that is of the desired circumferential order.

9.1.2 Projections

As explained in Section 4, the dispersion curves are traced in frequency - real wave number - atten-
uation space. However, we may view these lines of solutions in many different projections. Other
projections, such as phase velocity, group velocity, and angle of incidence, may make information
about the solutions more accessible. Figure 9.1 includes four common projections that are used
for systems that do not have any attenuation. These include real wave number, phase velocity,
group velocity, and angle of incidence. Another commonly used projection, attenuation, will be
used in Section 9.2, which describes a cylinder that leaks energy into a surrounding fluid.

9.1.3 Naming

In order to consistently refer to different modes in cylindrical systems, we use a modified ver-
sion of the system used by Silk and Bainton[93], which tracks the modes by their type, their
circumferential order, and their consecutive order. The labelling assigns each mode to one of three
types,

Longitudinal (L) modes are longitudinal axially symmetric modes. The fundamental longitu-
dinal mode is purely extensional (displacements in the axial direction) at zero frequency.

Torsional (T) modes are modes whose displacement is primarily in the θ̂ direction (These modes
correspond to the ’SH’ modes in a plate.).
Flexural (F) modes are non-axially symmetric bending modes. (The fundamental flexural mode
is a pure bending mode at zero frequency.)

In addition to the type of mode, a dual index system helps track the modes.

Disperse version 2.0.20a


9.1 Cylinders in Vacuum 165

(a) Real wavenumber (b) Phase velocity

(c) Group velocity (d) Angle of incidence

Figure 9.1: Various views of the dispersion curves for a 2 mm diameter steel bar in water: (a) real
wave number, (b) phase velocity, (c) group velocity, and (d) angle of incidence.

The first index refers to the circumferential order of the mode, which describes the integer number
of wavelengths around the circumference of the cylinder. For example, the displacements for zero
order modes are constant with angle but the displacements for first order modes vary in one
sinusoid around the circumference (thus the displacements at one position on the cylinder are 180
degrees out of phase with those at a position 180 degrees around the circumference). Therefore,
all of the longitudinal modes are of circumferential order 0, and all of the flexural modes have a
circumferential order greater than or equal to 1.

The second index is a counter variable. The fundamental modes (those modes that can exist at
zero frequency) are given the value 1 and the higher order modes are numbered consecutively. This
index roughly corresponds to the order of the modes of vibration within the wall of the cylinder.
Using this convention, for example, the third flexural mode of circumferential order 1 would be
named F(1,3).

The conventional names for the modes in a solid bar and in a cylinder are shown in Figures 9.2
and 9.3.

9.1.4 Nature of the Modes in Solid Cylinders

As explained in Section 5.2, the dispersion curves provide a lot of information about the velocities,
wave numbers, and attenuations of the solutions to the guided wave problem. However, this
information does not provide a full description of how the mode behaves. In order to better
understand the nature of the various guided waves and the names that are assigned to them,

Disperse version 2.0.20a


9.1 Cylinders in Vacuum 166

Figure 9.2 shows the mode shapes that are associated with various modes for a 2 mm diameter
steel bar. Figure 9.2(a) shows the behaviour of the fundamental longitudinal and flexural modes
as the frequency changes and (b) shows the mode shapes for some of the higher order modes.

The first longitudinal mode, L(0,1), behaves as a pure extensional mode at zero frequency. The
particle motion is completely in the axial direction and is uniform throughout the cross section
p of
the bar in Figure 9.2. The phase velocity at zero frequency is equal to the bar velocity, E/ρ,
where E is Young’s Modulus and ρ is the density. This expression is easily derived from a static
plane stress analysis. For the steel bar used in this example, the bar velocity is 5229 m/s, which
agrees with the value obtained from the model to six decimal places. As the frequency increases,
the particle motion becomes more complicated. The radial component of the motion grows and
the displacement is no longer constant through the thickness of the bar. As the frequency becomes
very high, the mode begins to act like a Rayleigh wave on the surface of the bar. The mode is
confined to the outer material of the bar to a depth of a few wavelengths and the core of the bar
acts like a semi-infinite half-space.

The fundamental flexural mode, F(1,1), begins as a pure bending, or transverse, mode at zero
frequency. However, as the frequency increases, the mode shapes again become more complicated.
When the frequency becomes large enough the fundamental flexural mode also begins to act like a
Rayleigh wave on the surface of the bar. However, whereas mode L(0,1) leads to a Rayleigh wave
that has the same phase all of the way around the bar, mode F(1,1) leads to a Rayleigh wave on
the top surface that is 180 degrees out of phase with that on the bottom of the bar and that does
not exist on the two sides of the bar.

The mode shapes for the torsional modes have not been shown here; their displacement is primarily
in the θ̂ direction. The fundamental torsional mode corresponds to a uniform twisting of the entire
bar that propagates at a constant velocity equal to the bulk shear velocity in steel.

Figure 9.2(b) displays the mode shapes for some of the higher order longitudinal and flexural
modes. These mode shapes demonstrate that as the second index increases, more minima and
maxima in the shapes appear through the diameter of the bar. These minima correspond to higher
order radial resonances or nodal diameters of the bar. Bar modes that have an even circumferential
order resemble the symmetric modes of a plate, for which the displacements at the top and the
bottom are in phase. On the other hand, modes that have an odd circumferential order resemble
anti-symmetric plate modes, whose top and bottom displacements are out of phase.

9.1.5 Nature of the Modes in Hollow Cylinders

The behaviour of the modes changes when the inside of the cylinder is hollow as shown in Figure
9.3, which shows the phase velocity and group velocity curves for a 1 mm thick steel pipe that
has an inner radius of 2 mm and is in vacuum. This example describes the type of the system
that Gazis described in 1959[29]. The modes in Figure 9.3 have all been labelled according to the
scheme discussed in Section 9.1.3.

As for a bar, the fundamental longitudinal mode, L(0,1), of p a hollow cylinder begins at zero
frequency, with a phase velocity equal to the bar velocity, E/ρ. However, as the frequency
increases, the phase velocity drops quickly and the mode behaves in a similar fashion as the A0
mode in a plate. This behaviour contrasts with the behaviour of the bar described above for
which the L(0,1) mode behaves more like the fundamental symmetric mode of a plate, S0. The
second longitudinal mode, L(0,2), begins at a non-zero frequency
p and an infinite phase velocity
(“cut-off”). The mode quickly drops down to Young’s velocity ( E/(ρ(1 − ν 2 ), where it begins to
act like the S0 mode in a plate. Like A0 and S0, L(0,1) and L(0,2) ultimately tend to the Rayleigh

Disperse version 2.0.20a


9.1 Cylinders in Vacuum 167

(a)

(b)

Figure 9.2: Phase velocity dispersion curves for a 2 mm diameter steel steel bar with mode shapes
super-imposed on the (a) fundamental modes and (b) higher order modes.

Disperse version 2.0.20a


9.1 Cylinders in Vacuum 168

(a) Phase velocity (b) Group velocity

Figure 9.3: Phase velocity and group velocity dispersion curves for an empty 1 mm thick steel
pipe with an inner radius of 2 mm.

velocity at high frequency.

The fundamental flexural mode, F(1,1), begins at zero frequency, phase velocity, and wave number.
At very low frequency, the F(1,1) mode represents a bending mode of the entire pipe that is similar
to the F(1,1) mode of a solid bar that has the same outer radius. The F(1,1) mode in a hollow
cylinder tends to increase in phase velocity and then drop back down and follow a path similar
to L(0,1) in a pipe and A0 in a plate. Although, both the F(1,1) mode and the L(0,1) mode in
a hollow cylinder behave like A0 in a plate, there is a major difference between the modes. The
displacements at the top and bottom of the cylinder are 180 degrees out of phase for the F(1,1)
mode, but they are in phase for the L(0,1) mode. Similar to the longitudinal modes, there is a
F(1,2) mode that behaves like S0 in a plate. This mode, which is not present in the solid cylinder,
differs from the L(0,2) mode because the displacements at the top and bottom of the cylinder are
out of phase.

For a hollow cylinder, the circumferential order has less effect on the high order modes than it
does on the fundamental modes at low frequency. As can be seen in Figure 9.3, the higher order
modes of different circumferential order, such as L(0,4) and F(1,4), nearly overlay each other. As
we will see later, this effect is even more pronounced as the inner radius is increased. For every
mode in an equivalent Cartesian system, there will exist an infinite number of cylindrical modes
corresponding to modes of different circumferential order. As the cylindrical system becomes more
and more like a plate, the infinite number of cylindrical modes all overlap the dispersion curves
for a plate. This trend is discussed more in Section 9.1.6

The fundamental torsional mode of circumferential order zero propagates at a constant phase
velocity that is equal to the bulk velocity of the shear wave in the material. However, the higher
order torsional modes show dispersion. The higher order torsional modes begin at a regular
frequency spacing corresponding to half multiples of the shear speed. Their behaviour is simpler
than the longitudinal and flexural modes and the zero order torsional modes can be expressed
analytically in a similar fashion as the ’SH’ modes in a plate.

9.1.6 Effect of Changing the Radius

Once we understand the general behaviour of the modes of a single layered hollow cylinder, we can
explore the effect of changing some of the parameters. One interesting parameter to investigate
is the effect of changing the ratio of the inner radius to the wall thickness. As the inner radius

Disperse version 2.0.20a


9.2 Cylinders Immersed in a Fluid 169

becomes infinitely large, the cylinder should act like a plate of the same thickness. The transition
of behaviour from a system that is dominated by its cylindrical nature and a system that behaves
like a plate provides an interesting subject to explore. This transition helps explain the behaviour
of cylindrical waves and reveals when it is acceptable to model a cylinder as a plate to simplify
the analysis.

As the ratio of the inner radius of a pipe to the wall thickness of the pipe increases the pipe begins
to behave more like a plate. The trend of the dispersion curves as the radius changes from one
wall thickness to 20 wall thicknesses is shown for the zero and first circumferential order modes in
Figure 9.4. The example models a one millimetre thick steel pipe that is surrounded by vacuum.
For comparison, the dispersion curves for an equivalent one millimetre thick plate, which can be
considered to be a pipe with an infinitely large radius, are shown in part (f). The effect of the
curvature of the pipe is most noticeable at low frequencies, which correspond to long wavelengths.
As the frequency increases and the wavelength decreases, the wave only ’sees’ a local section of
the pipe, which begins to appear as a plate, especially as the radius becomes larger. When the
radius is 10 wall thicknesses, the difference between the pipe and the plate dispersion curves is
only detectable below 0.5 MHz-mm frequency-wall thickness. Above that frequency-thickness, the
cylindrical curves nearly overlay those for a plate, and the curves remain the same regardless of
the circumferential order.

From this example, we can conclude that Cartesian dispersion curves can be used when working
with high frequencies and large radius pipes to improve the speed of calculation and the stability.
A rule of thumb derived from experience of tracing curves indicates that the minimum frequency
for which an equivalent plate case could be substituted is,
vlong
fmin = (9.1)
rinner
where fmin is the minimum frequency in MHz, rinner is the inner radius in mm, vlong is the bulk
longitudinal velocity in mm/µs. Therefore, if the target application will only use frequencies -
wall thicknesses above 0.25 MHz-mm, a plate case can only be substituted if the ratio of radius
to the wall thickness is greater than about 20.

9.2 Cylinders Immersed in a Fluid

The previous cases discuss an elastic cylinder in vacuum, which is a relatively simple case to solve.
When the cylinder is embedded in another material, calculating the dispersion curves becomes
much more complex. When an elastic system is in vacuum, there is no mechanism for energy to
be lost, so the attenuation is always zero. However, when the cylinder is surrounded by other
materials, energy can leak from the structure into the surrounding liquid or solid, where it is free
to propagate away (radiate)in the form of bulk waves which travel towards the infinite boundaries.
Therefore, the solutions of the dispersions are in three variables: frequency, real wavenumber, and
attenuation. This extends the complexity of the search operations in comparison with the two
parameter searches for a lossless problem.

9.2.1 Modelling the Fluid Inside the Cylinder

Figure 9.5 displays the dispersion curves for the zero order circumferential modes of a 1 mm thick
steel pipe with an inner radius of 10 mm that is surrounded by and filled with water. The plots
on the left show the phase velocity dispersion curves, as are shown in Figure 9.4. The plots on

Disperse version 2.0.20a


9.2 Cylinders Immersed in a Fluid 170

(a) Inner radius = 1 mm (b) Inner radius = 2 mm

(c) Inner radius = 5 mm (d) Inner radius = 10 mm

(e) Inner radius = 20 mm (f) Plate (inner radius = ∞)

Figure 9.4: The effect of increasing the inner radius of a 1 mm thick steel pipe. As the inner
radius increases, the dispersion curves begin to very closely match the dispersion curves for a plate
(shown in section f), except for very low frequencies.

Disperse version 2.0.20a


9.2 Cylinders Immersed in a Fluid 171

the right display the associated attenuation values. The three sets of dispersion curves represent
subtly different cases. The first two cases model the same pipe. However, in the second case, the
solution of the waves inside the pipe is modified so that there is a sink which absorbs any waves
that propagate at the origin. The significance of this modification is discussed below. The third
case models the pipe as a plate and is shown for comparison.

The top set of dispersion curves in Figure 9.5 models the fluid filled steel pipe as it actually exists.
The dispersion curves contain fluid modes, whose energy and motion is concentrated in the water.
These modes appear as the tight, regularly spaced curves. They tend to have low attenuation
values and travel more slowly than the bulk velocity of water. When the phase velocity and
frequency of the fluid modes correspond with a point on the dispersion curves for the unloaded
steel pipe, the energy of the fluid modes couples into the steel where a fast wave may propagate.
However, at these points, there will also be significant attenuation because the energy of the wave,
which is principally located in the steel, can easily leak into the surrounding water on the outside
of the pipe where it is lost. Figure 9.6 displays the axial displacement for two points on the
same mode, when it is behaving as fluid mode (Figure 9.6(a)) and when it is behaving as pipe
mode (Figure 9.6(b)). The top portion of the graph represents the centre of the system. The
displacement becomes very large towards the centre as the waves that are leaking from the steel
pipe into the centre of the pipe are focused. The middle section of the graph represents the wall of
the steel pipe. When the mode is behaving as a fluid mode, there is very little displacement in the
steel. However, there is a considerable amount of axial displacement when the mode deviates from
its path and momentarily behaves like a pipe mode. The bottom section of the graph represents
the first 10 mm of the surrounding fluid. The wave that is seen in this section in Figure 9.6(b)
corresponds to the leaky wave that causes attenuation.

The middle set of dispersion curves in Figure 9.5 represents the same system as the top set of
dispersion curves except that a sink has been placed at the axis of the system by representing the
waves in the internal fluid by incoming Hankel functions instead of by standing Bessel functions.
This sink absorbs the energy that propagates inwardly from the steel pipe and prevents the fluid
modes from forming. The attenuation values are much larger for this case as opposed to the
previous case because energy is lost in both the sink and in the external fluid. The sink is
implemented by choosing Trace : Solution type and then ticking the box Place sink at
origin.

Strictly speaking, for a perfect fluid filled pipe, this substitution does not represent the actual
system and is not valid. However, there is valuable information in the second set of dispersion
curves that is lost in the more strict first set. Some ultrasonic tests are designed to propagate
energy down the fluid inside a pipe and these tests must use the first set of dispersion curves.
However, the majority of ultrasonic testing is performed by sending waves down the walls of the
pipe. This choice is usually made because the testing procedure is searching for defects in the
pipe or because it is difficult to gain access to the inside of the pipe. For such studies it is useful
to be able to plot the dispersion curves for the pipe without cluttering the graphs with all of the
fluid modes. Furthermore, real filled pipes may often contain absorbing or scattering materials,
or may only be partially filled, in which case the radiated waves are not properly reabsorbed by
the opposite wall of the pipe; in such cases the sink model would be the more realistic choice.

The third set of dispersion curves in Figure 9.5 displays the phase velocity and attenuation curves
for a 1 mm steel plate in water. Comparison between the last two sets of curves shows that the
behaviour of a large radius pipe at high frequency is very similar to that of a plate, confirming
that the similarity seen in Figure 9.4 also holds for leaky systems.

Disperse version 2.0.20a


9.2 Cylinders Immersed in a Fluid 172

(a) 1 mm steel pipe (10 mm radius) with fluid modes shown

(b) 1 mm steel pipe (10 mm radius) with sink in centre

(c) 1 mm steel plate in water

Figure 9.5: A comparison of phase velocity and attenuation dispersion curves for three cases: a
fluid filled pipe, a fluid filled pipe with a sink in the centre, and a plate in water.

Disperse version 2.0.20a


9.2 Cylinders Immersed in a Fluid 173

(a) Fluid mode (b) Pipe mode

Figure 9.6: A comparison of the axial displacement for a fluid filled pipe for a mode that is (a)
predominantly in the fluid and (b) predominantly in the pipe wall. The top of the graph represents
the fluid filled core, the middle section the steel wall of the pipe, and the bottom section the first
10 mm of the surrounding fluid

9.2.2 Leakage into the Surrounding Medium

Much of the attenuation in Figure 9.5 can be attributed to the amount of radial displacement on
the outer surface of the cylinder. The radial displacement permits the energy to be coupled from
the guided wave into the surrounding liquid. Figure 9.7 demonstrates this effect for a 1 mm radius
bar immersed in water for the fundamental longitudinal and flexural modes. The top graph shows
the phase velocity curves for these two modes and the bottom graphs show the related attenuation
values. Four grid mode shapes display the nature of the mode at the circled points. The centre
section of the grid mode shapes represents a cross section through the centre of the bar and the
outer sections represent the first two millimetres of the infinitely large surrounding fluid.

The fundamental longitudinal bar mode, L(0,1), is very lightly attenuated at low frequencies since
the majority of the motion is in the axial and not the radial direction. However, as the frequency
increases, the attenuation also increases dramatically as the radial displacement at the surface
increases. As the frequency continues to increase, the attenuation approaches the attenuation
value for a leaky Rayleigh wave, which is linear with frequency.

The fundamental flexural mode has significant radial displacement components at the surface of
the bar across a wide frequency range. Correspondingly, the attenuation values remain relatively
high. However, when the phase velocity of the F(1,1) mode drops below the bulk velocity of water,
the attenuation becomes zero although there is still significant radial displacement.

This ’non-leaky’ mode can be explained by Figure 9.8. As the guided wave travels down the bar,
the radial displacements at the surface of the bar couple energy into the surrounding fluid. This
motion will create a bulk wave in the fluid, which then carries energy away from the system. The
wavelength of a bulk wave in the fluid is known, since bulk waves only travel at one speed and the
frequency is known. The wavelength of the guided wave is calculated from the dispersion curves.
Since the projection of the wave along the interface must match for the guided wave and the bulk
wave, the leakage angle, θleak , can be determined to be,

Disperse version 2.0.20a


9.2 Cylinders Immersed in a Fluid 174

Figure 9.7: Phase velocity dispersion curves for a 2 mm diameter steel steel bar immersed in
water with mode shapes super-imposed on the fundamental modes.

Disperse version 2.0.20a


9.3 Cylinders Embedded in a Solid 175

Figure 9.8: As a guided wave travels down a bar, it couples energy into the surrounding fluid,
creating a leaky bulk wave at a characteristic angle.

 
−1 λwater
θleak = sin (9.2)
λsteel
 
−1 v water /f
= sin
vph /f
 
−1 v water
= sin
vph

where λ is the wavelength, vwater is the bulk velocity of the water, vph is the phase velocity of the
guided wave, and f is the frequency. If the phase velocity, vph , is less than the bulk velocity in the
surrounding medium, then equation 9.2 implies that the leakage angle, θleak would be imaginary.
Physically, the imaginary angle means that instead of creating a wave that propagates away from
the bar, the displacements die away exponentially and the energy is trapped in the region of fluid
adjacent to the bar. Therefore, in general, a mode will not leak if its phase velocity is less than
the bulk velocity of any waves that can exist in the surrounding liquid.

9.3 Cylinders Embedded in a Solid

When the system is surrounded by a liquid, the energy leaks via a bulk compressional wave.
However, when the system is surrounded by a solid material, energy can be carried away from
the system by both shear and compressional (longitudinal) bulk waves. In addition, since the

Disperse version 2.0.20a


9.3 Cylinders Embedded in a Solid 176

Figure 9.9: Phase velocity and attenuation dispersion curves for an empty 1 mm thick steel pipe
with a 10 mm inner radius that is surrounded by soft stone (mudstone).

tangential displacements and stresses couple into the surrounding solid as well as the normal ones,
the attenuation values tend to be much higher and the minima in attenuation do not usually
reach down to zero attenuation. An example of a system that is surrounded by a solid is given in
Figure 9.9. The system represents an empty 1 mm steel pipe with a 10 mm inner radius that is
surrounded by a soft stone (mudstone).

Comparing the attenuation curves with those for a pipe in water in Figure 9.5 reveals some
dramatic differences.

First, the attenuation values are much larger, so much larger than the liquid loaded case that
different scales had to be used.

Second, many more modes are present on the attenuation curves for the embedded pipe than for
the immersed pipe. The torsional modes in an immersed pipe are not attenuated because the
theta displacement at the surface of the pipe does not couple into the surrounding fluid. However,
when the pipe is surrounded by a solid, the theta displacement does couple into the surrounding
medium, a bulk wave radiates, and attenuation is present. In addition, the pairs of modes at low
frequency represent the zero and first circumferential order solutions, although only the zero order
solutions are shown for the immersed pipe.

Third, the F(1,1) and L(0,1) modes in the embedded pipe are not continuous when they cross the
bulk longitudinal velocity of the surrounding mudstone, 2.6 mm/µs. At this point, there is a jump
in attenuation and frequency. The cause of this jump is that at a phase velocity above the bulk
longitudinal velocity of the surrounding solid, the pipe is leaking both shear and longitudinal waves.
However, below the bulk longitudinal velocity, the pipe is only leaking shear waves, which leads to
the discontinuity. The two portions of the curve may in fact be linked mathematically, however,
the linking section of the curve would be very dispersive and unsuitable for non-destructive testing,
so effort has not been expended to find these links. In addition, there is a gap in the L(0,2) and
the F(1,2) modes around 3 MHz. In reality these modes should continue, however, numerical
instabilities around the shear bulk velocity in steel do not allow them to be traced any farther at
present.

9.3.1 Different Types of Leakage

The different types of mode leakage for a cylinder embedded in a solid are shown in Figures 9.10
and 9.11. Figure 9.10 shows the phase velocity dispersion curves for a 1 mm radius steel bar

Disperse version 2.0.20a


9.3 Cylinders Embedded in a Solid 177

Figure 9.10: Phase velocity and attenuation dispersion curves for a 1 mm radius steel bar embedded
in a stiff material. The arrows and labels correspond to the positions for which the mode shapes
are given in the next figure. The straight lines correspond to the bulk velocities in the surrounding
solid and divide the dispersion curves into different leaky regimes.

(a) (b) (c)

Figure 9.11: Mode shapes for a steel bar embedded in a stiff material in (a) a non-leaky region,
(b) a region where only shear waves leak, and (c) a region where both shear and longitudinal bulk
waves leak into the surrounding solid.

embedded in a light stiff material. The existence of non-leaky portions of modes that have no
attenuation makes a significant difference between these curves and those in Figure 9.9. Two
straight lines are shown on the phase velocity diagram which correspond to the longitudinal and
shear bulk velocities in the surrounding rock. These bulk velocities divide the section of the
modes that are non-leaky, leak only shear, and leak both shear and longitudinal waves. The
points marked (a), (b), and (c) correspond to the points for which the mode shapes are shown
in Figure 9.11. Point (a) is in the non-leaky region, below both the shear and longitudinal bulk
velocities. Point (b) is the region where only bulk shear waves can leak and point (c) is in the
region above the longitudinal bulk velocity where both shear and longitudinal modes leak into the
surrounding solid. The mode shapes in Figure 9.11 demonstrate this effect. The centre part of
the grid represents the cross section of the bar and the outer section of the grid represents the
first two millimetres of the surrounding rock. In the non-leaky section shown in Figure 9.11(a),
the displacement in the surrounding solid dies away exponentially as described in the previous
section. In Figure 9.11(b) only a shear wave leaks from the bar into the surrounding solid. The
displacement field in the surrounding solid shows a single leaky wave. However, the displacement
field is much more complicated for Figure 9.11(c) because the leaky shear and longitudinal bulk
waves overlap at different leakage angles.

9.3.2 Multi-layered, Visco-elastic, Pipes, Embedded in a Solid

The final example in this chapter takes advantage of the visco-elastic and multi-layered nature of
this solution to the wave propagation problem. The system is the same as the example shown in
Figure 9.9, however a 0.5 mm layer of grout has been added to the inside of the pipe to simulate

Disperse version 2.0.20a


9.3 Cylinders Embedded in a Solid 178

Figure 9.12: Phase velocity and attenuation dispersion curves for a 1 mm thick steel pipe with
a 10 mm inner radius that is lined with visco-elastic grout and is surrounded by visco-elastic soft
stone (mudstone).

Table 9.1: Material constants used in the cylindrical wave propagation examples.
Material Density vl (m/s) vs (m/s) Young’s Poisson’s αl αs
(g/cm3 ) Modulus Ratio (np/λ) (np/λ)
(GPa)
Steel 7.932 5960 3260 217 0.29 0.0 0.0
Perspex 1.18 2730 1430 6.327 0.31 0.0 0.0
Water 1.0 1483 - - - 0.0 -
Mudstone 2.5 2592 1497 14 0.25 0.03 0.08
Grout 1.6 2810 1700 11.2 0.21 0.043 0.1
Stiff 2.2 5720 3300 60.0 0.25 0.0 0.0
Stone

a scaled down version of a mains water pipe. In addition, both the grout and the mudstone
are considered to be visco-elastic materials to account for their inherently lossy nature. The
attenuation due to the visco-elastic damping is super-imposed on the attenuation due to leakage.
The dispersion curves are shown in Figure 9.12.

It is immediately obvious in Figure 9.12 that many of the modes are incomplete. The complexity
of the system and the high losses associated with it render many of the regions insoluble. Many
of the instabilities have been improved by using appropriate scaling factors in the equations,
however, these instabilities have not been completely conquered and still cause problems as seen
in this example. Another obvious feature of Figure 9.12 is that the modes are shifted to the
left, compared to the previous dispersion curves. Since this system is thicker than the previous
one because the grout lining adds to the steel wall, the frequencies at which through thickness
resonances occur decrease.

Modes L(0,1) and F(1,1) show interesting behaviour. At low frequencies, they travel at a phase
velocity similar to the shear bulk velocity in the steel. However, as the frequency increases, their
phase velocity lowers to the bulk velocity of the grout liner. This behaviour is consistent with
the practice that the longer wavelengths associated with low frequencies ’look’ farther into the
system and are predominantly influenced by the steel layer while the short wavelengths and high
frequencies are predominantly affected by the grout lining.

Disperse version 2.0.20a


Chapter 10

Additional Information

10.1 File Structure

When Disperse is installed, a tree structure with three branches is created in the installation
directory, as follows:

• The Bin subdirectory. This directory contains all of the files for the operation of the program.
The main file is the executable Disperse.exe. However Disperse has been written in a modular
structure, such that each of the main functional areas is coded in its own dynamic linked
library (.DLL) and is called when needed by the program. Thus this directory contains a
number of .DLL files. Prior to version 2.0.20a, the materials database file materals.lst was
also kept here. From version 2.0.20a onwards, the materials database file materals.lst is kept
in the directory c:/programdata/dispersedata, see paragraph below. The materials database
file is an ASCII file which stores all of the material property information which has been
saved; initially this contains just the default material properties which are shipped with the
program, but any which are subsequently stored or deleted by the user are added or removed,
see Section 3.3. Since Disperse offers the functions to add or delete materials, it is strongly
recommended that these be used rather then manually editing the file.
• The Examples subdirectory. This directory contains all of the example files which are in-
stalled with Disperse and which are listed in Chapter B. The example files are simply files
which have been saved in the usual way after running Disperse. Thus the user may extend
this library by solving other cases, then saving the results in this directory (or save them
elsewhere if preferred).
• The Help subdirectory. This is where this help file is stored.

From version 2.0.20a onwards, Disperse stores the materials database file materals.lst by default
in the directory c:/programdata/dispersedata. This is to comply with MS Windows security pref-
erences. However, you can change the location for storing this file if you wish. To do so, edit the
CSV text file Dispersesettings.csv which you will find in the Bin subdirectory, and replace the text
string c:/programdata/dispersedata by your preferred path name.

Apart from this exception, Disperse does not use a configuration or settings file, but instead stores
information such as selected preferences in the registry of the computer; this is then read in again

Disperse version 2.0.20a


10.2 How to Get in Contact 180

at the start of future sessions. The information stored in this way is specific to each user of the
computer.

10.2 How to Get in Contact

The primary authors for this project are Mike Lowe and Brian Pavlakovic. We have also been
helped in development and testing by numerous members past and present of the Imperial College
NDT Lab, including David Alleyne, Jon Allin, Christophe Aristegui, Stephane Baly, Malcolm
Beard, Arnaud Bernard, Peter Cawley, Roger Dalton, Pierre Marty, Tom Pialucha, Tom Vogt,
Paul Wilcox.

Your comments and suggestions are always welcome and encouraged. All comments and question
should be addressed to

Mike Lowe
Department of Mechanical Engineering
Imperial College
London SW7 2AZ, UK
email: m.lowe@imperial.ac.uk Web page: www.imperial.ac.uk/nde

10.3 This User Manual

This manual has been written and is maintained using the Latex document preparation system
under the Windows 95/98/NT operating system. The source editor is WinEdt 1.414, the compiler
is pdfTeX 3.14159-14d with MiKTeX 1.20e. The figures are loaded as graphical objects which are
themselves .pdf format. The book marks and cross referencing use the hyperref package.

Disperse version 2.0.20a


Appendix A

Summary of Useful Relations

A.1 Relations between Lambda, Rho and Mu

For elastic (lossless) materials the bulk longitudinal velocity, CL , can be written in terms of the
Lamé constants, λ and µ, or Young’s Modulus, E, and Poisson’s ratio, ν,

  21  1/2
λ + 2µ E(1 − ν)
CL = = (A.1)
ρ ρ(1 + ν)(1 − 2ν)

And the bulk shear wave velocity, CS , can be written as,

  12
µ
CS = (A.2)
ρ

Young’s Modulus, E, and Poisson’s ratio, ν, can be expressed as combinations of the Lamé con-
stants, λ and µ,

3CL2 − 4CS2
   
3λ + 2µ
E=µ = ρCS2 (A.3)
λ+µ CL2 − CS2

  2
CL − 2CS2

λ 1
ν= = (A.4)
2(λ + µ) 2 CL2 − CS2

BulkM od = λ + 2/3µ (A.5)

A.2 Wave Speeds

Here follows some basic relations of wavenumber and speed, and some formulae for their values:

Disperse version 2.0.20a


A.3 Constitutive and Compatibility Relations for Elastic Isotropic Material 182


W avelength = (A.6)
|k|

ω
P hase speed = C = (A.7)
|k|

p
Ccircular−bar = E/ρ (A.8)

s
E
CS0ω=0 = (A.9)
(1 − ν 2 )ρ

 
0.87 + 1.12ν
CRayleigh ≈ CS (A.10)
1+ν

The angles of incidence, transmission and reflection of homogeneous bulk waves (L=longitudinal,
S=shear) in the layers are related to the wavenumber, frequency and phase velocity by:

|k| 1 sin(θL ) sin(θS )


= = = (A.11)
ω Cph CL CS

The Rayleigh wave phase velocity (CRayleigh ) can be determined by finding the roots of the
following polynomial, where T = (CR ayleigh/CS )2 and R = (CS /CL )2 :

T 3 − 8T 2 + (24 − 16R)T − 16(1 − R) = 0 (A.12)

A.3 Constitutive and Compatibility Relations for Elastic


Isotropic Material

The stress-strain relationship is:

 
    xx
σxx λ + 2µ λ λ  

yy

 
 
 

σyy λ λ + 2µ λ

 
   
 


 
   

σzz λ λ λ + 2µ
     
=  zz (A.13)
σ xy
 µ 
xy

 
   

σ µ
   
yz
    
yz

 
 
 

σxz µ 
   

xz
 

In Cartesian coordinates the strains are related to the displacements by:


∂u1 ∂u2 ∂u3
11 = , 22 = , 33 =
∂x1 ∂x2 ∂x3
∂u1 ∂u2 ∂u2 ∂u3 ∂u1 ∂u3
12 = + , 23 = + , 13 = + (A.14)
∂x2 ∂x1 ∂x3 ∂x2 ∂x3 ∂x1

Disperse version 2.0.20a


A.4 Relations between Velocity and Attenuation Constants for Isotropic Materials 183

In cylindrical coordinates the strains are related to the displacements by:

∂ur 1 ∂uθ ur ∂uz


rr = , θθ = + , zz =
∂r
 r ∂θ r ∂z
1 ∂ur ∂uz
rz = +
2 ∂z ∂r
 
1 ∂ uθ  1 ∂ur

rθ = r +
2 ∂r r r ∂θ
 
1 ∂uθ 1 ∂uz
θz = + (A.15)
2 ∂z r ∂θ

A.4 Relations between Velocity and Attenuation Constants


for Isotropic Materials

 12  12
λ + 2µ − i (λ0 + 2µ0 ) µ − i (µ0 )
 
CL = , CS = (A.16)
ρ ρ

The constants CL and CS are real and are exactly equal to the bulk velocities of longitudinal and
shear waves respectively if the damping parameters (primed terms) are zero. But for attenuating
waves they are complex and are related to the phase velocity and attenuation by:

ω Cph Cph
CL or CS = = = (A.17)
|kre | + i|kim | 1 + |kim |
|kre |
1 + iα/(2π)

where Cph is the phase velocity of the wave (which is always real) and α is the attenuation in
Nepers per wavelength, such that a wave of unit amplitude is reduced to an amplitude of e−α
after travelling one wavelength.

From this:
CL(im) CS(im)
Cph,L = CL(re) − αL , Cph,S = CS(re) − αS
2π 2π
CL(im) CS(im)
αL = −2π , αS = −2π (A.18)
CL(re) CS(re)

and:

2   2 
ρ (Cph,S ) 4π 2 4π 2 − αS2 0 ρ (Cph,S ) 4π 2 (4αS π)
µ= 2 , µ = 2
(αS2 + 4π 2 ) (αS2 + 4π 2 )
2   2 
ρ (Cph,L ) 4π 2 4π 2 − αL 2
ρ (Cph,L ) 4π 2 (4αL π)
λ= 2 − 2µ , λ0 = 2 − 2µ0 (A.19)
2 + 4π 2 )
(αL 2 + 4π 2 )
(αL

Disperse version 2.0.20a


A.5 Engineering and Cij Constants for Orthotropic Materials 184

A.5 Engineering and Cij Constants for Orthotropic Mate-


rials

The stiffness of an orthotropic material is often defined using a Cij matrix. The Cij matrix relates
the stress vector σ to the strain vector  according to the following relationship:

    
σxx C11 C12 C13 0 0 0 xx

 σyy  
  C21 C22 C23 0 0 0 
 yy 

 σzz  
= C31 C32 C33 0 0 0  zz 
   (A.20)

 σyz  
  0 0 0 C44 0 0 
 yz 

 σzx   0 0 0 0 C55 0  zx 
σxy 0 0 0 0 0 c66 xy

The Cij matrix for an orthotropic material is always symmetric, so that:


C12 = C21 ; C13 = C31 ; C23 = C32 (A.21)

Thus the Cij matrix for an orthotropic material has 9 independent constants.

A transversely isotropic material has 5 independent constants and is a special case of an orthotropic
material. The Cij matrix for a transversely isotropic material with symmetry about the X axis is:

    
σxx C11 C12 = C12 0 0 0 xx

 σyy  
  = C12 C22 C23 0 0 0 
 yy 

 σzz  
= = C12 = C23 = C22 0 0 0  zz 
   (A.22)

 σyz  
  0 0 0 = 12 (C22 − C23 ) 0 0 
 yz 

 σzx   0 0 0 0 C55 0  zx 
σxy 0 0 0 0 0 = C55 xy

A cubic material has 3 independent constants and is also a special case of an orthotropic material.
The Cij matrix for a cubic material is:

    
σxx C11 C12 = C12 0 0 0 xx

 σyy  
  = C12 = C11 = C12 0 0 0 
 yy 

 σzz  
= = C12 = C12 = C11 0 0 0  zz 
   (A.23)

 σyz  
  0 0 0 C44 0 0 
 yz 

 σzx   0 0 0 0 = C44 0  zx 
σxy 0 0 0 0 0 = C44 xy

Finally, an isotropic material has 2 independent constants and is the simplest special case of an
orthotropic material. The Cij matrix then reduces to:

    
σxx C11 C12 = C12 0 0 0 xx

 σyy  
  = C12 = C11 = C12 0 0 0 
 yy 

 σzz  
= = C12 = C12 = C11 0 0 0  zz 
   (A.24)

 σyz  
  0 0 0 = 12 (C11 − C12 ) 0 0 
 yz 

 σzx   0 0 0 0 = C44 0  zx 
σxy 0 0 0 0 0 = C44 xy

Disperse version 2.0.20a


A.5 Engineering and Cij Constants for Orthotropic Materials 185

An alternative way of expressing the stiffness information is to use the Engineering constants
which in fact define the properties according to a flexibility matrix. The coefficients are composed
of Eii (Young’s modulus), Gij (shear modulus) and νij (Poisson’s ratio) terms. The matrix of
engineering constants for an orthotropic material is:

1 −ν21 −ν31
 
0 0 0
  
xx E11 E22 E33 σxx
−ν12 1 −ν32
 yy   E11 E22 E33 0 0 0 
σyy 
   −ν13 −ν23 1
 
 zz  
= E11 E22 E33 0 0 0 
σzz 
(A.25)
  
1
yz   0 0 0 0 0 σyz
  
   G23  
zx 1 σzx
  0 0 0 0 0
  
 G13 
xy 1 σxy
0 0 0 0 0 G12

and symmetry limits the matrix again to 9 independent coefficients, requiring that:
ν12 ν21 ν13 ν31 ν23 ν32
= ; = ; = (A.26)
E11 E22 E11 E33 E22 E33

The 9 Engineering constants (Eii , Gii and νii ) for an orthotropic material are related to the Cij
constants as follows:
A A A
E11 = 2 ; E22 = 2 ; E33 = 2
C23 − C22 C33 C13 − C11 C33 C12 − C11 C22

G23 = C44 ; G31 = C55 ; G12 = C66

C13 C23 − C12 C33 C12 C23 − C13 C22


ν12 = 2 −C C ; ν13 = 2 −C C ;
C23 22 33 C23 22 33

C12 C13 − C11 C23


ν23 = 2 −C C (A.27)
C13 11 33

in which
2 2 2
A = C11 C23 + C22 C13 + C33 C12 − C11 C22 C33 − 2C12 C13 C23 (A.28)

The inverse of this relationship expresses the Cij constants in terms of the engineering constants
([45]):
1 − ν23 ν32 1 − ν13 ν31 1 − ν12 ν21
C11 = ; C22 = ; C33 =
E22 E33 B E11 E33 B E11 E22 B

C44 = G23 ; C55 = G31 ; C66 = G12

ν21 + ν31 ν23 ν31 + ν21 ν32


C12 = C21 = ; C13 = C31 = ;
E22 E33 B E22 E33 B
ν32 + ν12 ν31
C23 = C32 = (A.29)
E11 E33 B
in which
1 − ν12 ν21 − ν23 ν32 − ν31 ν13 − 2ν21 ν32 ν13
B= (A.30)
E11 E22 E33

Disperse version 2.0.20a


A.6 Attenuation of waves in materials with complex Cij constants 186

A.6 Attenuation of waves in materials with complex Cij


constants

The model which Disperse uses for damping of anisotropic materials mirrors that for isotropic
materials: that is, the fundamental parameter (the imaginary part of each Cij coefficient) defines
an attenuation of a bulk wave which is a constant loss per wavelength.

The attenuation of a bulk wave travelling along an axis of symmetry (for example direction n) of
an anisotropic material is parallel to the propagation direction and is given (approximately, for
Cnn(re)  Cnn(im) ), in Nepers per wavelength, as:

Cnn(im)
αn = −π (A.31)
Cnn(re)

This attenuation may be expressed in Nepers per meter at a particular frequency, f (Hz), by [12]:

Cnn(im) πf
αn,N p/m = − √ Cnn(re) (A.32)
Cnn(re) ρ

Disperse uses the following conversions from a set of isotropic material data (CL , CS , ρ) to an
orthotropic lossy Cij matrix. Again, these are approximate and valid only for small values of
attenuation.
2
C11(re) = ρCphase,L

2
C55(re) = ρCphase,S

C13(re) = C11(re) − 2C55(re)

!2
Cphase,L
C11(im) = Im ρ
1 + iαL

!2
Cphase,S
C55(im) = Im ρ
1 + iαS

C13(im) = C11(im) − 2C55(im) (A.33)

Conversion from Cij constants to isotropic constants is of course only possible in the special case
when the Cij constants define an isotropic material. However, in other cases the conversion may
be useful simply to provide a starting set of data for the isotropic analysis. Regardless of whether
or not the Cij material is isotropic, Disperse uses just the two complex coefficients ∗ C11 and ∗ C55 ,
and the density. The phase velocities are calculated by:
1/2
Re(∗ C11 )

Cphase,L =
ρ
1/2
Re(∗ C55 )

Cphase,S = (A.34)
ρ

Disperse version 2.0.20a


A.7 Bessel Function Recurrence Relations 187

in which the * denotes a complex quantity. Thus, for example, ∗ C11 = C11(re) + iC11(im)

The complex velocities are:


1/2
(∗ C11 )


CL =
ρ
 ∗ 1/2
∗ ( C55 )
CS = (A.35)
ρ

and these are used to calculate the attenuation constants (Nepers per wavelength):

Cphase,L
αL = 2π ∗C
L
Cphase,S
αS = 2π ∗ (A.36)
CS

A.7 Bessel Function Recurrence Relations

∂ h ∂2 h2
(Jn (hr)) = (Jn−1 (hr) − Jn+1 (hr)) 2 (Jn (hr)) = (Jn−2 (hr) − 2Jn (hr) + Jn+2 (hr))
∂r 2 ∂r 4
(A.37)

and,

2n
Jn+1 (hr) = Jn (hr) − Jn−1 (hr) (A.38)
hr

so, we can also specify that,

2(n − 1)
Jn−2 (hr) = Jn−1 (hr) − Jn (hr) (A.39)
hr

and inversely, that,

2(n + 1)
Jn+2 (hr) = Jn+1 (hr) − Jn (hr) (A.40)
hr

therefore, making the proper substitutions,

∂ n −n
(Jn (hr)) = Jn (hr) − hJn+1 (hr) = (Jn (hr)) + hJn−1 (hr) (A.41)
∂r r r

and,

∂2
 
n(n − 1) 2 h
(Jn (hr)) = − h Jn (hr) + Jn+1 (hr) (A.42)
∂r2 r2 r

Disperse version 2.0.20a


A.8 Operations in Cylindrical Coordinates 188

A.8 Operations in Cylindrical Coordinates

Vector derivatives:
∂Φ 1 ∂Φ ∂Φ
∇Φ = r̂ + θ̂ + ẑ (A.43)
∂r r ∂θ ∂z

1 ∂ 1 ∂Hθ ∂Hz
∇·H= (rHr ) + + (A.44)
r ∂r r ∂θ ∂z

     
1 ∂Hz ∂Hθ ∂Hr ∂Hz 1 ∂(rHθ ) 1 ∂Hr
∇ × H = r̂ − + θ̂ − + ẑ − (A.45)
r ∂θ ∂z ∂z ∂r r ∂r r ∂θ

1 ∂2Φ ∂2Φ
 
1 ∂ ∂Φ
∇2 Φ = r + + (A.46)
r ∂r ∂r r2 ∂θ2 ∂z 2

∇2 H = ∇(∇ · H) − ∇ × ∇ × H
 2
1 ∂ 2 Hr ∂ 2 Hr

∂ Hr 1 ∂Hr 2 ∂Hθ 1
= r̂ + + + − − Hr
∂r2 r ∂r r2 ∂θ2 ∂z 2 r2 ∂θ r2
 2
1 ∂ 2 Hθ ∂ 2 Hθ

∂ Hθ 1 ∂Hθ 2 ∂Hr 1
+ θ̂ + + 2 + + 2 − 2 Hθ
∂r2 r ∂r r ∂θ2 ∂z 2 r ∂θ r
 2
1 ∂ 2 Hz ∂ 2 Hz

∂ Hz 1 ∂Hz
+ ẑ 2
+ + 2 2
+ (A.47)
∂r r ∂r r ∂θ ∂z 2

Disperse version 2.0.20a


Appendix B

List of Examples

The following chapter gives a brief summary of each of the example files that are included with
the distribution of Disperse. Most of these examples are further explained in Chapters 8 and 9.

Note that Disperse version 2.0 files use the file extensions .ds2 which differs from the Disperse
version 1.0 files .dsp. This is to take account of the improvements to the file format in order to
store additional information such as user notes etc. Disperse version 2.0 can open both the new
.ds2 and existing .dsp files, but all files are now saved as .ds2. The new format .ds2 cannot be
opened by version 1.0 of Disperse.

plate-steel.ds2 A 1 mm steel plate in vacuum, a typical Lamb wave (8.1) case.


plate-steel-sh.ds2 The SH modes (8.1.2)of a 1 mm steel plate in vacuum.

plate-steel-in-water.ds2 A 1 mm steel plate in water, a typical Leaky Lamb wave (8.2) case.
ray-alum.ds2 A Rayleigh wave (8.3) on a semi-infinite half-space of aluminum.
ray-alum-water.ds2 A Rayleigh wave (8.3) on a semi-infinite half-space of aluminum that is water
loaded.

stonely-al-stl.ds2 A Stoneley wave that is propagating at the interface between steel and aluminum
half-spaces(8.3.3).
epoxy-on-alum.ds2 An epoxy layer on the top of an aluminium half-space (8.3.4).
hppe.ds2 A 1 mm layer of high loss visco-elastic (8.4) high performance polyethylene.

plate-adhesive-joint.ds2 An aluminium epoxy joint with 1 mm thick aluminum plates (8.5).


plate-adhesive-joint-leaky.ds2 An aluminium epoxy joint with 1 mm thick aluminum plates, im-
mersed in water, all of the possible modes (8.5).
plate-adhesive-joint-1gpa.ds2An aluminium epoxy joint with 1 mm thick aluminium plates and
epoxy stiffness reduced to 1 GPa(8.5).
plate-al-epx-al.ds2 An epoxy layer that is embedded (8.6) in aluminium half-spaces.
plate-gre-unidir-0deg.ds2 A single layer unidirectional Graphite Reinforced Epoxy composite,
with the wave propagation parallel to the fibres (8.7).

Disperse version 2.0.20a


190

plate-gre-unidir-45deg.ds2 A single layer unidirectional Graphite Reinforced Epoxy composite


with the wave propagation at 45 degrees to the fibres (8.7).
plate-gre-quasi-isotrop-0deg.ds2 A [45,-45,0,90]S Graphite Reinforced Epoxy composite with the
wave propagation in the direction defined by zero degrees (8.7).

plate-gre-quasi-isotrop-45deg.ds2 A [45,-45,0,90]S Graphite Reinforced Epoxy composite with the


wave propagation in the direction defined by 45 degrees (8.7).
bar-steel-2.ds2 A 2 mm diameter steel bar in vacuum (9.1.4).
bar-steel-2-in-water.ds2 A 2 mm diameter steel bar in water (9.1.4).

cyl-vac2-stl-vac-labeled.ds2 A 1 mm thick steel pipe (9.1) with an inner radius of 2 mm, with all
of the modes labelled.
cyl-vac1-stl-vac.ds2 A 1 mm thick steel pipe with an inner radius of 1 mm.
cyl-vac5-stl-vac.ds2 A 1 mm thick steel pipe (9.1 with an inner radius of 5 mm.

cyl-vac10-stl-vac.ds2 A 1 mm thick steel pipe (9.1 with an inner radius of 10 mm.


cyl-vac20-stl-vac.ds2 A 1 mm thick steel pipe (9.1 with an inner radius of 20 mm.
cyl-h20f10-stl-h20.ds2 A 1 mm thick steel pipe with an inner radius of 10 mm that is filled with
and surrounded by water (9.2).

cyl-h20s10-stl-h20.ds2 A 1 mm thick steel pipe with an inner radius of 10 mm that is filled with
and surrounded by water and has a sink in the center of the liquid core (9.2).
bar-stl-stiff-grout.ds2 A 2 mm diameter steel bar in a stiff grout-like material (9.3).
cyl-vac10-stl1-mudstone.ds2 A 1 mm thick steel pipe with an inner radius of 10 mm that is
surrounded by mudstone (9.3).
cyl-vac9.5-grt0.5-stl1-mudstone.ds2A 1 mm thick steel pipe with an inner radius of 10 mm that
is lined with grout and surrounded by mudstone (9.3).

Disperse version 2.0.20a


Appendix C

Common Material Properties

The following is tabulation of some common material properties. Most of this information was
taken from references [41, 38, 4, 37, 21, 11, 27]. Please note that these values are a compilation
from several sources and that inconsistencies are possible. Also, the range of possible values has
not been included for many of the materials listed, and some of the values are omitted because
they are not known. The shear attenuation (αs ) values for solids are very rarely available and so
are not shown. In general, the shear attenuation will be several times larger than the longitudinal
attenuation.

Disperse version 2.0.20a


192

Table C.1: Isotropic Solid Materials


Name Density (kg/m3 ) vl (m/s) vs (m/s) α (np/λ)
Metals
Aluminum 2700 6320 3130 0.0003
Bismuth 9800 2180 1100 –
Brass 8400 4400 2200 –
Cadmium 8600 2780 1500 –
Cast iron 6900to 7300 3500 to 5800 2200 to 3200 –
Constantan 8800 5240 2640 –
Copper 8900 4700 2260 –
Gold 19300 3240 1200 –
Iron (steel) 7700 5960 3260 0.003 –
Lead 11400 2160 700 –
Magnesium 1700 5770 3050 –
Manganin 8400 4660 2350 –
Nickel 8800 5630 2960 –
Platinum 21400 3960 1670 –
Silver 10500 3600 1590 –
Tin 7300 3320 1670 –
Titanium 4460 6060 3230 –
Tungsten 19100 5460 2620 –
Zinc 7100 4170 2410 –
Non-metals
Aluminum Ox- 3600 to 3950 9000 to 11000 5500 to 6500 –
ide
Bone,tibia 1900 4000 1970 0.64
Concrete 2200 3900 to 4700 2300 0.2
Epoxy Resin 1100 to 1250 2400 to 2900 1100 0.03
Fat 920 1450 – 0.008
Glass,flint 3600 4260 2560 –
Glass, crown 2500 5660 3420 –
Ice 900 3980 1990 –
Paraffin Wax 830 2200 – –
MDPE 950 2300 950 0.03 (and
αs =0.15)
Neoprene – 1690 – 0.15
Perspex (acrylic 1180 2730 1430 0.06
resin)
Polyamide (ny- 1100 2200 1100 0.007
lon)
Polystyrene 1060 2350 1150 0.02
Porcelain 2400 5600 to 6200 3500 to 3700 –
Quartz Glass 2600 5570 3520 –
Rock 2500 2500 to 4400 1500 to 2500 –
Rubber, soft 900 1480 – 0.6
Rubber, vulcan- 1200 2300 – –
ized
Skin 1110 1730 – 0.023
Teeth (enamel) 2900 to 3000 4500 to 6250 – 0.2
Teflon 2200 1350 550 0.12

Disperse version 2.0.20a


193

Table C.2: Fluid Materials


Name Temperature Density vl (m/s) Dynamic Kinematic
(0 C) (kg/m3 ) viscosity η viscosity ν
(N s/m2 ) (m2 /s)
Air 0 1.293 331 17.1x10−6 13.2x10−6
Air 20 1.205 344 18.1x10−6 15.0x10−6
Air 40 1.128 355 19.0x10−6 16.8x10−6
Air 100 0.946 388 21.8x10−6 23.1x10−6
Benzene 20 881 1117 0.65x10−3 0.74x10−6
Castor oil 20 971 1474 0.8 0.82x10−3
Castor oil 40 – – 0.2 –
Crude oil 20 860 – 8x10−3 9.3x10−6
Diesel oil – 800 1250 – –
Ethyl alcohol 20 790 1238 1.2x10−3 1.52x10−6
Glycerine 20 1258 1860 1.49 1.18x10−3
Kerosene 20 822 1319 1.5x10−3 1.82x10−6
Mercury 20 13550 1391 1.56x10−3 0.12x10−6
Methylene io- – 3230 980 – –
dide
Motor car oil – 870 1740 – –
SAE 10W oil 0 – – 0.4 –
SAE 10W oil 20 922 – 0.1 0.11x10−3
SAE 10W oil 60 – – 0.017 –
SAE 10W oil 100 – – 0.005 –
SAE 30 oil 20 – – 0.4 –
Water 0 999.8 1421 1.781x10−3 1.785x10−6
Water 20 998.2 1478 1.002x10−3 1.003x10−6
Water 50 988.0 1522 0.547x10−3 0.553x10−6
Water 100 958.4 1470 0.282x10−3 0.294x10−6

Table C.3: Anisotropic Materials


Name ρ E11 E22 E33 G12 G13 G23 ν12 ν13 ν23
kg/m3(GPa) (GPa) (GPa) (GPa) (GPa) (GPa)
Unidirectional Carbon Fi- 1600 172 11.6 11.6 7.8 7.8 3.9 0.36 – 0.48
bre/Epoxy - high strength
Unidirectional Carbon Fi- 1600 287 7.8 7.8 6.7 6.7 2.5 0.30 – 0.55
bre/Epoxy - high modulus
Unidirectional Glass Fi- – 47.0 16.4 – – – – 0.28 – –
bre/Epoxy - in-plane prop-
erties
Oak 600 4.7 1.1 1.4 0.92 0.89 0.3 0.75 0.8 .24

Disperse version 2.0.20a


Appendix D

Bibliography

[1] F. Abramovici. Diagnostic diagrams and transfer functions for oceanic wave-guides. Bulletin
of the Seismological Society of America, 58:427–456, 1968.
[2] L. Adler, M.d. Billy, and G. Quentin. Evaluation of friction-welded aluminum-steel bonds
using dispersive guided modes of a layered substrate. Journal of Physics D; Applied Physics,
68, 1990.

[3] L. Alsop. The leaky-mode period equation - a plane-wave approach. Bulletin of the Seismo-
logical Society, 6:1989–1998, 1970.
[4] B.A. Auld. Acoustic Fields and Waves in Solids, volume 2. Krieger Publishing Company
Malabar, Florida, 1990.

[5] M.J. Berliner and R. Solecki. Wave propagation in fluid loaded, transversely isotropic cylin-
ders. part ii. numerical results. Journal of the Acoustical Society of America, 99:1848–1853,
1996.
[6] M.J. Berliner and R. Solecki. Wave propagation in fulid-loaded, transversely isotropic cylin-
ders. part i. analytical formulation. Journal of the Acoustical Society of America, 99:1841–
1847, 1996.
[7] A. Bernard, M. Deschamps, and M.J.S. Lowe. Energy velocity and group velocity for guided
waves propagating within an absorbing or non-absorbing plate in vacuum. In D.O. Thompson
and D.E. Chimenti, editors, Review of Progress in Quantitative NDE, volume 18, pages 183–
190. Plenum Press, New York, 1999.

[8] A. Bernard, M. Deschamps, and M.J.S. Lowe. Comparison between the dispersion curves
calculated in complex frequency and the minima of the reflection coefficients for an embedded
layer. J. Acoust. Soc. am., 107:793–800, 2000.
[9] L.M. Brekhovskikh. Waves in layered media. Academic Press, New York, 1980.

[10] L.M. Brekhovskikh and V. Goncharov. Mechanics of continua and wave dynamics. Springer-
Verlag, 1985.
[11] J. Bugler. Fluid mechanics for technologists. Longman Cheshire, Melbourne, 1989.
[12] M. Castaings. Propagation ultrasonore dans les milieux stratifies plans constitues de mate-
riaux absorbants et orthotropes. PhD thesis, University of Bordeaux I, 1993.

Disperse version 2.0.20a


Bibliography 195

[13] M. Castaings and B. Hosten. Delta operator technique to improve the thomson-haskell
method stability for propagation in multilayered anisotropic absorbing plates. Journal of
the Acoustical Society of America, 95:1931–1941, 1994.
[14] P. Cervenka and P. Challande. A new efficient algorithm to compute the exact reflection and
transmission factors for plane waves in layered absorbing media (liquids and solids). Journal
of the Acoustical Society of America, 89:1579–1589, 1991.
[15] C.W. Chan and P. Cawley. Lamb waves in highly attenuative plastic plates. Journal of the
Acoustical Society of America, 104(2):874–881, 1998.
[16] D. Chimenti and A. Nayfeh. Ultrasonic reflection and guided wave propagation in biaxially
laminated composite plates. Journal of the Acoustical Society of America, 87:1409–1415,
1990.
[17] D.E. Chimenti, A.H. Nayfeh, and D. Butler. Leaky rayleigh waves on a layered halfspace.
Journal of Applied Physics, 53:170–176, 1982.
[18] C. Chree. The equations on an isotropic elastic solid in polar and cylindrical coordinates,
their solutions, an applications. Trans. Cambridge Philos. Soc., 14:250–369, 1889.
[19] M.D. Cochran, A.F. Woeber, and J.-C. Bremaecker de. Body waves as normal and leaking
modes. Reviews of Geophysics and Space Physics, 8:321–357, 1970.
[20] A.M. Dainty. Leaking modes in crust with a surface layer. Bulletin of the Seismological
Society of America, 61:93–107, 1971.
[21] J. R. L. Daugherty, J. B. Franzini, and J. Finnemore, E. Fluid mechanics with engineering
applications. McGraw-Hill, s i metric edition edition, 1989.
[22] R. M. Davies. A critical study of the hopkinson pressure bar. Phil. Trans. Roy. Soc. London
A, 240:375–457, 1948.
[23] M. Deschamps. L’onde plane heterogene et ses applications en acoustique lineaire. Journal
Acoustique, 4:269–305, 1991.
[24] M. Deschamps and B.E. Hosten. The effects of viscoelasticity on the reflection and trans-
mission of ultrasonic waves by an orthotropic plate. J. Acoust. Soc. am., 91:1–29, 1992.
[25] W.M. Ewing, W.S. Jardetzky, and F. Press. Elastic waves in layered media. McGraw-Hill,
1957.
[26] G.W. Farnell and E.L. Adler. Elastic wave propagation in thin layers. In W.P. Mason and
R.N. Thunston, editors, Physical Acoustics - principles and methods, pages 35–127, 1972.
[27] R. W. Fox and A. T. McDonald. Introduction to fluid mechanics. John Wiley, 4 edition,
1992.
[28] D. C. Gazis. Three-dimensional investigation of the propagation of waves in hollow circular
cylinders. ii. numerical results. JASA, 31(5):573–578, may 1959.
[29] D.C. Gazis. Three dimensional investigation of the propagation of waves in hollow circular
cylinders. Journal of the Acoustical Society of America, 31:568–578, 1959.
[30] F. Gilbert. Propagation of transient leaking modes in a stratified elastic waveguide. Reviews
of Geophysics, 2:123–153, 1964.
[31] K.F. Graff. Wave Motion in Elastic Solids. Dover Publicaitons inc., New York, 1973.
[32] N. Guo and P. Cawley. The interaction of lamb waves with delaminations in composite
laminates. Journal of the Acoustical Society of America, 94:2240–2246, 1993.

Disperse version 2.0.20a


Bibliography 196

[33] N.A. Haskell. The dispersion of surface waves on multi-layered media. Bulletin of the
American Seismological Society, 43:17–34, 1953.
[34] B. Hosten. Bulk heterogeneous plane wave propagation through viscoelastic plates and
stratified media with large values of frequency domain. Ultrasonics, 29:445–450, 1991.

[35] B. Hosten and M. Castings. Transfer matrix of multilayered absorbing and anisotropic media.
measurements and simulations of ultrasonic wave propagation through composite materials.
Journal of the Acoustical Society of America, 94:1488–1495, 1993.
[36] G. E. Hudson. Dispersion of elastic waves in solid circular cylinders. Phys. Rev., 63:46–51,
1943.

[37] S. C. Hunter. Mechanics of continuous media. Ellis Horwood, Chichester, 1976.


[38] G.W.C. Kaye and T.H. Laby. Tables of Physical and Chemical Constants. Longman, Essex,
16 edition, 1995.
[39] L. Knopoff. A matrix method for elastic wave problems. Bulletin of the Seismological Society
of America, 54:431–438, 1964.
[40] H. Kolsky. Stress waves in solids. Dover Publications, New York, 1963.
[41] J. Krautkramer and H. Krautkramer. Ultrasonic testing of materials. Springer-Verlag, 1983.
[42] E. Kreyszig. Advanced Engineering Mathematics. John Wiley + Sons, inc., 1993.

[43] R. Kumar. Flexural vibrations of fluid-filled circular cylindrical shells. Acustica, 24:137–146,
1971.
[44] R. Kumar. Dispersion of axially symmetric waves in empty and fluid-filled cylindrical shells.
Acustica, 27:317–329, 1972.

[45] W.M. Lai, D. Rubin, and E. Krempl. Introduction to Continuum Mechanics. Pergamon
Press, Oxford, 3 edition, 1993.
[46] H. Lamb. On waves in an elastic plate. In Conference of the Royal Society, pages 114–128,
1917.
[47] S. Laster, J. Foreman, and F. Linville. Theoretical investigation of modal seismograms for
a layer over a half-space. Geophysics, pages 571–596, 1965.
[48] O.I. Lobkis and D.E. Chimenti. Elastic guided waves in plates with surface roughness. i.
model calculation. J. Acoust. Soc. am., 102:143–149, 1997.
[49] O.I. Lobkis and D.E. Chimenti. Elastic guided waves in plates with surface roughness. ii.
experiments. J. Acoust. Soc. am., 102:150–159, 1997.
[50] A. E. H. Love. Some problems of geodynamics. Cambridge University, London, 1911.
[51] M. Lowe. Matrix techniques for modeling ultrasonic waves in mutilayered media. IEEE
Transactions on Ultrasonics, Ferroelectrics and Frequency control, 42:525–542, 1995.

[52] M. Lowe and P. Cawley. The detection of a brittle layer at the bondline in diffusion bonded
titanium. In D.O. Thompson and D.E. Chimenti, editors, Review of Progress in Quantitative
NDE, pages 1653–1659. Plenum Press, New York, 1993.
[53] M.J.S. Lowe. Plate waves for the NDT of diffusion bonded titanium. PhD thesis, University
of London, 1993.

Disperse version 2.0.20a


Bibliography 197

[54] M.J.S. Lowe and P. Cawley. The applicability of plate wave techniques for the inspection
of adhesive and diffusion bonded joints. Journal of Non-destructive Evaluation, 13:185–199,
1994.
[55] Malvern. Introduction to the mechanics of a continuous medium. Prentice-Hall, 1969.

[56] D.F. McCammon and S.T. McDaniel. The influence of the physical properties of ice on
reflectivity. Journal of the Acoustical Society of America, 77:499–507, 1985.
[57] T. R. Meeker and A. H. Meitzler. Guided wave propagation in elongated cylinders and
plates. In W. P. Mason and R. N. Thurston, editors, Physical Acoustics, Principles and
Methods, volume 1A, pages 111–167, New York, 1972. Academic Press.

[58] I. Mirsky. Wave propagation in transversely isotropic circular cylinders part i: Theory.
Journal of the Acoustical Society of America, 37:1016–1021, 1965.
[59] P.M. Morse and H. Feshbach. Methods of Theoretical Physics. McGraw-Hill Book Company,
1953.

[60] P. B. Nagy and A.H. Nayfeh. Viscosity-induced attenuation of longitudinal guided waves in
fluid-loaded rods. J. Acoust. Soc. am., 100:1501–1508, 1996.
[61] P.B. Nagy. Longitudinal guided wave propagation in a transversely isotropic rod immersed
in fluid. Journal of the Acoustical Society of America, 98:454–457, 1995.
[62] P.B. Nagy and L. Adler. Non-destructive evaluation of adhesive joints by guided waves.
Journal of Applied Physics, 66:4658–4663, 1989.
[63] A.H. Nayfeh. Wave propagation in layered anisotropic media with application to composites.
Elsevier, 1995.
[64] A.H. Nayfeh and D.E. Chimenti. Ultrasonic leaky waves in the presence of a thin layer.
Journal of Applied Physics, 52:4985–4994, 1981.
[65] M. Onoe, H.D. McNiven, and R.D. Mindlin. Dispersion of axially symmetric waves in elastic
solids. Journal of Applied Mechanics, 29:729–734, 1962.
[66] Y. H. Pao. The dispersion of flexural waves in an elastic, circular cylinder – part 2. J. Appl.
Mech., 29:61–64, 1962.

[67] Y.-H. Pao and R.D. Mindlin. Dispersion of flexural waves in an elastic, circular cylinder.
Journal of Applied Mechanics, 27:513–520, 1960.
[68] B.N. Pavlakovic and M.J.S. Lowe. A general purpose approach to calculating the longitudinal
and flexural modes of multi-layered, embedded, transversely isotropic cylinders. In D.O.
Thompson and D.E. Chimenti, editors, Review of Progress in Quantitative NDE, volume 18,
pages 239–246. Plenum Press, New York, 1999.
[69] B.N. Pavlakovic, M.J.S. Lowe, D.N. Alleyne, and P. Cawley. Disperse: A general purpose
program for creating dispersion curves. In D.O. Thompson and D.E. Chimenti, editors,
Review of Progress in Quantitative NDE, volume 16, pages 185–192. Plenum Press, New
York, 1997.
[70] C. Pecorari and P. Kelly. The quasi-static approximation for cracked interfaces in layered
systems. In D.O. Thompson and D.E. Chimenti, editors, Review of Progress in Quantitative
NDE, volume 18. Plenum Press, New York, 1999.
[71] W. J. Percival and E. A. Birt. A study of lamb wave propagation in carbon-fibre composites.
Insight, 39:728–735, 1997.

Disperse version 2.0.20a


Bibliography 198

[72] R.A. Phinney. Leaking modes in the crystal waveguide. Journal of Geophysical Research,
66:1445–1469, 1961.
[73] R.A. Phinney. Propagation of leaking interface waves. Bulletin of the Seismological Society
of America, 51:527–555, 1961.

[74] T.P. Pialucha. The reflection coefficient from interface layers in NDT of adhesive joints.
PhD thesis, University of London, 1992.
[75] W.L. Pilant. Complex roots of the stonely wave equation. Bulletin of the Seismological
Society of America, 62:285–299, 1972.
[76] A. Pilarski. Ultrasonic evaluation of the adhesion degree in layered joints. Materials Evalu-
ation, 43:765–770, 1985.
[77] A. Pilarski, J.L. Rose, J. Ditri, D. Jiao, and K. Rajana. Lamb wave mode selection for
increased sensitivity to interfacial weaknesses of adhesive bonds. In D.O. Thompson and
D.E. Chimenti, editors, Review of Progress in Quantitative NDE. Plenum Press, New York,
1992.

[78] J. Pochhammer. Uber die fortpflanzungsgeschwindigkeiten kleiner schwingungen in einem


unbergrenzten isotropen kreiscylinder. J fur reine und angewandte Math., 81:324–336, 1876.
[79] O. Poncelet and M. Deschamps. Lamb waves generated by complex harmonic inhomogeneous
plane wave. Submitted to Journal of the Acoustical Society of America, 1996.

[80] M.J. Randall. Fast programs for half-space problems. Bulletin of the Seismological Society
of America, 57:1299–1315, 1967.
[81] L. Rayleigh. On waves propagating along the plane of an elastic solid. Proc. London Math.
Soc., 17, 1885.

[82] J.H. Rosenbaum. The long-time response of a layered elastic medium to explosive sound.
Journal of Geophysical Research, 65:1577–1613, 1960.
[83] A. Safaai-Jazi, C.-K. Jen, and G.W. Farnell. Cutoff conditions in an acoustic fiber with
infinitely thick cladding. IEEE Transactions on Ultrasonics, Ferroelectrics and Frequency
control, UFFC-33:69–73, 1986.

[84] H. Schmidt and F. Jensen. Efficient numerical solution technique for wave propagation in
horizontally stratified environments. Computers and Maths with Applications, 11:699–715,
1985.
[85] H. Schmidt and F.B. Jensen. A full wave solution for propagation in multilayered viscoelastic
media with application to gaussian beam reflection at liquid-solid interfaces. Journal of the
Acoustical Society of America, 77:813–825, 1985.
[86] H. Schmidt and G. Tango. Efficient global matrix approach to the computation of synthetic
seismograms. Geophysics Journal of the Royal Astronomical Society, 84:331–359, 1986.
[87] M. Schoenberg. Elastic wave behaviour across linear slip interfaces. Journal of the Acoustical
Society of America, 68:1516–1521, 1980.
[88] J.G. Scholte. The range of existence of rayleigh and stoneley waves. Geophysics, 5:120–126,
1947.
[89] F. Schwab and L. Knopff. Surface-wave dispersion computations. Bulletin of the Seismolog-
ical Society of America, 60:321–344, 1970.

Disperse version 2.0.20a


Bibliography 199

[90] F. Schwab and L. Knopff. Surface waves on multilayered anelastic media. Bulletin of the
Seismological Society of America, 61:893–912, 1971.
[91] F.A. Schwab and L. Knopoff. Fast surface wave and free mode computations. In Methods
in Computational Physics, volume 11. Academic Press, 1972.

[92] M.G. Silk. Ultrasonic transducers for nondestructive testing. Adam Hilger, Bristol, 1984.
[93] M.G. Silk and K.F. Bainton. The propagation in metal tubing of ultrasonic wave modes
equivalent to lamb waves. Ultrasonics, pages 11–19, 1979.
[94] J.A. Simmons, E. Drescher-Krasicka, and H.N.G. Wadley. Leaky axisymmetric modes in
infinite clad rods. i. Journal of the Acoustical Society of America, 92:1061–1090, 1992.

[95] C. Simon, H. Kaczmarek, and D. Royer. Elastic wave propagation along arbitrary direction
in free orthotropic plates. application of composite materials. In 4th French Congress of
Acoustics, Marseille, France, 1997.
[96] R. Stoneley. Elastic waves at the surface of separation of two solids. In Conference of the
Royal Society, pages 416–428, 1924.
[97] W.T. Thomson. Transmission of elastic waves through a stratified solid medium. Journal
of Applied Physics, 21:89–93, 1950.
[98] R.N. Thurston. Elastic waves in rods and clad rods. Journal of the Acoustical Society of
America, 64:1–37, 1978.

[99] M. Viens, Y. Tshukahara, C.K. Jen, and J.D.N. Cheeke. Leaky torsional modes in infinite
clad rods. Journal of the Acoustical Society of America, 95:701–707, 1994.
[100] I. A. Viktorov. Rayleigh and Lamb waves. Plenum Press, New York, 1970.
[101] T.H. Watson. A real frequency complex wave-number analysis of leaking modes. Bulletin
of the Seismological Society of America, 62:369–384, 1972.
[102] P.D. Wilcox, M.J.S. Lowe, and P. Cawley. Long range lamb wave inspection: the effect of
dispersion and modal selectivity. In D.O. Thompson and D.E. Chimenti, editors, Review of
Progress in Quantitative NDE, volume 18, pages 151–158. Plenum Press, New York, 1999.

[103] P.-C. Xu and A.K. Mal. Calculation of the inplane green’s functions for a layered viscoelastic
solid. Bulletin of the Seismological Society of America, 77:1823–1837, 1987.

Disperse version 2.0.20a


Appendix E

Index

Abbrev., 24 dB/wl, 53
Actions, 14, 71, 82 Default scale type, 60
Actions : Copy text, 65 Define, 25
Actions : Custom zoom, 57 Description, 53
Add to List, 24, 35 Hysteretic damping per meter, 25
Adhesive joint Hysteretic structural damping, 25
Defining, 149 Material damping, 147
Example, 149 Nepers, 53
Tracing, 149 np/wl, 53
Weakened, 150 Relations between properties, 183
Advanced, 65–67, 71, 142 Surface roughness, 26
Air, 55 Theory, 90
All points, 61 Viscous damping, 25
Amp, 73 Attenuation per wavelength, 75
Amp mod, 70, 71 Attenuation-thickness, 54
Analytical solutions, 47 Automatic tracing, 37
Leaky Lamb Waves, 138 Autoscale, 14, 57, 71
And also to, 41 Axis scaling
Angle, 85 f*d, 61
Angle of incidence, 76
Definition, 55 Backward travelling waves, 41, 46
Angles Bar
Rotation, 26 Defining, 163
Animate, 67 Bessel functions, 22
Animation delay, 68 Reoccurance relations, 187
Anisotropic Bibliography, 194
Cubic, 28 Black grid, 67
Example, 153 Bond transformation matrix, 26
Group velocity, 53 Boundary conditions
Orientation, 26 External, 35
Orthotropic, 28 Sliding, 119
Solid, 26 Bulk velocities, 78
Transversely, 28 Spurious roots, 39, 118
Trigonal, 28
Anisotropic materials Calculate function, 73, 76
complex Cij, 29, 186 Sampling, 78
relations between constants, 184 Calculating at every:, 78
Antisymmetric, 40 Change name of mode, 84
Arrow, 58, 59, 85 Check convergence, 78
Attachment, 85 Check energy consistency, 78, 79
Attenuation, 25, 41, 54 Check energy/group velocity, 79

Disperse version 2.0.20a


Index 201

Cij, 28 Limitations, 22
complex, 29, 186 Modelling of Core, 169
Circle, 85 Multi-layered, 177
Circumferential, 20 Naming convention, 164
Circumferential order Nature of modes
Selecting for solution, 164 Bars, 165
Circumferential waves, 48, 52, 54 Pipes, 166
Clipboard Radius, 21
Colour, 60 Substitution of Cartesian, 169
Copying, 82 Tracing, 163
Col 1:, 82 Types of Leakage, 176
Colour, 56, 82, 85 Cylindrical modes
Change, 59 Flexural, 164
Colour axis Longitudinal, 164
Default colours, 60 Mode shapes, 166
Coloured grid, 67 Torsional, 164, 171
Column order, 82
Combine, 60 Debug, 42
Compare, 83 Decoupling SH modes, 135
Complex Cij, 29, 186 Default directories, 12
Complex frequency, 46, 55 Defining the structure, 20
Consecutive, 61 Delay, 71
Constant distance, 67 Delete, 21, 83, 149
Constant wavelength, 67, 68 Curve, 59
Constitutive relations, 182 Foreign modes, 56
Continuously variable properties, 48 Point, 59
Control, 58 Delete an entire mode, 148
Convergence, 42 Delete label, 85
Convergence parameters, 42 Delete point, 59, 148
Copy, 71, 82, 83 Delimiter, 82
Copy all, 83 Derived data, 56
Copy Metafile, 82 Determinant
Copy Picture, 83 Display, 55
Copy picture, 82 Different types, 176
Copy text, 82 Displacement mode shapes, 62
Cross, 85 Display, 13, 51, 52, 54, 56, 73, 135
Cubic, 28 Amplitude of mode shape, 68
Cursor, 85 Angle of incidence, 55
Curve Attenuation, 53
Combine, 60 Compare, 83
Delete, 59 Determinant, 55
Split, 59 Energy velocity, 55
Undelete, 59 Function result, 55
Curves, 81 Group slowness, 55
Custom zoom, 57 Group velocity, 52
Cut-off frequencies, 135 Mode shapes in half spaces, 66, 68
Cycles, 70, 71 Partial waves, 69
Cylinder, 20, 23 Phase of mode shape, 68
Cylindrical, 22 Phase slowness, 55
Defining leaky, 163 Phase velocity, 52
Embedded, 175 Preferences, 60
Examples, 162 Types, 52
Immersed, 169 Wavenumber, 54
Large radius, 168 Display : Attenuation, 53, 60

Disperse version 2.0.20a


Index 202

Display : Autoscale, 57 Rayleigh waves, 140


Display : Custom zoom, 57 SH modes, 135
Display : Derived data, 55, 73 Stoneley wave, 143
Display : Group velocity, 52 Summary, 189
Display : Hide curves, 56 Thin layer on half-space, 144
Display : Phase velocity, 52 Visco-elastic, 147
Display : Preferences, 53, 60, 67 Excitation, 70, 72
Display : Real wave number, 54 Export data, 80
Display : Second graph, 54 Export Movie, 69
Display : Second graph : Attenuation, 60 Exporting, 80
Display : Second graph : Hide, 54 Point header, 61
Display : Select axes, 54, 57, 60 Points, 61
Display : Sweep results, 40 Settings, 61
Display : Unzoom, 57, 58 Side by side, 61
Display as phase, 65, 66 Title header, 61
Display in half spaces, 67
Display logic, 13 f*d, 61
Do not modulate frequency, 72 F10, 41
Dynamic viscosity, 93 F11, 43
F3, 58
Edit, 24, 39, 59 F4, 57
Name of mode, 84 F7, 37
Edit : Copy, 61, 82 F9, 40
Edit : Copy all, 61, 82 fd, 61
Edit : Delete, 59 Fiber fraction, 29
Edit : Delete : All labels, 85 File : Import text file, 81
Edit : Delete : Bulk velocity lines, 78 File : New, 17, 18, 20, 31, 134, 149, 154, 163
Edit : Delete : Foreign modes, 56 File : Open, 20, 35, 80
Edit : Delete : Next point selected, 39, 149 File : Print, 18, 19, 63, 71, 84
Edit : Insert new object, 83 File : Print preview, 84
Edit curves, 14 File : Print setup, 84
Edit curves : Change colour, 59 File : Save, 18, 19, 80
Edit curves : Combine, 60 File : Save as, 80
Edit curves : Hide curve, 56 File extension, 15, 80
Edit curves : Split, 59 File format, 80
Edit label, 84 .ds2, 15
Edit list, 22 File structure, 179
Embedded layers File: New, 156, 158
Example, 150 Find minima, 14, 41, 145
Tracing, 152 Fixed boundary, 35
Energy velocity, 52, 75, 131 Flat top Gaussian, 71
Display, 55 Flat top ring down, 71
Enhanced Metafile, 81 Flat top sin, 71
Erase mode, 59 Fluid, 24
Eraser, 59 Modelled as solid, 25
Example files, 12 Viscous, 24, 91
Examples, 20 Fluid filled core, 169
Anisotropic, 153 Sink, 171
Cartesian, 133 Foreign, 56
Cylindrical, 162 Foreign modes, 56, 83
Embedded layers, 150 Formatting
Lamb waves, 133 Display, 51
Leaky Lamb waves, 138 Export, 80
Multi-layered structures, 149 File, 80

Disperse version 2.0.20a


Index 203

Free, 35 Inc. Angle (dft), 55


Free boundary, 21, 22, 35 Include point count, 61
Freq span, 72 Include text headers, 61
Freq-wavn combo, 43, 44 Increment, 43, 134
Frequency, 70, 71, 79 Information about manual, 180
Frequency (constant phase velocity), 40 Input and output, 80
Frequency modulation, 72 Insert after, 21
Frequency-thickness, 54, 61 Insert before, 21
From, 40 Insert new object, 83
Function, 73 Installation, 12
Display, 55 Interface, 35
Fundamental modes Spring, 119
Bar Introduction, 10
F(1,1), 166 Isotropic solid, 25
L(0,1), 166 Iteration type, 44
Lamb waves, 135
Leaky Lamb waves, 140 Job priority boost, 48
Pipe Jobs, 39, 42, 45, 51, 52
F(1,1), 168 Jumping, 42
L(0,1), 166
Kelvin-Voigt viscous damping, 25
Gaussian, 70 Kinematic viscosity, 93
Geometry, 34, 51, 52 Kinetic energy, 75
Global matrix, 36, 47 Kinetic energy density, 65, 130
Go, 79
Green circle, 40 Label text, 85
Grid, 63–67 Labels, 84
Amplitude of deformed shape, 67 Lamb waves
Colour, 67 Analytical solution, 47
Line every, 68 Circumferential, 48, 52, 54
Mode shapes Cut-off frequencies, 135
Example, 67 Defining, 133
Grid line every, 68 Examples, 133
Group slowness, 75 Fundamental modes, 135
Group velocity, 54, 79 High order modes, 135
Definition, 52 Interpreting, 134
Labelling, 134, 135
Half space, 20 Leaky, 138
Display, 67 Quick start, 18
Mode shapes, 66 Quick start, 17
Ratio, 68 Tracing, 134
Hanning, 71 Large radius, 168
Helmholtz, 98 Layout, 81
Hide, 54 Leakage
All but selected, 56 Cylindrical, 173
Curves, 56 Embedded layers, 150
Next selected, 56 Examples, 176
Other orders/symmetry, 56 Fluid filled core, 169
Holding the following constant:, 40 Into a Solid, 175
HPPE, 147 Into surrounding media, 173
Hysteretic damping (per meter), 25 Leaky Lamb Waves, 140
Hysteretic structural damping, 24, 25 Matching wave numbers, 173
Non-leaky, 173
Import text file, 81 Rayleigh waves, 142

Disperse version 2.0.20a


Index 204

Leaky Lamb waves, 138 Leaky Rayleigh waves, 143


Defining, 139 Lines, 64
Example, 138 Movie, 69
Leaky Rayleigh wave, 142 Moving along curve, 63
Mode shape, 143 Number of cycles, 68
Leaky systems, 41, 45 Number of points, 65
Leaky Waves Numerical values, 65
Cylindrical, 107 Orientation of graph, 64
Length, 85 Partial wave, 68
Line plots, 68 Phase, 64
Lines, 64, 67, 68, 131 Points per layer, 68
Bulk velocities, 78 Power, 69
Mode shapes, 64 Rayleigh waves, 142
Liquid, 24 Scale, 62, 66
Cartesian, 112 Significance of grid, 66
Cylindrical, 112 Stoneley wave, 143
Long. atten., 24 Time step, 68
Units, 65
Manual Modes/points to verify, 78
Issue date, 180 Mouse, 13
Manual tracing, 39 Left button, 13
Marker, 85 Right button, 13, 40, 57
Material definition, 24 Move, 57
Material properties Move label, 85
Anisotropic, 26, 193 Multi-layered
Fluid, 24, 193 Cylindrical, 177
Isotropic solid, 25, 192 Multi-layered structures
List, 191 Example, 149
New, 24 Multi-mode, 70, 72
Spring, 34 Multiple mode, 41
Variable, 48 Multiplication, 76
Variation with frequency, 33
Material.lst, 13, 22, 179 Name, 24
Materials database, 13, 179 Name of a mode, 85
Max amp, 71 Naming, 39
Mixed boundary, 35 Cylindrical modes, 164
Mode, 73 Native modes, 56
Mode jumping, 42 Negative wavenumber, 46
Mode names, 58 Nepers
Cylindrical, 164 Conversion to dB, 53
Mode shapes, 62, 63, 67 New features for version 2.0, 14
Advanced, 67 New label, 84
Amplitude or phase, 68 New material, 24
Animate, 67 No material damping, 25
Animation delay, 68 No point count, 61
Cylindrical, 166 No text headers, 61
Density of grid lines, 68 Non-leaky mode, 173
Display in half spaces, 68 Non-propagating roots, 46
Exporting, 65 Norm. Thick., 22
Grid, 66 Normalized frequency, 54
Half space ratio, 68 Notes, 51, 52
Half spaces, 65 Nudge zoom, 57
Immersed bar, 173 Num header lines, 82
Leakage into a Solid, 177 Number of cycles, 68

Disperse version 2.0.20a


Index 205

Object Linking and Embedding, 82 Projections, 52, 164


OK, 80 Lamb waves, 134
OLE, 82 Prop dist, 71
One mode, 41 Properties vary as a function of frequency or
Only axis titles, 61 user variable, 48
Order, 40 Properties vary between modes, 48
Orthotropic, 28 Pulse-echo, 72
Orthotropic materials
complex Cij, 29, 186 Quick start
relations between constants, 184 Lamb waves, 17
Leaky Lamb waves, 18
Partial waves, 36, 68, 69
Paste, 82, 83 Radius
Perform 2-D attenuation sweep, 40 Effect of changing, 168
Perspex, 55 Rayleigh waves, 20, 140
Phase, 68 Defining, 142
Mode shapes, 66 Equation, 142
Phase slowness, 75 Leaky, 142
Phase velocity, 52, 54, 79 Re-assemble as more layers and re-check, 79
Pitch-catch, 72 Reaches:, 41
Pitch-catch/Pulse-echo, 72 Rectangular, 70
Place sink at origin, 171 Red circle, 40
Plane: RZ, RQ, 67 Registry, 60, 179
Plate, 20, 23 Relations between constants, 181
Point Repeated groups, 21
Delete, 59 Resample, 79
Undelete, 59 Resample data, 78, 79
Point header, 81 Results
Point number, 58 Compare, 83
Point on a mode, 85 Rigid, 35
Pointer, 13, 14, 57–59, 71, 134 Root verification
Move, 58 Energy, 76
Points on graph, 61 Velocity, 76
Points per layer, 68 Roots
Points through thickness:, 78 Searching, 127
Position on X, 85 Spurious, 39, 118
Position on Y, 85 Roughness, 26
Postscript File, 81 RQ plane, 67
Power, 64, 66, 69, 131 RZ plane, 67
Power flow, 66, 69, 75, 130
Power normalised, 62 Sample every, 79
Predefined equation, 75 Sampling
Preferences, 60, 81, 84 Number of points on curve, 79
Display, 60 Saving, 80
Exporting, 61 Search width, 44
Printing, 60 Second graph, 54
Print preview, 84 Hide, 54
Print setup, 84 Select axes, 53, 54
Printing, 84 Select curve, 58
Colour, 60 Select point, 58
Geometry notes, 60 Select types to show, 56
Line width, 60 Setup, 12
Orientation, 60 SH modes, 46, 135
Settings, 60 Shift, 59

Disperse version 2.0.20a


Index 206

Show Symmetric System, 22


All hidden modes, 56 Symmetry, 40
Constant distance, 67
Constant wavelength, 67 Take ratio, 76
Determinant, 42 Take value, 74, 76
Only modes, 56 Taper, 71
Undeformed mode shapes, 67 Text (All variables), 81
Show determinant, 14, 42 Text (As displayed), 81
Side by side, 61 Text (Stnd variables), 81
Simulated signal, 70 Thickness
Sin/Cos, 71 Top layer, 21
Sink at origin, 47, 107, 171 Thickness normalisation, 22
Sliding boundaries, 119 Thickness*value, 61
Slowness Time step
Group, 55 Mode shapes, 68
Phase, 55 Title header, 81
Solid Bar Tolerance, 44
Defining, 163 Toolbar, 59
Solution Tools, 56, 62
Special options, 46 Tools : Calculate function, 52, 55, 56, 66
Solution stability, 42, 45–47 Tools : Export, 80
Solution type, 46 Tools : Export Data, 18, 19
Split, 59 Tools : Labels, 84
Split one mode into two, 59, 148 Tools : Labels : Edit labels, 135
Spring interface, 34, 119 Tools : Labels : Edit mode names, 39, 58
Infinite stiffness, 122 Tools : Mode Shapes, 63
Top or bottom of structure, 122 Tools : Resample data, 79
Zero stiffness, 122 Tools : Show Bulk Velocities, 78
Spurious roots, 39, 118 Tools : Simulated signal, 70
Start, 13 Tools : Verify solution, 78
Start : Programs, 17, 18 Torsional modes, 47
Stoneley wave Total energy, 75
Example, 143 Total energy density, 65, 130
Mode shapes, 143 Trace : Automatic Tracing, 159
STOP, 42, 63 Trace : Automatic tracing, 17, 18, 37, 134, 164
Strain energy, 75 Trace : Convergence parameters, 38, 41, 43, 64,
Strain energy density, 65, 130 148, 152
Stress equations Trace : Edit Geometry, 20
Cartesian, 97 Trace : Solution type, 22, 46, 138, 164, 171
Stress mode shapes, 62 Trace : Stop all jobs, 39
Structure is embedded, 21, 23 Trace : Stop Jobs, 42
Structure is in vacuum, 21, 23 Trace : Sweep, 40, 143, 145
Subtraction, 76 Trace : Trace from Selected Point, 41
Sum, 76 Trace a function of Frequency, 49
Sweep, 164 Trace from, 13, 14, 145
Plot results, 56 Trace from point, 17, 18, 41, 149
Sweep and Search, 14, 19, 40, 145 Trace from selected, 39
Value of attenuation, 40 Trace from selected point, 134
Sweeping, 40 Trace from swept point, 145
Convergence, 42 Trace limits, 38, 41, 45
Leaky systems, 40 Tracing, 36, 41
Mouse, 40 Automatic, 37
Symmetric, 40 Naming, 39
Symmetric boundary, 35 Convergence, 42

Disperse version 2.0.20a


Index 207

Cylindrical, 164 xy xy xy, 82


Logic, 36 xyz xyz, 82
Manual, 39 xyz...., 82
New system, 20
Stopping, 42 Y, 85
Sweeping, 40 y y y y, 82
Transversely anisotropic, 28 Young’s modulus
Triangular, 70 Example, 149
Trigonal, 28
Zoom, 14, 57, 58, 71
Undefined, 24 Autoscale, 57
Undeformed, 67 Box, 57
Undelete Custom, 57
Curve, 59 Move, 58
Point, 59 Unzoom, 58
Units, 65
Use frequency chirp, 72
Use frequency modulation, 72
User, 35
User variable, 48

Value variation is automatic, 25


Values are constant, 25
Values change automatically with frequency, 26
Values depend on frequency, 25, 26
Values depend on user variable, 25, 26
Values do not vary with frequency, 26
Velocity
Bulk, 78, 191
Default units, 60
Velocity affects long. atten., 24
Verification tolerance (%), 79
Verify selected mode (10 points), 78
Verify selected mode (all points), 78
Verify selected point, 78
Verify solution, 78
Version 2.0, 14
Visco-elastic, 25
Example, 147
Viscosity, 24
Dynamic, 93
Kinematic, 93
Vl, 55
Vs, 55

Water, 55
Wave number
Matching at layer interfaces, 173
Wavelength, 75
Wavenumber, 79
Wavenumber-thickness, 54
Window, 72
Windows 3.11 Metafile, 81

X, 85

Disperse version 2.0.20a

Das könnte Ihnen auch gefallen