You are on page 1of 267

CIP-GEGEVENS KONINKLIJKE BIBLIOTHEEK, DEN HAAG

Daalen, Edwin Frank George van


Numerical and theoretical studies of water waves and oating bodies /
Edwin Frank George van Daalen. - [S.l. : s.n.]. - Ill., g., tab.
Proefschrift Enschede. - Met index, lit. opg. - Met samenvatting in
het Nederlands.
ISBN 90-9005656-4
Trefw.: hydrodynamica / integraalvergelijkingen / variatierekening
Numerical and Theoretical Studies of
Water Waves and Floating Bodies
Proefschrift
ter verkrijging van
de graad van doctor aan de Universiteit Twente,
op gezag van de rector magnicus
prof.dr. Th.J.A. Popma,
volgens besluit van het College van Dekanen
in het openbaar te verdedigen
op vrijdag 8 januari 1993 te 16.45 uur
door
Edwin Frank George van Daalen
geboren op 13 juni 1965 te Amsterdam
Dit proefschrift is goedgekeurd door de promotor:
prof.dr.ir. P.J. Zandbergen
(Faculteit der Toegepaste Wiskunde)
en de overige leden van de promotiecommissie:
prof.dr.ir. E.W.C. van Groesen
(Faculteit der Toegepaste Wiskunde)
prof.dr.ir. L. van Wijngaarden
(Faculteit der Technische Natuurkunde)
prof.dr.ir. A.J. Hermans
(Technische Universiteit Delft)
prof.dr.-ing. O. Mahrenholtz
(Technische Universit at Hamburg-Harburg)
to Sonja
Contents
1 Introduction 1
1.1 Historical background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.2 The investigations: past and present . . . . . . . . . . . . . . . . . . . . . 3
1.3 Dissertation outline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.4 Suggested references . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.5 Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
I Mathematical Formulation and Numerical Algorithm 11
2 Mathematical Statement of the Problem 13
2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.2 Force mechanisms in uid ow . . . . . . . . . . . . . . . . . . . . . . . . 14
2.3 Potential ow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.4 Boundary conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
2.4.1 Impermeable xed boundaries . . . . . . . . . . . . . . . . . . . . . 19
2.4.2 Impermeable moving boundaries . . . . . . . . . . . . . . . . . . . 20
2.4.3 Free boundaries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
2.4.4 Open boundaries . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
2.5 Wave-body formulations . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.5.1 Ship wave-making and wave-resistance . . . . . . . . . . . . . . . . 23
2.5.2 Linearized oscillatory motion . . . . . . . . . . . . . . . . . . . . . 24
2.5.3 Forward speed radiation-diraction . . . . . . . . . . . . . . . . . . 26
2.5.4 Nonlinear motion at zero forward speed . . . . . . . . . . . . . . . 28
2.6 Problem denitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
2.7 Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
3 Boundary Integral Equation Formulations 35
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
3.2 Integral equations in potential theory . . . . . . . . . . . . . . . . . . . . . 37
3.3 Well-posedness: existence and uniqueness . . . . . . . . . . . . . . . . . . 41
3.4 Choice of integral equation method . . . . . . . . . . . . . . . . . . . . . . 43
3.5 Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
v
4 Algorithm for Wave-Body Simulations 49
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
4.2 Review of methods for nonlinear ship motions . . . . . . . . . . . . . . . . 50
4.3 Basic algorithm for nonlinear water waves . . . . . . . . . . . . . . . . . . 52
4.4 Extension to nonlinear ship motions . . . . . . . . . . . . . . . . . . . . . 55
4.5 Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
II Numerical Results 65
5 Impulsive Wavemaker Motion 67
5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
5.2 Nonlinear analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
5.2.1 Governing equations and small time expansions . . . . . . . . . . . 70
5.2.2 Leading order solutions . . . . . . . . . . . . . . . . . . . . . . . . 71
5.2.3 Initial and small time behaviour . . . . . . . . . . . . . . . . . . . 73
5.3 Numerical results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
5.3.1 Initial behaviour on the wavemaker . . . . . . . . . . . . . . . . . . 77
5.3.2 Initial behaviour on the bottom . . . . . . . . . . . . . . . . . . . . 77
5.3.3 Initial behaviour on the free surface . . . . . . . . . . . . . . . . . 78
5.3.4 Free surface elevation for small time . . . . . . . . . . . . . . . . . 78
5.4 Concluding remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
5.5 Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
6 Hydrodynamic Mass and Damping 95
6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
6.2 Mathematical model for two-dimensional motion . . . . . . . . . . . . . . 97
6.3 Numerical results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
6.3.1 Circular cylinder in heaving motion . . . . . . . . . . . . . . . . . 100
6.3.2 Circular cylinder in swaying motion . . . . . . . . . . . . . . . . . 103
6.4 Concluding remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
6.5 Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
7 Cylinders in Free Motion 113
7.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
7.2 Circular heaving cylinder . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
7.3 Rectangular heaving cylinder . . . . . . . . . . . . . . . . . . . . . . . . . 117
7.4 Rectangular rolling cylinder . . . . . . . . . . . . . . . . . . . . . . . . . . 118
7.5 Concluding remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
7.6 Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
III Variational Principles and Hamiltonian Formulations 129
8 Lagrangian and Hamiltonian Formulations 131
8.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132
8.2 Lukes variation principle for water waves . . . . . . . . . . . . . . . . . . 132
8.3 Zakharov & Broers Hamiltonian formulation . . . . . . . . . . . . . . . . 136
8.4 Wave-body formulations . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
8.4.1 A variation principle for wave-body interactions . . . . . . . . . . . 142
8.4.2 A Hamiltonian formulation for wave-body interactions . . . . . . . 146
8.5 Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 152
9 Symmetries and Conservation Laws 157
9.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 158
9.2 Conserved densities for water waves . . . . . . . . . . . . . . . . . . . . . 159
9.3 Conservation laws for the wave-body problem . . . . . . . . . . . . . . . . 164
9.3.1 Invariants for the two-dimensional case . . . . . . . . . . . . . . . . 164
9.3.2 Invariants for the three-dimensional case . . . . . . . . . . . . . . . 171
9.4 The virial and conservation law no.8 . . . . . . . . . . . . . . . . . . . . . 175
9.4.1 The virial connection . . . . . . . . . . . . . . . . . . . . . . . . . . 175
9.4.2 The circulation alternative . . . . . . . . . . . . . . . . . . . . . . 179
9.4.3 Fluid-lled deformable bodies . . . . . . . . . . . . . . . . . . . . . 180
9.4.4 The broken symmetry argument . . . . . . . . . . . . . . . . . . . 181
9.5 Symmetry-breaking boundaries . . . . . . . . . . . . . . . . . . . . . . . . 182
9.6 Numerical validation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 185
9.7 Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 190
10 Radiation Boundary Conditions 193
10.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193
10.2 Theory for continuous systems . . . . . . . . . . . . . . . . . . . . . . . . 194
10.3 Application to one-dimensional wave equations . . . . . . . . . . . . . . . 202
10.4 Application to the water-wave problem . . . . . . . . . . . . . . . . . . . . 218
10.4.1 The Lagrangian-Hamiltonian approach . . . . . . . . . . . . . . . . 219
10.4.2 The Eulerian approach . . . . . . . . . . . . . . . . . . . . . . . . . 226
10.5 Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 230
11 Conclusions and Recommendations 233
IV Appendices 235
A An Expression for
tn
on the Body 237
B A Body Surface Integral Condition 243
C The Angular Body Momentum 247
Preface
This thesis was written within the framework of a research project entitled
Further development of a three-dimensional model for nonlinear surface waves,
with respect to the interaction with oating objects and the inuence of bound-
aries.
As already indicated by the title, one of the main objectives was the extension of an
existing method developed by Romate
1
to an algorithm for the numerical simulation
of nonlinear free surface waves in hydrodynamic interaction with oating bodies; this
particular part of the investigations is described here.
In the development of the numerical model and the computer code, many problems
of both theoretical and practical nature were encountered. From time to time the daily
search for engineering solutions was interrupted by periods in which these problems
were considered from a more fundamental point of view. This theoretical research also
recorded here resulted in some basic changes in the numerical model. In addition, a
number of theories concerning water waves and oating bodies was developed.
The investigations were nancially supported by the Netherlands Technology Foundation
(STW) a subdivision of the Dutch Foundation for Scientic Research (NWO) under
grants TWI88.1460 and TWI80.1460.
Most of the computations were done on supercomputers; from 1989 to 1990 on a
CRAY-XMP, and from 1991 on a CRAY-YMP. Computation time was granted by the
Dutch Foundation for Supercomputer Facilities (NCF). Technical support from the Stich-
ting Academisch Rekencentrum Amsterdam (SARA) is acknowledged.
A committee of future users of the computer code accompanied the project. The commit-
tee members were from the Maritime Research Institute Netherlands (MARIN), Delft
Hydraulics, the Royal Dutch Shell Laboratory Amsterdam (KSLA), the University of
Twente, and the Netherlands Technology Foundation.
I would like to thank the board of directors of MARIN for the unique opportunity to do
this work in such an inspiring research institute.
Thanks go to Hans P. van der Kam (Automation and Instrumentation Department)
and Hans Zeller (Audio-Visual Department) for their enthusiasm and assistance in visu-
alization and video-editing, and to Henk Luisman (Oshore Research Department) for
making the illustrations.
1
Romate, J.E. 1989. The Numerical Simulation of Nonlinear Gravity Waves in Three Dimensions
using a Higher Order Panel Method. Ph.D. thesis, University of Twente. Enschede, The Netherlands.
ix
Special thanks go to Ir. Rene H.M. Huijsmans (Oshore Research Department) and
Ir. Hoyte C. Raven (Ship Research Department) for introducing me to the eld of ship
hydrodynamics.
Also I want to thank Dr. Huib J. de Vriend (Delft Hydraulics) for his interest and
participation in this project.
Very special thanks go to Ir. Jan Broeze for his good-fellowship during the past four
years, and for our fruitful cooperation which made our investigations so successful.
Next, I would like to thank Dr. Johan E. Romate (KSLA) for his invaluable contributions
to this research, both in the past and in the present.
Furthermore, I would like to thank some members of the sta of the Faculty of Applied
Mathematics of the University of Twente.
I am grateful to Dr. Douwe Dijkstra for his helpful comments in numerical matters.
I feel very much indebted to professor van Groesen for his interest and active partic-
ipation in this research; our frequent discussions culminated in four joint papers, which
in turn formed a sound basis to part III of this thesis.
It is a great pleasure to express my deep gratitude to professor Zandbergen for his crit-
icism, advice, and encouragement; it has been a privilege to do this research under his
supervision.
I thank my parents and my brother for their stimulation in my study and work.
Finally, I would like to express my sincere love to my wife Sonja; in the end, her enormous
support and endless patience have helped me to complete this work.
Deventer, The Netherlands Ed van Daalen
January 8th, 1993
The Argonautica
A narrow escape: the Argo passes through the Symplegades.
Turning over the pages of this thesis, the interested reader inevitably will encounter sev-
eral passages from the Argonautica, an ancient Greek tale written by Apollonius Rhodius.
To those who are familiar with the mathematics and physics described here, each selected
fragment may come as a pleasant break while reading consecutive chapters. Perhaps even
more important is the fact that this dissertation, despite its technical nature, still oers
something worthwhile to read to those who are not familiar with exact sciences.
Several motives induced me to lard this thesis with a selection of passages from the
Argonautica. First of all, the tale of the Argonauts has a pronounced nautical character.
In this sense, the Argonautica is in close harmony with the investigations described here.
The second reason is that ever since my rst lesson in classical languages I have felt
a natural interest in Greek and Roman literature. This particular part of the literary
classics is full of fetching parables and moralizing legends; immortal gods and mortal
heroes are involved in an eternal struggle for the sympathy of the reader. Finally, in
my opinion the Argonautica has unjustly been eclipsed by Homers well-known Iliad and
Odyssee; this is the third motive for this personal mixture of literature and exact sciences.
xi
Cited from the jacket of the Loeb Classical Library publication, with an English transla-
tion by R.C. Seaton:
Apollonius of Rhodes was a Greek grammarian and epic poet of Alexandria
in Egypt and lived late in the third century and early in the second century
b.c. While still young he composed his extant epic poem of four books on the
story of the Argonauts. When this work failed to win acceptance he went to
Rhodes where he not only did well as a rhetorician but also made a success of
his epic in a revised form, for which the Rhodians gave him the freedom of
their city; hence his surname. On returning to Alexandria he recited his poem
again, with applause. In 196 b.c. Ptolemy Epiphanes made him the librarian
of the Museum (the University) at Alexandria. His Argonautica is one of the
better minor epics, remarkable for originality, powers of observation, sincere
feeling, and depiction of romantic love. His Jason and Medea are natural and
interesting, and did much to inspire Virgil (in a very dierent setting) in the
fourth book of the Aeneid.
The motive of the voyage of the band of Greek heroes (named Argonauts after their ship,
the Argo) is the command of Pelias, king of Iolcus, to bring back the golden eece from
Colchis. This command is based on Pelias desire to destroy the hero Jason, while the
divine aid given to Jason results from the intention of the goddess Hera to punish Pelias
for his neglect of the honour due to her.
The rst and second books describe the history of the voyage to Colchis, where
Apollonius interweaves with his narrative legends, accounts of local customs. This part
of the Argonautica is lled with many exciting episodes, such as the rape of Hylas, the
boxing match between Polydeuces and Amycus, the prophecies and counsels of Phineus,
and the passing through the Symplegades. The third book is occupied for the greater
part by the episode of the love of Jason and Medea
2
, and the accomplishment of Jasons
task with her aid. The fourth book, describing the return voyage, is invaluable for its
amazing geography, such as the supposed junction of the Rhine, Rhone, and Po rivers,
the Libyan desert, and the so-called Tritonian Lake. Each book has its own specic topic
and character, and the unity of the legend is that of the voyage itself.
May it please the reader!
2
Medea was a daughter of the king of Colchis.
Chapter 1
Introduction
Beginning with thee, O Phoebus
1
, I will recount the famous deeds of men
of old, who, at the behest of King Pelias, down through the mouth of Pontus
2
and between the Cyanean rocks
3
, sped well-benched Argo in quest of the golden
eece.
Such was the oracle that Pelias heard, that a hateful doom awaited him
to be slain at the prompting of the man whom he could see coming forth
from the people with but one sandal. And no long time after, in accordance
with that true report, Jason crossed the stream of wintry Anaurus on foot,
and saved one sandal from the mire, but the other he left in the depths held
back by the ood. And straightway he came to Pelias to share the banquet
which the king was oering to his father Poseidon and the rest of the gods,
though he paid no honour to Pelasgian Hera. Quickly the king saw him and
pondered, and devised for him the toil of a troublous voyage, in order that on
the sea or among strangers he might lose his home-return.
The ship, as former bards relate, Argus wrought by the guidance of Athena.
But now I will tell the lineage and the names of the heroes, and of the long
sea-paths and the deeds they wrought in their wanderings; may the Muses be
the inspirers of my song!
Argonautica, Book I, Verses 1-22.
1
i.e. Apollo
2
i.e. the Black Sea
3
i.e. the Symplegades
1
2 CHAPTER 1. INTRODUCTION
1.1 Historical background
Over the past three centuries, surface wave problems have interested a considerable
number of mathematicians, beginning in the early eighteenth century with Euler and the
Bernoullis in Switzerland, and continuing in the late eighteenth and the early nineteenth
century with Lagrange, Cauchy, Navier, and Poisson in France. Later the British school
of mathematical physicists paid attention to these problems, and notable contributions
were made by Airy, Stokes, Rayleigh, Kelvin, Michell, and Lamb. In the latter part of the
nineteenth century the French once more took up the subject; the work done by Boussi-
nesq in this eld is historic, as well as the Dutch contributions from Korteweg and de
Vries. Later, Poincare made excellent contributions with regard to gures of equilibrium
of rotating and gravitating liquids. One of the most outstanding accomplishments from
the purely mathematical point of view the proof of the existence of progressing waves
of nite amplitude was made by Levi-Civita; the extension of this proof to waves in
a canal of nite depth was accomplished by Struik.
4
The subject of surface gravity waves covers a wide range, whether regarded from the
viewpoint of physical problems which occur, or from the point of view of the mathemat-
ical ideas and methods to solve these problems. The physical problems vary from wave
motion over sloping beaches to ood waves in rivers, the motion of ships in a sea-way,
free oscillations of enclosed bodies of water such as lakes and harbors, to mention just
a few. The mathematical tools employed comprise the whole of the methods developed
in the classical linear mathematical physics concerned with partial dierential equations,
as well as a good part of what has been learned about the nonlinear problems of math-
ematical physics. Thus potential theory and the theories of linear and nonlinear wave
equations, together with tools such as conformal mapping and complex variable methods
in general, the Laplace and Fourier transform techniques, methods employing a Greens
function, integral equations, etc. are used. The nonlinear problems are of both elliptic
and hyperbolic type.
Nowadays, the evolution of nonlinear gravity driven water waves interacting with xed
or freely oating objects is an important eld of research in ocean engineering. For large
objects with characteristic dimensions of the order of the wave length, viscous eects
of the uid ow can be neglected, as well as eects of compressibility and surface ten-
sion. This is the so-called diraction regime of uid-structure interaction. Under the
additional assumption of irrotational uid ow, a velocity potential can be introduced to
describe the ow characteristics. This potential satises a linear eld equation, namely
Laplaces equation, by continuity. The free surface boundary conditions render the prob-
lem nonlinear.
In many cases it is sucient to linearize the free surface conditions and solve the linear
problem. However, this applies only to waves with small amplitude compared with the
wavelength and the mean water depth. For steep waves this linearization procedure
can not be justied and other techniques must be developed. In this case the mutual
interaction of waves due to the nonlinearity can be very strong; this may result, for
instance, in the overturning of waves. If a oating body is involved, nonlinear eects
may have strong inuence on the wave evolution and the body motion; in such a case a
linearized approximation would provide inadequate results. In many of these problems
higher order approximations to the nonlinear equations are also inadequate. Typical
4
Of course this historical survey is incomplete; for the classical publications mentioned here the
interested reader is referred to the bibliography at the end of this chapter.
1.2. THE INVESTIGATIONS: PAST AND PRESENT 3
examples where approximations to the nonlinear equations give unsatisfactory results
are:
wave slamming on xed and oating structures,
wave run-up on slopes,
problems in which a part of the structure is submerged, causing the waves on top
of it to break or nearly break.
Of course the assumptions of potential ow are not valid at all stages of the physical
process under consideration, but in such a case the potential solution can be very useful
as input for a more general model.
1.2 The investigations: past and present
The main objective of these investigations is the extension of a higher order panel method
for nonlinear gravity wave simulations to water waves in hydrodynamic interaction with
oating bodies.
The major diculty in solving the nonlinear potential model for this particular prob-
lem is the presence of a moving free surface and a freely oating body. Due to the
time dependent free surface conditions both the potential at, and the position of the
free surface are unknown variables, which have to be determined as part of the solution.
Similarly, due to the equations of motion for the freely oating body, the potential on
the wetted body surface and the body position and orientation are unknown; these vari-
ables have to be determined as part of the solution too. The diculty in solving the
nonlinear potential model is also indicated by the fact that this particular problem has
both elliptic and hyperbolic properties, through the eld equation and the free surface
conditions respectively.
In the past decades numerous attempts have been made to obtain analytical solu-
tions for general free surface wave problems. Substitution of perturbation series for the
variables into the governing equations and the use of expansions to sometimes very high
order provided many nonlinear approximations. However, analytical solutions of the fully
nonlinear equations have not been found so far. Therefore, it seems that the development
of numerical methods is a more promising way towards the solution of these problems.
Romate (1989) developed a very ecient boundary element method for three-dimensional
nonlinear gravity wave simulations, using a Greens formulation for the velocity poten-
tial. Higher order approximations for the singularity distributions and the geometry,
combined with a robust time stepping scheme, ensured the accuracy and stability of the
algorithm. By the end of 1988, his panel method gave excellent results for linear and
weakly nonlinear waves.
From 1989 on, manpower was doubled; Broeze and van Daalen continued Romates
research, aiming to t his method for highly nonlinear waves and wave interactions with
xed and freely oating objects. As a rst step towards these goals, a simplied panel
method for two-dimensional nonlinear gravity wave simulations was derived from Ro-
mates original method. With a number of basic modications this method was made
suitable for highly nonlinear waves and waves interacting with xed structures. Af-
ter a period of close cooperation, it was decided that Broeze focused his attention on
4 CHAPTER 1. INTRODUCTION
highly nonlinear waves in three dimensions. The extension of the panel method to two-
dimensional waves in hydrodynamic interaction with freely oating bodies was entrusted
to van Daalen; the results of these investigations cover the larger part of this thesis.
During the investigations many engineering problems were encountered, ranging from
grid motion control to the development of eective radiation boundary conditions. If it
was felt that these problems called for a more fundamental approach, a reasonable amount
of time was spent on gaining insight through theoretical considerations. Sometimes these
views induced signicant adaptations to the numerical algorithm; at other times, ideas
were profoundly worked out to novel theories. The ndings of these investigations are
embodied in the remaining part of this thesis.
1.3 Dissertation outline
This thesis comprises a number of numerical and theoretical studies of gravity driven
water waves and the interaction with xed structures and oating bodies. In spite of the
diversity of the material, it is not merely a collection of disconnected topics, lacking unity
and coherence. Considerable eort was made to supply the fundamental background
in hydrodynamics and also some of the mathematics needed and to plan this
dissertation such that a self-contained and readable whole was arrived at.
This work is split up into four main parts:
I. Mathematical Formulation and Numerical Algorithm (ch. 24)
II. Numerical Results (ch. 57)
III. Variational Principles and Hamiltonian Formulations (ch. 810)
IV. Appendices (AD)
Each part (excepting part IV) has been written such that to some extent it can be
read independently of the other parts. We have also striven to write separate chapters;
an outline will be given next.
Parts I and II reect the investigations with respect to the development of the math-
ematical model and the numerical (boundary element) algorithm for water waves and
oating bodies.
In chapter 2 the governing equations for the nonlinear water-wave problem, including
the interaction with xed objects and oating bodies, are presented. The transition from
this set of governing equations to a boundary integral equation formulation is described
in chapter 3. In chapter 4 the numerical algorithm for water waves and oating bodies
is outlined, with a short discussion of the discrete (boundary element) approximation of
the problem.
Numerical results obtained with our computer code TIPHYS
5
for an impulsively
started wavemaker are discussed in chapter 5. The numerically computed hydrodynamic
mass and damping coecients for two-dimensional cylinders in forced harmonic motion
have been compared to experimental data and analytical solutions; the ndings of this
study are reported in chapter 6. In chapter 7 another validation study for two-dimensional
cylinders but now in free motion is presented.
Part III is concerned with special descriptions of the hydrodynamic problems considered
5
TIPHYS: a time domain panel method for nonlinear gravity waves and oating bodies; named after
Tiphys, the helmsman of the Argo.
1.4. SUGGESTED REFERENCES 5
here. Chapter 8 reviews so-called variation principles and Hamiltonian formulations
for the classical water-wave problem; novel extensions of these theories to water waves
interacting with oating bodies are presented. Chapter 9 is devoted to invariants and
conservation laws for wave-body problems. A theory for radiation boundary conditions
for wave problems that are governed by a Lagrangian principle that conserve a
characteristic density (for instance, the energy density) is presented in chapter 10.
Concluding remarks and recommendations for future research are given in chapter 11.
Part IV consists of four appendices containing extensive derivations (Appendices A-C)
and a ow-chart of the TIPHYS-code (Appendix ??).
The bibliographical footnotes have been borrowed from the New Websters Dictionary
and Thesaurus of the English Language (1991 edition, Lexicon Publications), and from
C.B. Boyers A History of Mathematics (1985, Princeton University Press).
Finally, it is noted that the equations have not been made dimensionless, unless stated
otherwise; all variables are expressed in SI-units.
1.4 Suggested references
For a general introduction to water waves and wave-body interactions, the reader is re-
ferred to the many books and (review) articles on these subjects; a number of them will
be mentioned hereafter.
All of the basics of uid dynamics used in this thesis is covered by the works of Batchelor
(1967) and Milne-Thomson (1968). Valuable information regarding the eld of hydro-
dynamics is gathered in the historic work of Lamb (1932), of which almost a third is
concerned with surface gravity waves. A mathematically oriented treatment of water
waves is given by Stoker (1957). Another important source of information on wave the-
ories is the work of Wehausen and Laitone (1960). Wind-generated water waves are
discussed extensively by Kinsman (1965). Whitham (1974) discusses waves in a more
general context; his work includes a number of chapters on water waves and the use of
variational principles. An introduction to the science of wave motions in uids, with a
chapter devoted to water waves only, is given by Lighthill (1978). A very readable text on
the dynamics of ocean surface waves is due to Mei (1983). An introductory presentation
of the basic theories of water waves, using direct mathematical techniques, is given by
Crapper (1984).
The following reviews on various wave types should also be brought to the attention
of the reader: solitary waves, by Miles (1980); trapped waves, by Mysak (1980); wave
instabilities, by Yuen and Lake (1980); strongly nonlinear waves, by Schwartz and Fen-
ton (1982); breaking waves, by Peregrine (1983) and Battjes (1988); tsunamis (i.e. huge
waves caused by large submarine earthquakes or landslides), by Voit (1987). Numerical
methods in free surface ows are reviewed by Mei (1978) and Yeung (1982). The use
of boundary integral equation methods in inviscid uid mechanics is discussed by Hess
(1990). A survey of Hamiltonian formulations in uid mechanics is due to Salmon (1988).
Mathematical treatments of the motion of bodies in waves have been given by Landwe-
ber (1961) and Wehausen (1971). Newmans (1977) textbook on the hydrodynamics of
6 CHAPTER 1. INTRODUCTION
marine objects is recommendable. Wave-structure interactions have been dealt with by
Sarpkaya and Isaacson (1981); more recent contributions with regard to this subject are
due to Faltinsen (1990ab). An update of the theory of oating bodies since John (1949,
1950) is given by Kleinman (1982). Timman, Hermans, and Hsiao (1985) have written a
very readable textbook on water waves and ship hydrodynamics.
For a survey of the dierent numerical methods used in the solution of free surface
ow problems, with and without the presence of solid structures or oating bodies, the
proceedings of the International Conferences on Numerical Ship Hydrodynamics, the
proceedings of the Symposia on Naval Hydrodynamics, and the abstracts of the Inter-
national Workshops on Water Waves and Floating Bodies should be consulted. These
publications include almost any method in most areas concerning uid-structure inter-
action, such as potential ows, nonlinear waves, lifting bodies, vortex ows, cavitation,
ship wave-making and wave-resistance, propulsion, boundary layers, and viscous ows.
1.5 Bibliography
Abstracts of the International Workshops on Water Waves and Floating Bodies. 1986
. . .
Airy, Sir G.B. 1845. On tides and waves. Encyclopaedia Metropolitana, Series 5,
5:241-396.
Batchelor, G.K. 1967. An Introduction to Fluid Dynamics. Cambridge University
Press.
Battjes, J.A. 1988. Surf-zone dynamics. Annual Review of Fluid Mechanics 20:257-
293.
Bernoulli, D. 1738. Hydrodynamica. Argentorati.
Boussinesq, J. 1871. Theorie de lintumescence liquide, appelee onde solitaire ou de
translation, se propageant dans un canal rectangulaire. Institut de France, Academie des
Sciences, Comptes Rendus 72:755-759.
Boussinesq, J. 1877. Essai sur la theorie des eaux courantes. Memoires par divers
savants, Series 2, 23:1-680.
Cauchy, A.-L. 1816. Memoire sur la theorie des ondes. Memoires de lAcademie Royale
des Sciences.
Crapper, G.D. 1984. Introduction to Water Waves. John Wiley and Sons.
Euler, L. 1755. Principes generaux du mouvement des uides. Histoires de lAcademie
de Berlin.
Faltinsen, O.M. 1990a. Wave loads on oshore structures. Annual Review of Fluid
Mechanics 22:35-56.
Faltinsen, O.M. 1990b. Sea loads on ships and oshore structures. Cambridge Univer-
sity Press.
Hess, J.L. 1990. Panel methods in computational uid dynamics. Annual Review of
Fluid Mechanics 22:255-274.
John, F. 1949. On the motion of oating bodies I. Communications on Pure and Applied
1.5. BIBLIOGRAPHY 7
Mathematics 2:13-57.
John, F. 1950. On the motion of oating bodies II. Simple harmonic motions. Commu-
nications on Pure and Applied Mathematics 3:45-101.
Kelvin, Lord. 1886/1887. On stationary waves in owing water. Philosophical Maga-
zine, Series 5, 22:353-357, 445-452, 517-530, 23:52-57.
Kinsman, B. 1965. Wind Waves. Their Generation and Propagation on the Ocean
Surface. Prentice-Hall.
Kleinman, R.E. 1982. On the mathematical theory of the motion of oating bodies: an
update. David Taylor National Ship Research and Development Center, Report 82/074.
Bethesda, Maryland.
Korteweg, D.J., and de Vries, G. 1895. On the change of form of long waves ad-
vancing in a rectangular canal and on a new type of long stationary waves. Philosophical
Magazine, Series 5, 39:422-443.
Lagrange, J.L. 1781. Memoire sur la theorie du mouvement des uides. Nouvelles
memoires de lAcademie de Berlin.
Lamb, Sir H. 1904. On deep water waves. Proceedings of the London Mathematical
Society, Series 2, 2:388.
Lamb, Sir H. 1913. Some cases of wave motion on deep water. Annali di Matematica,
Series 3, 21:237.
Lamb, Sir H. 1932. Hydrodynamics. Cambridge University Press.
Landweber, L. 1961. Motion of immersed and oating bodies. In Handbook of Fluid
Dynamics. McGraw-Hill.
Levi-Civita, T. 1925. Determination rigoreuse des ondes permanentes dampleur nie.
Mathematische Annalen 93:264-314.
Lighthill, Sir M.J. 1978. Waves in Fluids. Cambridge University Press.
Mei, C.C. 1978. Numerical methods in water-wave diraction and radiation. Annual
Review of Fluid Mechanics 10:393-416.
Mei, C.C. 1983. The Applied Dynamics of Ocean Surface Waves. John Wiley and Sons.
Michell, J.H. 1893. The highest waves in water. Philosophical Magazine, Series 5,
56:430.
Miles, J.W. 1980. Solitary waves. Annual Review of Fluid Mechanics 12:11-43.
Milne-Thomson, L.M. 1968. Theoretical Hydrodynamics. MacMillan.
Mysak, L.A. 1980. Topographically trapped waves. Annual Review of Fluid Mechanics
12:45-76.
Navier, C.L.M.H. 1822. Memoire sur les lois du mouvement des uides. Memoires de
lAcademie des Sciences 6:389.
Newman, J.N. 1977. Marine Hydrodynamics. The MIT Press.
Peregrine, D.H. 1983. Breaking waves on beaches. Annual Review of Fluid Mechanics
15:149-178.
Poincar e, H. 1885. Sur lequilibre dune masse uide animee dun mouvement de ro-
8 CHAPTER 1. INTRODUCTION
tation. Acta Mathematica 7:259.
Poisson, S.D. 1816. Memoire sur la theorie des ondes. Memoires de lAcademie Royale
des Sciences.
Proceedings of the International Conferences on Numerical Ship Hydrodynamics. 1975
. . .
Proceedings of the Symposia on Naval Hydrodynamics. 1956. . .
Rayleigh, Lord. 1876. On periodical irrotational waves at the surface of deep water.
Philosophical Magazine, Series 5, 1:381-389.
Romate, J.E. 1989. The Numerical Simulation of Nonlinear Gravity Waves in Three
Dimensions using a Higher Order Panel Method. Ph.D. thesis, University of Twente.
Enschede, The Netherlands.
Sarpkaya, T., and Isaacson, M. de St. Q. 1981. Mechanics of Wave Forces on
Oshore Structures. Van Nostrand Reinhold.
Salmon, R. 1988. Hamiltonian uid mechanics. Annual Review of Fluid Mechanics
20:225-256.
Schwartz, L.W., and Fenton, J.D. 1982. Strongly nonlinear waves. Annual Review
of Fluid Mechanics 14:39-60.
Stoker, J.J. 1957. Water Waves. The Mathematical Theory with Applications. Inter-
science Publishers.
Stokes, Sir G.G. 1847. On the theory of oscillatory waves. Transactions of the Cam-
bridge Philosophical Society 8(4):441-455.
Struik, D.J. 1926. Determination rigoreuse des ondes irrotationelles periodiques dans
un canal `a profondeur nie. Mathematische Annalen 95:595-634.
Timman, R., Hermans, A.J., and Hsiao, G.C. 1985. Water Waves and Ship Hy-
drodynamics. An Introduction. Kluwer Academic Publishers.
Voit, S.S. 1987. Tsunamis. Annual Review of Fluid Mechanics 19:217-236.
Wehausen, J.V., and Laitone, E.V. 1960. Surface Waves. In Handbook of Physics,
Volume IX, pp.445-778. Springer-Verlag.
Wehausen, J.V. 1971. The motion of oating bodies. Annual Review of Fluid Mechan-
ics 3:237-268.
Whitham, G.B. 1974. Linear and Nonlinear Waves. John Wiley and Sons.
Yeung, R.W. 1982. Numerical methods in free surface ows. Annual Review of Fluid
Mechanics 14:395-442.
Yuen, H.C., and Lake, B.M. 1980. Instabilities of waves on deep water. Annual
Review of Fluid Mechanics 12:303-334.
1.5. BIBLIOGRAPHY 9
First then let us name Orpheus whom once Calliope bare, it is said, wedded
to Thracian Oeagrus, near the Pimpleian height. Men say that he by the music
of his songs charmed the stubborn rocks upon the mountains and the course
of rivers . . .
Tiphys, son of Hagnias, left the Siphaean people of the Thespians, well
skilled to foretell the rising wave on the broad sea, and well skilled to infer
from sun and star the stormy winds and the time for sailing. Tritonian Athena
herself urged him to join the band of chiefs, and he came among them a
welcome comrade. She herself too fashioned the swift ship; and with her
Argus, son of Arestor, wrought it by her counsels. Wherefore it proved the
most excellent of all ships that have made trial of the sea with oars . . .
Next to him came a scion of the race of divine Danaus, Nauplius. He
was the son of Clytonaeus son of Naubolus; Naubolus was son of Lernus;
Lernus we know was the son of Proetus son of Nauplius; and once Amymone
daughter of Danaus, wedded to Poseidon, bare Nauplius, who surpassed all
men in naval skill . . .
After them from Taenarus came Euphemus whom, most swift-footed of
men, Europe, daughter of mighty Tityos, bare to Poseidon. He was wont to
skim the swell of the grey sea, and wetted not his swift feet, but just dipping
the tips of his toes was borne on the watery path . . .
Yea, and two other sons of Poseidon came; one Erginus, who left the
citadel of glorious Miletus, the other proud Ancaeus, who left Parthenia, the
seat of Imbrasion Hera; both boasted their skill in sea-craft and in war . . .
So many then were the helpers who assembled to join the son of Aeson
6
.
All the chiefs the dwellers thereabout called Minyae, for the most and the
bravest avowed that they were sprung from the blood of the daughters of
Minyas; thus Jason himself was the son of Alcimede who was born of Clymene
the daughter of Minyas
7
. . .
Now when all things had been made ready by the thralls, all things that
fully-equipped ships are furnished withal when mens business leads them to
voyage across the sea, then the heroes took their way through the city to the
ship where it lay on the strand that men call Magnesian Pagasae . . .
Argonautica, Book I, Fragments from Verses 23-238.
6
i.e. Jason
7
Minyas was a son of Aeolus, who was a son of Zeus; hence, Jason was a descendant of the supreme
god.
10 CHAPTER 1. INTRODUCTION
Part I
Mathematical Formulation and
Numerical Algorithm
11
Chapter 2
Mathematical Statement of the
Problem
. . . And they heaped their garments, one upon the other, on a smooth stone,
which the sea did not strike with its waves, but the stormy surge had cleansed
it long before. First of all, by the command of Argus, they strongly girded the
ship with a rope well twisted within, stretching it tight on each side, in order
that the planks might be well compacted by the bolts and might withstand the
opposing force of the surge. And they quickly dug a trench as wide as the
space the ship covered, and at the prow as far into the sea as it would run
when drawn down by their hands. And they ever dug deeper in front of the
stem, and in the furrow laid polished rollers; and inclined the ship down upon
the rst rollers, that so she might glide and be borne on by them. And above,
on both sides, reversing the oars, they fastened them round the thole-pins, so
as to project a cubits space. And the heroes themselves stood on both sides at
the oars in a row, and pushed forward with chest and hand at once. And then
Tiphys leapt on board to urge the youths to push at the right moment; and
calling on them he shouted loudly; and they at once, leaning with all their
strength, with one push started the ship from her place, and strained with
their feet, forcing her onward; and Pelian Argo followed swiftly; and they on
each side shouted as they rushed on. And then the rollers groaned under the
sturdy keel as they were chafed, and round them rose up a dark smoke owing
to the weight, and she glided into the sea; but the heroes stood there and kept
dragging her back as she sped onward. And round the thole-pins they tted
the oars, and in the ship they placed the mast and the well-made sails and the
stores.
Argonautica, Book I, Verses 364-393.
13
14 CHAPTER 2. MATHEMATICAL STATEMENT OF THE PROBLEM
2.1 Introduction
In this chapter the complete set of governing equations for the motion of an ideal uid
with a free surface is presented. These equations include the eld equation for the interior
uid ow and the conditions on all physical and articial boundaries. The additional
equations describing the hydrodynamic interaction with oating bodies, either partially
or totally submerged, are also presented. These equations comprise the appropriate
boundary conditions on the body surface and the equations of motion for the body.
In order to give full account for assumptions that would have been made tacitly other-
wise, the present derivation of the governing equations is rather extensive. In section 2.2
we discuss assumptions with respect to various types of forces which are active in uid
ow. Then, the set of equations describing potential ow is derived in section 2.3. In
section 2.4 the conditions for several types of boundaries are presented. The governing
equations for a number of formulations describing three-dimensional wave-body interac-
tions are discussed in section 2.5. Finally, formal denitions of the particular water-wave
and wave-body problems treated in this thesis are given in section 2.6.
2.2 Force mechanisms in uid ow
The motion of a uid, like the motion of rigid bodies, is governed by opposing actions of
dierent forces. In uid dynamics, these forces are distributed continuously throughout
a volume lled with innitesimal uid particles. One may distinguish force mechanisms
associated with the uid inertia, the uid weight, the uid viscosity, and other (sec-
ondary) eects such as surface tension.
1
In order to analyze the three principal force
mechanisms inertial, gravitational, and viscous it is useful to estimate the orders of
magnitude. Suppose that the problem under consideration is characterized by a physical
length L, a velocity U, a uid density , a gravitational acceleration g, and a dynamic
uid viscosity . An order estimation of the force magnitudes is given in the table below.
Type of Force Symbol Order of Magnitude
Inertial F
i
U
2
L
2
Gravitational F
g
gL
3
Viscous F
v
UL
Table 2.1: Order estimation of force magnitudes.
1
In section 8.2 we show that surface tension eects are negligible for all waves but the shortest or
ripple waves.
2.3. POTENTIAL FLOW 15
These order estimates merely indicate how changes in any of the physical parameters L,
U, , g, or aect the balance of the various force mechanisms. For instance, doubling
the length scale L corresponds to multiplicative factors of 2
2
, 2
3
, and 2
1
for the inertial,
gravitational, and viscous forces respectively. The above considerations are useful not
only in predicting full-scale phenomena from tests with a scale model, but they also
indicate which eects can be neglected in the mathematical model of the problem under
consideration.
In order to gain insight in the relative magnitudes of the various forces, three nondi-
mensional parameters are introduced to describe the uid ow:
F
i
F
g
=
U
2
L
2
gL
3
= U
2
/gL , (2.1)
F
i
F
v
=
U
2
L
2
UL
= UL/ , (2.2)
F
g
F
v
=
gL
3
UL
= gL
2
/U . (2.3)
From any two of these ratios the third can be calculated. Hence, any pair of these ratios
determines the balance of forces in the uid motion. Usually, the rst two are used to
dene the characteristic numbers
Fr =
U
(gL)
1/2
(Froude number) , (2.4)
Re =
UL

=
UL

(Reynolds number) , (2.5)


where = / is the kinematic uid viscosity; common values of are 10
6
m
2
/s for
water and 1.5 10
5
m
2
/s for air. For typical values of the characteristic velocity and
the length scale, say U = 1 m/s and L = 10 m, this implies that the Reynolds number
Re will be large, and hence that viscous forces will be small compared to inertial forces.
Therefore, eects of viscosity can be neglected for the bulk of the uid. However, in
special regions such as the boundary layer
2
very close to a body viscosity should
be included.
2.3 Potential ow
Assuming the laws of classical mechanics and thermodynamics to apply, the motion of
a uid can be described by a set of partial dierential equations expressing conservation
of mass, conservation of momentum, and conservation of energy per unit volume of the
uid. If necessary, this set of equations is complemented with an equation of state.
2
For a detailed discussion of ship boundary layers we refer to the review article written by Landweber
and Patel (1979).
16 CHAPTER 2. MATHEMATICAL STATEMENT OF THE PROBLEM
Let e
1
, e
2
, e
3
denote a Cartesian
3
coordinate system xed in space, with corresponding
spatial coordinates (x, y, z), and let v = (u, v, w)
T
denote the uid velocity eld. Follow-
ing the motion of an innitesimal control volume V , the equation of mass conservation
or continuity equation reads
D
Dt
[V ] =
D
Dt
V +
DV
Dt
= 0 , (2.6)
where D/Dt denotes the material derivative (i.e. the substantial derivative or the total
time derivative) and = (x, y, z; t) is the local time dependent uid density.
The control volume is subject to changes due to strain only; hence, in terms of the
velocity eld the material derivative of the control volume is given by
DV
Dt
=

u
x
+
v
y
+
w
z

V = ( v) V , (2.7)
where represents the three-dimensional gradient-operator:
=


x
,

y
,

z

T
. (2.8)
With (2.7) the equation of mass conservation (2.6) can be simplied to
D
Dt
+ ( v) = 0 . (2.9)
Then, assuming the uid to be absolutely incompressible and homogeneous, we have
= constant throughout the uid domain, and the continuity equation reduces to
v = 0 or divv = 0 , (2.10)
expressing that the velocity eld is free of strain (or divergence-free).
Next, for the control volume V , the equation of momentum conservation reads
D
Dt
[v ] =

F
i
+

F
g
+

F
v
, (2.11)
where we have dropped the incompressibility assumption. This equation expresses the
change of momentum due to forces acting on the control volume. These forces are, in
general, inertial (pressure) forces (

F
i
), gravity forces (

F
g
), and frictional forces due to
viscosity (

F
v
). Substitution of explicit expressions for these dierent forces yields the
well-known Navier-Stokes equations:
D
Dt
u +
Du
Dt
=
p
x
+

2
u +
v
x

, (2.12)
D
Dt
v +
Dv
Dt
=
p
y
+

2
v +
v
y

, (2.13)
D
Dt
w +
Dw
Dt
=
p
z
+

2
w +
v
z

g , (2.14)
3
Descartes, Rene (1596-1650), French philosopher, physicist and mathematician. He founded the
science of analytical geometry (la Geometrie, 1637) and discovered the laws of geometric optics. In
Discours de la Methode (1637) he divests himself of all previously held beliefs, to rebuild on his own
basis of certitude, i.e. the fact of his self-conscious existence: dubito ergo cogito: cogito ergo sum (I
doubt, therefore I think: I think, therefore I am).
2.3. POTENTIAL FLOW 17
where p denotes the pressure, is the uniform dynamic uid viscosity, g is the gravita-
tional acceleration (acting downwards along the z-axis), and
2
is the Laplace-operator:
4

2
= =


2
x
2
+

2
y
2
+

2
z
2

. (2.15)
Under the assumption of incompressibility, the above set of equations reduces to the
Navier-Stokes equations for incompressible uid ow:
Du
Dt
=
u
t
+v u =
1

p
x
+
2
u , (2.16)
Dv
Dt
=
v
t
+v v =
1

p
y
+
2
v , (2.17)
Dw
Dt
=
w
t
+v w =
1

p
z
+
2
w g , (2.18)
where = / denotes the kinematic viscosity of the uid. If = 0, these equations
simplify to the Euler
5
equations for incompressible and inviscid uid ow:
Dv
Dt
=
v
t
+ (v ) v =
1

p ge
3
. (2.19)
In general, a uid ow problem is characterized by a pressure p, a density , a tempera-
ture T, a viscosity , and a velocity v = (u, v, w). For an incompressible, homogeneous,
and inviscid uid with uniform temperature, and T are constant and equals zero,
and only four equations are needed to solve the ow problem. Summarizing, it can be
stated that the motion of an ideal and isothermal uid is governed by the continuity
equation (2.10) and the three Euler equations (2.19).
4
Laplace, Pierre Simon, Marquis de (1749-1827), French physicist and astronomer. With Lavoisier
he made the rst determination of the coecient of expansion of a metal rod, and initiated the study of
thermochemistry.
5
Euler, Leonhard (1707-1783), Swiss mathematician. He was responsible for the revision of nearly
all the branches of pure mathematics then known and the foundation of new methods of analysis. The
mathematical symbols , e, and i were introduced by Euler; his celebrated equality e
i
+1 = 0 combines
the ve most important numbers in mathematics in a single equation.
18 CHAPTER 2. MATHEMATICAL STATEMENT OF THE PROBLEM
Next, consider the uid vorticity
= v or (
x
,
y
,
z
) =

w
y

v
z
,
u
z

w
x
,
v
x

u
y

. (2.20)
Assuming that the uid ow is irrotational at some time t = t
0
, then it follows from the
curl of equation (2.19) that the ow will remain irrotational at all times:
=

0 for t t
0
. (2.21)
This conclusion is important because an irrotational vector eld can be represented as
the gradient of a scalar eld.
6
Thus we are allowed to introduce a potential for the
velocity eld:
v = . (2.22)
With (2.22) the continuity equation (2.10) becomes

2
= 0 (2.23)
i.e. Laplaces equation for the velocity potential .
With (2.22) the Euler equations (2.19) can be integrated to a general form of Bernoullis
7
equation:

t
+
1
2
( ) +gz =
p p
0

+B(t) (2.24)
where p
0
is a constant reference pressure and B(t) is an arbitrary function of time.
When, as in the cases which will be considered, the boundary conditions are kinematical,
the solution process consists in nding a harmonic potential satisfying (2.24) and the
prescribed boundary conditions. The pressure p is then indeterminate to the extent of
an additive function of t. It becomes determinate when the value of p at some point of
the uid is given for all values of t. Since the term B(t) has no inuence on resultant
pressures, it is frequently omitted.
8
Thus it has been shown that under the assumptions of an incompressible, homogeneous,
and inviscid (i.e. ideal) uid and in the absence of vorticity (i.e. an irrotational ow), the
motion of the uid is governed by Laplaces equation for the velocity potential and the
nonlinear Bernoulli equation. If the potential is known throughout the uid domain,
the velocity eld v and the pressure eld p are easily obtained from (2.22) and (2.24)
respectively. From these quantities the potential and kinetic energies can be determined,
due to the absence of friction.
6
This is a direct result of Helmholtz theorem in vector analysis, which states that any continuous
and nite vector eld can be expressed as the sum of the gradient of a scalar function and the curl of
a zero-divergence vector. The divergence-free vector vanishes if the original vector eld is irrotational.
The proof of this theorem can be found in Morse and Feshbach (1953).
7
Bernoulli, Swiss family of Dutch extraction, several members of which made distinguished contri-
butions to physics and mathematics. Jacques (1654-1705) worked on analytical geometry, his brother
Jean (1667-1748) discovered the exponential calculus and a method of integrating rational functions,
and Jeans son Daniel (1700-1782) developed the kinetic theory of gases.
8
Suppose, for instance, that a solid body moves through a uid completely enclosed by xed bound-
aries, and that it is possible say, by means of a piston to apply an arbitrary pressure at some point
of the boundary. Whatever variations are made in the magnitude of the force applied to the piston, the
motion of both the uid and the solid will be absolutely unaected, since at all points the pressure will
instantaneously rise or fall by equal amounts. Physically, the origin of this paradox is that the uid is
treated as absolutely incompressible. In actual uids changes of pressure propagate with very great, but
not innite, velocity.
2.4. BOUNDARY CONDITIONS 19
2.4 Boundary conditions
To complete the set of governing equations, initial conditions are needed, as well as
boundary conditions for the elliptic eld equation (2.23). The initial conditions depend
on the specic problem under consideration, and are incorporated in the nal problem
specications. Obviously, the initial conditions must be compatible with the governing
equations, in order to obtain a well-posed problem.
Figure 2.1: Fluid domain and bounding surfaces.
The boundary conditions depend on the type of boundary under consideration. In gen-
eral, we have a certain amount of uid occupying a simply connected
9
transient domain
(t), see Figure 2.1. The problem of gravity driven water waves introduces a free sur-
face F, which is one part of the domain boundary (t). Other parts are, for instance,
the bottom B and the hull of a ship S. These and other types of boundaries and the
corresponding conditions are discussed hereafter.
2.4.1 Impermeable xed boundaries
In most practical congurations, at least one xed boundary is present; the bottom
denoted by B in Figure 2.1 which is not necessarily even. If such a xed boundary is
impermeable, then uid particles can not penetrate it; hence, the normal component of
the uid velocity must vanish there:
v
n
= v n = 0 . (2.25)
9
A region such that any simple closed curve can be shrunk to a point without leaving the region is
called simply (or singly) connected.
20 CHAPTER 2. MATHEMATICAL STATEMENT OF THE PROBLEM
Here n is the unit normal vector along B, pointing outwards.
Under the assumptions of potential ow, this so-called zero-ux condition can be
reformulated to

n
= n = 0 (2.26)
stating that the normal derivative of the velocity potential vanishes.
2.4.2 Impermeable moving boundaries
The motion of uid particles in the neighbourhood of impermeable moving boundaries
such as a wavemaker or the hull of a ship is inuenced by the boundary velocity.
Eects of viscosity can be taken into account by introducing a boundary layer, wherein
stress forces inuence the uid motion. However, assuming the uid to be inviscid, the
motion of a uid particle on an impermeable moving boundary with velocity

V satises
v
n
= v n =

V n , (2.27)
expressing that the normal velocity of the uid particle and the normal boundary velocity
are equal.
In case of potential ow, this condition reads

n
= n =

V n (2.28)
Of course condition (2.26) for impermeable xed boundaries is a special case of condi-
tion (2.28).
2.4.3 Free boundaries
The free surface denoted by F in Figure 2.1 diers from other physical boundaries
since it is a free boundary indeed; not only the potential must be determined there,
but also the position of the free surface itself. Therefore, it is obvious that two boundary
conditions are needed for a proper mathematical description of the free surface.
In the absence of surface tension, the pressure at the free surface must equal the atmo-
spheric pressure p
0
; the rst condition is then obtained from Bernoullis equation (2.24)
by substitution of p = p
0
= 0 and by the choice B(t) = 0:

=

t
+
1
2
( ) +gz = 0 , (2.29)
where z = 0 corresponds to the mean water level. Condition (2.29) is known as the
dynamic free surface condition, since it has been deduced from a momentum equation,
i.e. a balance of forces.
In Lagrangian notation (2.29) reads
D
Dt
=
1
2
( ) gz (2.30)
The second condition concerns the free surface velocity; uid particles at the free sur-
face have the property that they remain part of the free surface until they encounter a
2.4. BOUNDARY CONDITIONS 21
boundary of another type. Following such a free surface particle, this implies that its
transient position x = x
F
is determined by the Lagrangian expression
Dx
F
Dt
= v = (2.31)
which is known as the kinematic free surface condition, since it is concerned with the
velocity eld only.
In many applications the free surface elevation can be assumed to be a single-valued
function of the horizontal coordinates and the time. In these cases the free surface is
described in terms of its elevation by
F (x, y, z; t) z (x, y; t) = 0 . (2.32)
The requirement that the material derivative of z vanishes on the free surface then
gives

t
+

x
u +

y
v = w , (2.33)
or, in terms of the elevation and the potential :

t
+

x

x
+

y

y
=

z
. (2.34)
Condition (2.31) is usually preferred to (2.34), since the latter condition excludes phe-
nomena as overturning waves.
For waves with small amplitude A compared to the wavelength and the mean water
depth h, the free surface conditions (2.29) and (2.34) can be linearized about the mean
water level:

t
= g

t
=
z

at z = 0 , (2.35)
where a sux denotes partial dierentiation, i.e.
t
= /t etc. The second approxi-
mate boundary condition states that the vertical velocity of a free surface particle coin-
cides with the vertical velocity of the free surface itself, thus ignoring the small horizontal
deviations.
In case of a horizontal bottom, the zero-ux condition (2.26) gives

z
= 0 at z = h . (2.36)
A two-dimensional periodic solution satisfying Laplaces equation (2.23), the linearized
free surface conditions (2.35), and the bottom condition (2.36) is given by the velocity
potential
(x, z; t) = A

k
cosh k (z +h)
sinh kh
cos (kx t) , (2.37)
and the corresponding free surface elevation
(x; t) = Asin (kx t) , (2.38)
where is the wave frequency and k = 2/ is the wave number.
22 CHAPTER 2. MATHEMATICAL STATEMENT OF THE PROBLEM
Substitution of (2.37-2.38) into (2.35) yields the rst order dispersion relation for
Stokes waves:

2
= gk tanh kh . (2.39)
For nite-amplitude waves on deep water we have kh and consequently
2
= gk.
Hence, for deep water waves the phase velocity c
f
and the group velocity c
g
are given by
c
f


k
= (g/k)
1/2
, c
g

d
dk
=
1
2
(g/k)
1/2
=
1
2
c
f
. (2.40)
For nite-amplitude waves on shallow water we have kh 0, and as a result
2
= ghk
2
.
It follows that for shallow water waves c
f
and c
g
are equal and independent of the wave
number k:
c
f
= c
g
= (gh)
1/2
. (2.41)
2.4.4 Open boundaries
In general, the uid domain extends to innity in the horizontal directions. In this
case a so-called radiation condition at innity is required to obtain a uniquely solvable
problem. This condition based on conservation of energy states that the waves
should behave at innity like progressing waves moving away from the source of the
disturbance. In problems concerning electromagnetic wave propagation this condition is
known as the Sommerfeld radiation condition, see Sommerfeld (1964) or Stoker (1957).
With regard to the water-wave problem, this radiation condition states that the
solution corresponds to outgoing waves only. However, for computational reasons (limited
cpu-time and computer memory) it will be necessary to truncate the uid domain at
some distance from the area of interest. The articial (i.e. non-physical) boundaries
thus introduced are part of the bounding surface . To obtain a well-posed problem,
appropriate conditions are needed on these boundaries.
There are two main criteria for a radiation condition on an articial boundary. First of all,
the condition should yield a well-posed problem. This boils down to the requirement that
the solution to the problem exists, is unique, and depends continuously upon the initial
and boundary conditions, i.e. well-posed in the sense of Hadamard (1923). Secondly,
the radiation condition should simulate the behaviour of the excluded domain as well as
possible; the solution should approximate the solution that one would have obtained if
the boundaries would have been chosen at innity. With regard to the free surface wave
problem, this comes down to the requirement that radiated surface waves approaching an
articial boundary should be fully transmitted or absorbed. In this sense, the radiation
condition should provide an open boundary.
Romate (1992) has reviewed methods for the numerical simulation of open boundaries
for linear and nonlinear water waves. On the basis of his literature survey, he decided
to use Higdons (1987) rst- and second-order partial dierential equations as absorbing
boundary conditions for the linearized model; the question of well-posedness was ad-
dressed too. The numerical implementation in his panel method for linear free surface
wave simulations and the stability of these rst- and second-order absorbing boundary
conditions have been discussed in a subsequent paper by Broeze and Romate (1992).
More recently, van Daalen, Broeze, and van Groesen (1992) introduced an analytical
technique for the development of radiation boundary conditions for general wave systems.
2.5. WAVE-BODY FORMULATIONS 23
At the outset of a variational principle for the wave problem under consideration, addi-
tional boundary conditions are derived such that a characteristic density (for instance,
the energy density) is conserved. In chapter 10 this theory is presented, and applications
to a number of wave problems including the three-dimensional nonlinear water-wave
problem are discussed.
2.5 Wave-body formulations
The presence of a oating body, either partly or totally submerged, implies that extra
equations are needed for a correct mathematical description of the uid-body interaction.
Usually, a second coordinate system e
1

, e
2

, e
3

is introduced, which moves with the


body, see Figure 2.2. The origin of this coordinate system is located at some xed point
inside the body for example, the centre of mass G and the unit vectors are chosen
along the body principal axes of inertia. The wetted (submerged) part of the body
surface is denoted by S.
Figure 2.2: Free surface with a oating body.
Generally, dierent assumptions on the motion of the body and the free surface lead to
dierent sets of governing equations. To illustrate this dependence, four frequently used
wave-body formulations are discussed next.
2.5.1 Ship wave-making and wave-resistance
Ever since Kelvin (1886) analyzed the potential waves generated by a pressure point
moving with constant velocity in otherwise calm water and Michell (1898) derived the
potential wave-resistance of a thin ship, much interest has been paid to theoretical (i.e.
analytical) ship wave-resistance calculations. With the development of fast computers
this problem has also been addressed numerically.
Consider a body moving at a constant speed in otherwise calm water. For this
particular system a stationary problem can be formulated, using a moving frame of
24 CHAPTER 2. MATHEMATICAL STATEMENT OF THE PROBLEM
reference xed to the body. Assuming that the uniform inow velocity is given by

V = (U, 0, 0)
T
, (2.42)
the following steady-state formulation is obtained:

2
= 0 in , (2.43)
1
2
( ) +gz =
1
2
U
2
on F , (2.44)

n
= 0 on F and S , (2.45)
=

V for r =

x
2
+y
2

1/2
. (2.46)
Condition (2.44) is the dynamic free surface condition, and is easily obtained by applica-
tion of Bernoullis theorem to a free surface streamline extending to innity in horizontal
direction. The kinematic condition (2.45) expresses that both the free surface ow and
the ow around the body are stationary.
Linearization of the free surface conditions about the mean water level z = 0 and
about the uniform ow =

V results in the so-called Neumann-Kelvin problem. This
problem is usually solved with integral equation techniques; in the Havelock-source (or
Kelvin-source) approach the Greens function satises all boundary conditions except
the hull boundary condition, while in the Rankine-source approach the Greens function
satises none of the boundary conditions.
Another solution procedure is the linearization about the ow at zero Froude number.
Successful computations for this so-called slow ship wave-making and wave-resistance
problem have been initiated by Gadd (1976) and Dawson (1977), followed by Sclavounos
and Nakos (1988), Raven (1988), and many others. In recent literature a shift towards
solution methods for the nonlinear problem is observed, see Ni (1987), Jensen, Bertram,
and Soding (1989), Campana, Lalli, and Bulgarelli (1989), and Raven (1992).
2.5.2 Linearized oscillatory motion
If the body motion is time-harmonic and non-translatory, the amplitude of the motion
can be used to linearize the governing equations. This has the important advantage that
the free surface grid is xed in time, thus simplifying a numerical solution procedure
considerably. Introducing the body frequency by
(x, y, z; t) = Re

(x, y, z) e
it

, (2.47)
V
n
(x, y, z; t) = Re

V
n
(x, y, z) e
it

, (2.48)
where the body velocity V
n
is imposed, the governing equations can be Fourier
10
trans-
formed in time to obtain a much simpler frequency domain problem:
10
Fourier, Jean Baptiste Joseph, Baron de (1768-1830), French physicist, mathematician and politician,
important for his theorem that any periodic function may be resolved into sine and cosine terms.
2.5. WAVE-BODY FORMULATIONS 25

= 0 in , (2.49)

2
g

z
on

F , (2.50)

n
=

V
n
on

S , (2.51)
where

F and

S denote the mean positions of the free surface and the wetted body
surface respectively. Condition (2.50) is obtained by substitution of (2.47-2.48) into the
(combined) linearized free surface conditions (2.35).
In order to render this problem uniquely solvable, Sommerfelds radiation condition
at innity must be imposed:
lim
r
(kr)
1/2

r
ik

= 0 , r =

x
2
+y
2

1/2
, (2.52)
where k is the wave number. Mei (1978) reviews two approaches towards the solution of
this so-called harmonic radiation problem: hybrid (coupled nite element and boundary
element) methods and integral equation methods.
The use of integral equation methods in the study of wave elds goes back to Lamb
(1932), who examined the scattering of linear waves by a surface piercing body. The
potential of the wave eld can be described in terms of a source distribution on the mean
wetted part of the body surface, where the Greens function satises Laplaces equation,
the linearized free surface conditions, and the radiation condition at innity. Greens
functions of this type are rather complicated and have been developed in the fties by
John (1950) and Stoker (1957); for a more recent discussion, see Noblesse (1983). The
corresponding integral equation methods have been employed by, for instance, Brown,
Eatock Taylor, and Patel (1983), and Nestegard and Sclavounos (1984) in the frequency
domain, and by Adachi and Ohmatsu (1980), Beck and Liapis (1987), and Newman
(1985) in the time domain.
Unfortunately, this method fails to give a unique solution at the so-called irregular
frequencies which are the eigenfrequencies of the interior Dirichlet
11
problem of the
body even though the solution of the original boundary value problem is unique.
Much eort has been spent to solve this problem; Sclavounos and Lee (1985) give a short
survey of methods for the removal of irregular frequencies. For further results on this
(non-physical) phenomenon, see Ursell (1981), Martin (1981), Hulme (1983), and Forbes
(1984). The question of well-posedness for this specic problem was addressed by John
(1950); more general existence and uniqueness proofs are given by Lenoir and Martin
(1981), and Simon and Ursell (1984).
Another way to treat this problem is to employ a simpler Greens function, which only
satises the boundary condition on the bottom. In this case singularity distributions
are needed on all other boundaries. Angell, Hsiao, and Kleinman (1986) show that
for a restricted class of three-dimensional body geometries this formulation has
no irregular frequencies, provided that the original boundary value problem is uniquely
solvable. Liu (1991) proved in a dierent setting that the same integral equation
for the two-dimensional problem does not suer from irregular frequencies either.
11
Dirichlet, Peter Gustav Lejeune (1805-1859), German mathematician who contributed to the theory
of numbers and to the establishment of Fouriers theorem.
26 CHAPTER 2. MATHEMATICAL STATEMENT OF THE PROBLEM
2.5.3 Forward speed radiation-diraction
Another important problem in oshore technology is the slow drift motion of a oat-
ing marine structure, such as a moored ship or a moored oil platform. The motion is
generated by resonance between the moored structure and slowly oscillating waves, and
may result in very large horizontal displacements. The viscous damping and the wave
(radiation) damping are often small; in many sea states the so-called wave drift damping
dened as the increase in wave drift forces due to a small forward velocity of a body
moving in waves may be the dominant damping eect.
The concept of wave drift damping has been introduced by Wichers and Sluijs (1979).
It has been discussed further by Wichers and Huijsmans (1984), and Wichers (1988). An
extensive study among many other contributions of wave drift forces was published
by Pinkster (1980).
Consider a body moving horizontally with constant forward speed U in response to
incoming regular waves with small amplitude-wavelength ratio A/. Let the reference
frame be xed to the body, with the undisturbed free surface in the (x, y)-plane, the
x-axis in the direction of forward motion, and the z-axis vertically upwards. In this
frame the body performs small oscillations due to the incoming waves and is embedded
in a uniform current with speed U along the x-axis. Assuming the uid to be ideal
and of innite extent in the lower half-space, and the ow to be free of vorticity, and
neglecting eects of viscosity and surface tension, then there exists a velocity potential
that satises Laplaces equation:

2
= 0 . (2.53)
This potential can be split up as follows:
(x; t) =
s
(x) +
d
(x; t) +
r
(x; t) . (2.54)
The rst part
s
represents the steady ow and is independent of time. The unsteady
parts
d
and
r
are time-harmonic with frequency of encounter , determined by the
angle of incidence and the orbital frequency of the incoming waves:
= Uk cos , (2.55)
where k =
2
/g is the deep water approximation of the zero-speed wave number. The case
= 0 corresponds to following waves, while = corresponds to head (i.e. opposing)
waves.
The steady potential
s
can be put as

s
(x) = U
s
(x) = U [(x) x] , (2.56)
where the term Ux accounts for the uniform current, and U(x) is the steady distur-
bance due to the presence of the body.
The total radiation potential
r
represents the eects of the body oscillations; it can
be written as

r
(x; t) = Re

ie
it
6

j=1

j
(x)

, (2.57)
where
j
is the amplitude in the j-mode of motion (corresponding to surge, sway, heave,
roll, pitch, and yaw) and
j
is the corresponding radiation potential for unit amplitude
of motion.
2.5. WAVE-BODY FORMULATIONS 27
The total diraction potential
d
can be expressed as

d
(x; t) = Re

Ae
it

0
(x) +
7
(x)

, (2.58)
where
7
is the scattering potential, and
0
is the incoming wave potential:

0
(x) =
ig

e
kz
e
ik(x cos +y sin )
. (2.59)
The steady potential (2.56) fullls

s
n = 0

n
= n
1
on S , (2.60)
corresponding to a zero ux through the wetted body surface S; the normal vector
n = (n
1
, n
2
, n
3
)
T
points out of the uid domain.
The boundary conditions for the unknown potentials
j
are see, for instance,
Newman (1978)

j
n
= n
j
+
U
i
m
j
, j = 1, 2, . . . , 6 , (2.61)

7
n
=

0
n
, (2.62)
where (n
4
, n
5
, n
6
)
T
x n. The so-called m-terms m
j
are dened as
(m
1
, m
2
, m
3
)
T
= n (
s
) , (2.63)
(m
4
, m
5
, m
6
)
T
= n (x
s
) . (2.64)
Thus, the normal derivative of each radiation potential consists of two parts. The rst,
the n-term, represents the oscillatory normal velocity of the body, while the second, the
m-term, represents the change in the local steady eld due to the motion of the body.
Let L be the characteristic dimension of the body. If the Froude number Fr =
U/ (gL)
1/2
is small, the free surface condition for the steady potential can be approxi-
mated to rst order in Fr by

s
z
= 0 at z = 0 . (2.65)
The steady-state problem dened by (2.53), (2.60), and (2.65) can now easily be solved
by means of a source distribution method.
The radiation potentials
j
, with j = 1, 2, . . . , 6, and the diraction potential
d
=

0
+
7
will then have to satisfy to rst order in Fr the combined kinematic and
dynamic free surface conditions:

j
+ 2i

j
+i
j

s
+g

j
z
= 0 at z = 0 , (2.66)
where

(/x, /y)
T
denotes the horizontal gradient operator.
When
s
is known, this is a linear boundary condition with spatially dependent
coecients. Far away from the body, we have
s
= Ux, and (2.66) simplies to

j
2iU

j
x
+g

j
z
= 0 at z = 0 , (2.67)
28 CHAPTER 2. MATHEMATICAL STATEMENT OF THE PROBLEM
which contains constant coecients only.
A radiation condition at innity stating that all potentials
j
must correspond
to outgoing waves renders this forward speed radiation-diraction problem uniquely
solvable.
Recently, Nossen, Grue, and Palm (1991) presented a boundary integral equation method
to compute the rst order unsteady forces and wave drift forces for arbitrary three-
dimensional bodies at small forward speed. Following the approach introduced by Hui-
jsmans and Hermans (1985), the forward velocity U is put in nondimensional form as
U/g. Typical values for the wave period and the body velocity in oshore problems
are T = 10 s and U = 1 m/s respectively, giving 0.06. The velocity potential and
the Greens function with the latter satisfying the radiation condition at innity
are then expressed in power series of , retaining linear terms only. The solution can be
expressed as an integral over the wetted body surface and the free surface.
Another way to solve the above radiation-diraction problem has been indicated
by Zhao and Faltinsen (1989). They use a hybrid method, where close to the body a
boundary element method with Rankine-sources is applied. This region is matched to
an outer regime where a multipole expansion is used. Finally, we mention the approach
of Wu and Eatock Taylor (1990), wherein the velocity potential is developed in a series
of over the whole free surface. Their examples, however, are all two-dimensional.
A discussion of the state of aairs with regard to the solution of the radiation-
diraction problem also in coherence with the steady-state ship wave-making and
wave-resistance problem has recently been given by Newman (1991).
2.5.4 Nonlinear motion at zero forward speed
Finally, we present the governing equations for an unrestrained body at zero forward
speed, i.e. a freely oating body. Let the position and the orientation of the body
be specied by the three-component vectors x
G
= (x
1
, x
2
, x
3
)
T
and

G
= (
1
,
2
,
3
)
T
respectively, as shown in Figure 2.2.
The motion of the uid near the wetted body surface S is determined by the same
mechanisms which are active in the vicinity of an impermeable moving boundary. So,
under the assumptions of an ideal uid and an irrotational ow, the proper boundary
condition is given by (2.28), repeated here for convenience:

n
= n =

V n , (2.68)
where

V is the velocity of a point on S and the unit normal vector n points out of the
body. In terms of the body position and orientation, this condition reads

n
= n =

x
G
n + (r n)

G
(2.69)
where a dot denotes dierentiation with respect to time and r = x
S
x
G
represents the
position of a point x
S
on S with respect to the centre of gravity G, see Figure 2.2. In case
of a oating body under forced motion,

x
G
and

G
are prescribed, and the uid-body
interaction is completely determined by (2.69).
However, if the body is oating freely, then x
G
(t) and

G
(t) are not known a priori,
and these variables have to be determined as part of the solution. In this case, the set of
2.6. PROBLEM DEFINITIONS 29
governing equations is supplemented with the equations of motion for a rigid body with,
say, mass M and moment of inertia

I about its principal axes. The gravitational force
acting on the body is given by

F
g
= Mge
3
. (2.70)
The inertial forces and moments acting on the body are directly obtained from hydro-
dynamical considerations; integration of the pressure exerted by the uid over the body
surface gives

F
i
=

pndS ,

L
i
=

p (r n) dS , (2.71)
where the pressure p equals the Bernoulli pressure from (2.24), where we have chosen
p
0
= 0 and B(t) = 0:
p =

t
+
1
2
( ) +gz

. (2.72)
Thus, in the absence of viscosity, the equations of motion for the freely oating body
read
M

x
G
=

pndS Mge
3
,

I

G
=

p (r n) dS (2.73)
where the symbol is used to dene a component-wise product of two vectors:
a

b (a
1
b
1
, a
2
b
2
, a
3
b
3
)
T
. (2.74)
The Neumann boundary condition (2.69) and the hydrodynamic equations of motion (2.73)
are the additional governing equations for a body oating freely at zero forward speed
in or below the free surface of an ideal and vorticity-free uid.
2.6 Problem denitions
The potential problem that will be dealt with in this thesis is described as
Nonlinear gravity driven water waves travelling over a bottom, and the hy-
drodynamic interaction with solid xed structures and rigid bodies oating in
or below the free surface at zero forward speed.
The evolution of this particular type of dynamical systems is governed by the following
set of (partial) dierential equations:
1. Laplaces equation for the velocity potential throughout the uid domain:

2
= 0 . (2.75)
2. Nonlinear dynamic and kinematic conditions on the free surface:
D
Dt
=
1
2
( ) gz ,
Dx
F
Dt
= . (2.76)
30 CHAPTER 2. MATHEMATICAL STATEMENT OF THE PROBLEM
3. Zero-ux condition on impermeable xed boundaries:

n
= 0 . (2.77)
4. Contact condition on impermeable moving boundaries:

n
=

V n . (2.78)
5. Hydrodynamic equations of motion for freely oating bodies:
M

x
G
=

pndS Mge
3
,

I

G
=

p (r n) dS . (2.79)
6. Bernoullis equation throughout the uid domain:

=

t
+
1
2
( ) +gz . (2.80)
In general, supplementary initial conditions and the imposition of a Sommerfeld radiation
condition at innity render these problems uniquely solvable.
From now on, the problem of free surface waves travelling on water of nite depth
governed by (1-3) is referred to as a (classical) water-wave problem. Likewise, a
system of free surface waves in hydrodynamic interaction with oating bodies at zero
forward speed governed by (1-6) is labelled a wave-body system, and any problem
matching this denition is called a wave-body problem hereafter.
The above set of equations constitutes an elliptic spatial problem with time dependent
boundary conditions. In most numerical procedures these water-wave and wave-body
problems are treated as such; the elliptic boundary value problem is solved at a certain
time level, and then the boundary conditions and, if a oating body is involved, the
equations of motion for the body are integrated in time to obtain a new boundary
value problem, and so forth. In chapter 4 we shall outline a method, proceeding along
these lines, for the numerical simulation of the physical systems described above. This
panel method is based on a Greens formulation for the velocity potential; the transition
from an elliptic boundary value problem to a boundary integral equation is treated in
the next chapter.
2.7 Bibliography
Adachi, H., and Ohmatsu, S. 1980. On the time dependent potential and its appli-
cation to wave problems. Proceedings of the Thirteenth Symposium on Naval Hydrody-
namics, Tokyo, Japan.
Angell, T.S., Hsiao, G.C., and Kleinman, R.E. 1986. An integral equation for
the oating-body problem. Journal of Fluid Mechanics 166:161-171.
Beck, R.F., and Liapis, S. 1987. Transient motions of oating bodies at zero forward
speed. Journal of Ship Research 31(3):164-176.
Broeze, J., and Romate, J.E. 1992. Absorbing boundary conditions for free surface
2.7. BIBLIOGRAPHY 31
wave simulations with a panel method. Journal of Computational Physics 99(1):146-158.
Brown, D.T., Eatock Taylor, R., and Patel, M.H. 1983. Barge motions in ran-
dom seas - a comparison of theory and experiment. Journal of Fluid Mechanics 129:385-
407.
Campana, E., Lalli, F., and Bulgarelli, U. 1989. Some numerical computa-
tions about free surface boundary layer and surface tension eects on nonlinear waves.
Proceedings of the Fifth International Conference on Numerical Ship Hydrodynamics,
Hiroshima, Japan.
van Daalen, E.F.G., Broeze, J., and van Groesen, E. 1992. Variational prin-
ciples and conservation laws in the derivation of radiation boundary conditions for wave
equations. Mathematics of Computation 58(197):55-71.
Dawson, C.W. 1977. A practical computer method for solving ship-wave problems.
Proceedings of the Second International Conference on Numerical Ship Hydrodynamics,
Berkeley, California.
Forbes, L.K. 1984. Irregular frequencies and iterative methods in the solution of steady
surface-wave problems in hydrodynamics. Journal of Engineering Mathematics 18(4):299-
313.
Gadd, G.E. 1976. A method for computing the ow and surface wave pattern around
full forms. Transactions of the Royal Institution of Naval Architects 118:207-216.
Hadamard, J. 1923. Lectures on Cauchys Problem. Oxford University Press.
Havelock, T.H. 1909. The wave-making resistance of ships. Proceedings of the Royal
Society of London, Series A, 82:276-300.
Higdon, R.L. 1987. Numerical absorbing boundary conditions for the wave equation.
Mathematics of Computation 49(179):65-90.
Huijsmans, R.H.M., and Hermans, A.J. 1985. A fast algorithm for computation
of 3-D ship motions at moderate forward speed. Fourth International Conference on Nu-
merical Ship Hydrodynamics, Washington, DC.
Hulme, A. 1983. A ring-source/integral-equation method for the calculation of hydrody-
namic forces exerted on oating bodies of revolution. Journal of Fluid Mechanics 128:387-
412.
Jensen, G., Bertram, V., and S oding, H. 1989. Ship wave-resistance computations.
Proceedings of the Fifth International Conference on Numerical Ship Hydrodynamics,
Hiroshima, Japan.
John, F. 1950. On the motion of oating bodies II. Simple harmonic motions. Commu-
nications on Pure and Applied Mathematics 3:45-101.
Kelvin, Lord. 1887. On the waves produced by a single impulse in water of any depth,
or in a dispersive medium. Proceedings of the Royal Society of London, Series A, 42:80-
85.
Lamb, Sir H. 1932. Hydrodynamics. Cambridge University Press.
Landweber, L., and Patel, V.C. 1979. Ship boundary layers. Annual Review of
Fluid Mechanics 11:173-205.
Lenoir, M., and Martin, D.A. 1981. An application of the principle of limiting
32 CHAPTER 2. MATHEMATICAL STATEMENT OF THE PROBLEM
absorption to the motions of oating bodies. Journal of Mathematical Analysis and Ap-
plications 79:370-383.
Liu, Y.W. 1991. A boundary integral equation for the two-dimensional oating-body
problem. SIAM Journal on Mathematical Analysis 22(4):973-981.
Martin, P.A. 1981. On the null-eld equations for water-wave radiation problems.
Journal of Fluid Mechanics 113:315-332.
Mei, C.C. 1978. Numerical methods in water-wave diraction and radiation. Annual
Review of Fluid Mechanics 10:393-416.
Michell, J.H. 1898. The wave resistance of a ship. Philosophical Magazine, Series 5,
45:106
Morse, P.M., and Feshbach, H. 1953. Methods of Theoretical Physics I and II.
McGraw-Hill.
Nestegard, A., and Sclavounos, P.D. 1984. A numerical solution of two-dimensional
deep water wave-body problems. Journal of Ship Research 28(1):48-54.
Newman, J.N. 1978. The theory of ship motions. Advances in Applied Mechanics
18:221-280.
Newman, J.N. 1985. Transient axisymmetric motion of a oating cylinder. Journal of
Fluid Mechanics 157:17-33.
Newman, J.N. 1991. The quest for a three-dimensional theory of ship-wave interactions.
Philosophical Transactions of the Royal Society of London, Series A, 334:213-227.
Ni, S.-Y. 1987. A Method for Calculating Non-linear Free Surface Potential Flows using
Higher Order Panels. Ph.D. thesis, Chalmes University. Gothenborg, Sweden.
Noblesse, F. 1983. Integral identities of potential theory of radiation and diraction of
regular water waves by a body. Journal of Engineering Mathematics 17(1):1-13.
Nossen, J., Grue, J., and Palm, E. 1991. Wave forces on three-dimensional oating
bodies with small forward speed. Journal of Fluid Mechanics 227:135-160.
Pinkster, J.A. 1980. Low Frequency Second Order Wave Exciting Forces on Floating
Structures. Ph.D. thesis, Delft University of Technology. Delft, The Netherlands. Also:
MARIN Publication No.650. Wageningen, The Netherlands.
Raven, H.C. 1988. Variations on a theme by Dawson. Proceedings of the Seventeenth
Symposium on Naval Hydrodynamics, The Hague, The Netherlands.
Raven, H.C. 1992. A practical nonlinear method for calculating ship wavemaking and
wave resistance. Proceedings of the Nineteenth Symposium on Naval Hydrodynamics,
Seoul, Korea.
Romate, J.E. 1992. Absorbing boundary conditions for free surface waves. Journal of
Computational Physics 99(1):135-145.
Sclavounos, P.D., and Lee, Ch.-H. 1985. Topics on boundary element solutions of
wave radiation-diraction problems. Proceedings of the Fourth International Conference
on Numerical Ship Hydrodynamics, Washington, DC.
Sclavounos, P.D., and Nakos, D.E. 1988. Stability analysis of panel methods for
free-surface ows with forward speed. Proceedings of the Seventeenth Symposium on
Naval Hydrodynamics, The Hague, The Netherlands.
2.7. BIBLIOGRAPHY 33
Simon, M.J., and Ursell, F. 1984. Uniqueness in linearized two-dimensional water-
wave problems. Journal of Fluid Mechanics 148:137-154.
Sommerfeld, A. 1964. Partial Dierential Equations in Physics. Volume VI of Lec-
tures on Theoretical Physics. Academic Press.
Stoker, J.J. 1957. Water Waves. The Mathematical Theory with Applications. Inter-
science Publishers.
Ursell, F. 1981. Irregular frequencies and the motion of oating bodies. Journal of
Fluid Mechanics 105:143-156.
Wichers, J.E.W. 1988. A Simulation Model for a Single Point Moored Tanker. Ph.D.
thesis, Delft University of Technology. Delft, The Netherlands. Also: MARIN Publica-
tion No.797. Wageningen, The Netherlands.
Wichers, J.E.W., and Huijsmans, R.H.M. 1984. On the low frequency hydrody-
namic damping forces acting on oshore moored vessels. Proceedings of the Oshore
Technology Conference, Houston, Texas, OTC 4831.
Wichers, J.E.W., and Sluijs, M.F. 1979. The inuence of waves on the low
frequency hydrodynamic coecients of moored vessels. Proceedings of the Oshore Tech-
nology Conference, Houston, Texas, OTC 3625.
Wu, G.X., and Eatock Taylor, R. 1990. The hydrodynamic force on an oscillating
ship with low forward speed. Journal of Fluid Mechanics 211:333-353.
Zhao, R., and Faltinsen, O.M. 1989. Interaction between current, waves and
marine structures. Fifth International Conference on Numerical Ship Hydrodynamics,
Hiroshima, Japan.
34 CHAPTER 2. MATHEMATICAL STATEMENT OF THE PROBLEM
Now when gleaming dawn with bright eyes beheld the lofty peaks of Pelion,
and the calm headlands were being drenched as the sea was rued by the
winds, then Tiphys awoke from sleep; and at once he roused his comrades to
go on board and make ready the oars. And a strange cry did the harbour of
Pagasae utter, yea and Pelian Argo herself, urging them to set forth. For
in her a beam divine had been laid which Athena had brought from an oak
of Dodona and tted in the middle of the stem. And the heroes went to the
benches one after the other, as they had previously assigned for each to row
in his place, and took their seats in due order near their ghting gear. In the
middle sat Ancaeus and mighty Heracles, and near him he laid his club, and
beneath his tread the ships keel sank deep. And now the hawsers were being
slipped and they poured wine on the sea. But Jason with tears held his eyes
away from his fatherland. And just as youths set up a dance in honour of
Phoebus either in Pytho or haply in Ortygia, or by the waters of Ismenus,
and to the sound of the lyre round his altar all together in time beat the earth
with swiftly-moving feet; so they to the sound of Orpheus lyre smote with
their oars the rushing sea-water, and the surge broke over the blades; and on
this side and on that the dark brine seethed with foam, boiling terribly through
the might of the sturdy heroes. And their arms shone in the sun like ame
as the ship sped on; and ever their wake gleamed white far behind, like a path
seen over a green plain. On that day all the gods looked down from heaven
upon the ship and the might of the heroes, half-divine, the bravest of men then
sailing the sea; and on the topmost heights the nymphs of Pelion wondered as
they beheld the work of Itonian Athena, and the heroes themselves wielding
the oars. And there came down from the mountain-top to the sea Chiron,
son of Philyra, and where the white surf broke he dipped his feet, and, often
waving with his broad hand, cried out to them at their departure, Good speed
and a sorrowless home-return!
Argonautica, Book I, Verses 519-556.
Chapter 3
Boundary Integral Equation
Formulations
Now when they had left the curving shore of the harbour through the cun-
ning and counsel of prudent Tiphys son of Hagnias, who skilfully handled the
well-polished helm that he might guide them steadfastly, then at length they set
up the tall mast in the mast-box, and secured it with forestays, drawing them
taut on each side, and from it they let down the sail when they had hauled it to
the top-mast. And a breeze came down piping shrilly; and upon the deck they
fastened the ropes separately round the well-polished pins, and ran quietly past
the long Tisaean headland. And for them the son of Oeagrus
1
touched his lyre
and sang in rhythmical song of Artemis, saviour of ships, child of a glorious
sire, who hath in her keeping those peaks by the sea, and the land of Iolcos;
and the shes came darting through the deep sea, great mixed with small, and
followed gambolling along the watery paths. And as when in the track of the
shepherd, their master, countless sheep follow to the fold that have fed to the
full of grass, and he goes before gaily piping a shepherds strain on his shrill
reed; so these shes followed; and a chasing breeze ever bore the ship onward.
Argonautica, Book I, Verses 559-579.
1
i.e. Orpheus
35
36 CHAPTER 3. BOUNDARY INTEGRAL EQUATION FORMULATIONS
3.1 Introduction
In this chapter the (numerical) solution of the time independent eld equation (2.23) is
considered. Solutions of Laplaces equation

2
(x) = 0 , x in (3.1)
are referred to as harmonic functions.
The velocity potential satisfying (3.1) in the entire uid domain is subject to
conditions on the bounding surface . These conditions can be written in the general
form

,

n
, x

= 0 , x on (3.2)
In the following it is assumed that the boundary value problem (3.1-3.2) is well-posed.
In the past, numerous methods both analytical and numerical have been developed
to solve the above problem. Since the eld equation (3.1) is linear and elliptic, one might
think that this problem can easily be solved with standard methods. However, diculties
may arise from the supplementary boundary conditions (3.2); for instance, geometrical
singularities (sharp edges, corners, etc.) hamper the solution of this boundary value
problem.
As far as numerical techniques are concerned, nite dierence methods, nite vol-
ume methods, nite element methods, and integral equation methods are used most
frequently. The rst three types of methods are so-called eld discretization methods;
in these techniques the eld equation here Laplaces equation (3.1) is discretized
directly. In integral equation methods the rst step is to transform the eld equation
into a (set of) boundary integral equation(s) by using a fundamental identity here
Greens second identity for the problem under consideration. The next step in inte-
gral equation methods is the discretization of these boundary integral equations.
So far, none of the above methods has been proven to be superior to all other meth-
ods. Each method has its specic advantages and disadvantages; for instance, integral
equation methods lower the problem dimension by one, thus reducing data storage and
allowing a more ecient use of computer memory. On the other hand, the (derived)
integral equations are more dicult to solve than the (original) eld equation; in gen-
eral, a full system matrix is obtained, while eld discretization methods usually yield
band matrices. As a consequence, relatively expensive matrix solvers have to be used
in integral equation techniques. Furthermore, integral equation methods are sensitive
to the discretization, and later extensions to the solution of problems that are not gov-
erned by the eld equation concerned, are dicult or even impossible. However, for
problems with moving boundaries, like free surface wave problems, integral equation
methods have a clear advantage compared to the eld discretization methods; the mo-
tion of (two-dimensional) surface grids is much easier to compute than the motion of
(three-dimensional) volume grids. A discussion of the advantages and disadvantages of
eld discretization methods and integral equation methods including an analysis of
the computational eort involved has been given by Romate and Zandbergen (1989).
The use of integral equation methods goes back to the fties, when these techniques
were introduced in various elds, such as aerodynamics and elastostatics. The sound
theoretical basis of potential theory and the fast growing knowledge of integral equations
3.2. INTEGRAL EQUATIONS IN POTENTIAL THEORY 37
since the pioneering work of Fredholm
2
made integral equation methods good competitors
of other numerical techniques.
For the above and other reasons Romate (1989) decided to use a boundary integral
equation method to solve the nonlinear free surface wave problem. Elaborating on his
panel method, a new simulation procedure for water waves interacting with xed and
oating bodies was developed; this numerical algorithm is outlined and discussed in the
next chapter.
The aim of this chapter is to present a compact introduction to integral equations
in potential theory. In section 3.2 a number of integral equations is derived using basic
concepts of potential theory. The well-posedness of the boundary value problem (abbre-
viated: BVP) (3.1-3.2) is considered in section 3.3. The choice of the integral equation
method to solve this boundary value problem is accounted for in section 3.4. For a more
detailed introduction to integral equation methods for potential problems we refer to
Romate (1989).
3.2 Integral equations in potential theory
In this section the BVP (3.1-3.2) will be reformulated in terms of boundary integral
equations. Those elements of potential theory that are essential for the panel method
described in chapter 4 will be discussed briey, assuming that the reader is familiar
with the basic concepts. For an extensive mathematical treatment of potential theory
the reader is referred to Kellogg (1953), and Courant and Hilbert (1962). An excellent
introduction to integral equation methods in potential (scattering) theory is given by
Colton and Kress (1983).
The well-posedness of the elliptic BVP (3.1-3.2) implies that solutions (x) and all
spatial derivatives of these solutions are nite and continuous throughout , except possi-
bly at some points on . Therefore, it can be expected that the properties of a harmonic
solution depend strongly on the shape of and the boundary conditions (3.2). Indeed,
these boundary data determine the solution , as will be shown later on.
The three-dimensional singly connected
3
domain in Figure 3.1 is bounded by the
piecewise smooth surface . For two arbitrary scalar elds and which are C
2
-
continuous
4
throughout , Greens second identity states:

[ ] d =

dS

, (3.3)
where
2
is Laplaces operator and n

is the unit normal vector in



on , pointing
into .
Substituting the unknown potential (x) for , with satisfying (3.1), and replacing
by the Greens function
G

; x

=
1
4r
, r = |x

| (3.4)
2
Fredholm, Ivar (1866-1927), Swedish mathematician who made important contributions to the theory
of integral equations.
3
See section 2.5 for a denition.
4
This means that and are continuously dierentiable and have continuous spatial derivatives up
to second order.
38 CHAPTER 3. BOUNDARY INTEGRAL EQUATION FORMULATIONS
Figure 3.1: Denition of , , and n

.
with x xed and

on , we obtain the following identity:

(x) G

; x

d =

; x

dS

G
n

; x

dS

. (3.5)
If x is in the interior of , then the Greens function G

; x

has a singularity in

= x.
Introducing a neighbourhood

(x) of x with boundary S

(x), a small sphere with centre


at x and radius , identity (3.5) applied to `

(x) with boundary S

(x)
reduces in the limit 0 to
(x) =

; x

G
n

; x

dS

. (3.6)
If x is on the boundary , then identity (3.5) reduces in a similar way to
(x)
4
(x) =

; x

G
n

; x

dS

, (3.7)
where (x) denotes the interior space angle of in x; if x is on a smooth part of ,
then (x) = 2. The symbol

denotes the nite part of the integral

in the sense of
Hadamard (1923); see also Courant and Hilbert (1962).
Since all surface integrals in (3.5) are regular, the normal derivative operator /n =
n can be brought under the integral signs; then, by the above token a second integral
equation is obtained:
(x)
4

n
(x) =

G
n
x

; x


2
G
n
x
n

; x

dS

, (3.8)
where the normal derivative operator is applied to G both in the eld point x and in the
source point

.
3.2. INTEGRAL EQUATIONS IN POTENTIAL THEORY 39
The surface also bounds the exterior domain

= IR
3
` ( ) which extends to
innity and is singly connected. Let S
R
(x) denote a sphere with centre at x and radius
R suciently large to enclose , see Figure 3.1. Then, application of Greens second
identity (3.3) to the part of

bounded by and S
R
(x) leads to

(x)
4

(x) =

G
n

; x

; x

dS

S
R
(x)

G
r

; x

; x

dS

, (3.9)

(x)
4

n
(x) =


2
G
n
x
n

; x

G
n
x

; x

dS

S
R
(x)


2
G
n
x
r

; x

G
n
x

; x

dS

,
(3.10)
where

satises Laplaces equation in

, and

(x) denotes the exterior space angle


of in x; obviously, (x) +

(x) = 4 for all x on . For increasing radius R, the


contributions of the spherical boundary S
R
(x) vanish, assuming that

(R) = O

R
1

for R .
From now on it is assumed that x is on a regular (smooth) part of , and hence
(x) =

(x) = 2. Addition of (3.7) and (3.9), and (3.8) and (3.10) results in
1
2
( +

) (x) =

; x

dS

G
n

; x

dS

, (3.11)
1
2

n
+

(x) =

G
n
x

; x

dS


2
G
n
x
n

; x

dS

. (3.12)
Then, if we put

,

on (3.13)
(

,

on (3.14)
the identities (3.11-3.12) can be rewritten in terms of the source distribution and the
normal dipole distribution on the surface :
(x) +
1
2
(x) =

; x

G
n

; x

dS

(3.15)
40 CHAPTER 3. BOUNDARY INTEGRAL EQUATION FORMULATIONS

n
(x)
1
2
(x) =

G
n
x

; x


2
G
n
x
n

; x

dS

(3.16)
In order to have a unique solution

of Laplaces equation, that is

(x) = 0 , x in

, (3.17)
boundary conditions on are needed; written in a general form, we have

n
, x

= 0 , x on . (3.18)
Since

is ctitious
5
, any condition that renders the exterior boundary value prob-
lem (3.17-3.18) well-posed is adequate. From (3.13-3.14) and (3.15-3.16) it is clear that
dierent choices for the boundary conditions (3.18) lead to dierent boundary integral
equations:
1. Choose

= on = 0
i.e. a source-only distribution (x on ):
(x) =

; x

dS

, (3.19)

n
(x) =
1
2
(x) +

G
n
x

; x

dS

. (3.20)
2. Choose

n
=

n
on = 0
i.e. a dipole-only distribution (x on ):
(x) =
1
2
(x) +

G
n

; x

dS

, (3.21)

n
(x) =


2
G
n
x
n

; x

dS

. (3.22)
3. Choose

= 0 on

n
= 0 on

=

n
, =
i.e. a Greens or mixed source-dipole distribution (x on ):
1
2
(x) =

; x

G
n

; x

dS

, (3.23)
1
2

n
(x) =

G
n
x

; x


2
G
n
x
n

; x

dS

.
(3.24)
5
This means that the potential

does not represent a real physical phenomenon.


3.3. WELL-POSEDNESS: EXISTENCE AND UNIQUENESS 41
The nal choice of the boundary integral equation formulation will be made in section 3.4.
First, we discuss the well-posedness of the BVP (3.1-3.2) and present some results on the
uniqueness of integral equations (3.7-3.8) for various boundary conditions and singularity
distributions.
3.3 Well-posedness: existence and uniqueness
For the well-posedness of the BVP (3.1-3.2) the denition due to Hadamard reads
Denition 1 : Well-posedness (Hadamard, 1923)
The boundary value problem (3.1-3.2) is well-posed if for all x in a unique solution ,
depending continuously on the boundary data, exists.
The question of existence and uniqueness of a solution to (3.1-3.2) has been treated
in many books and papers; for completeness, some important results are presented here.
For rigorous proofs of these results the reader is referred to Kellogg (1953), Jaswon and
Symm (1977), and Courant and Hilbert (1962).
Suppose that the simply connected domain is bounded by a piecewise smooth surface
=
M

i=1
S
i
, (3.25)
with all subsurfaces S
i
twice continuously dierentiable. In addition, it is assumed that
satises the so-called external cone condition:
Condition 1 : External cone condition
For each point x on there exists a circular cone lying outside with angle > 0 and
vertex at x.
Furthermore, it is assumed that the boundary conditions

,

n
, x

= 0 , x on S
i
, i = 1, 2, . . . , M , (3.26)
are all continuous.
A necessary condition for the existence of a solution to (3.1-3.2) is obtained by applying
Gauss theorem
6
to the vector velocity eld :


2
d =

ndS . (3.27)
By continuity we have
2
= 0 throughout , giving


n
dS = 0 . (3.28)
6
Gauss, Johann Karl Friedrich (1777-1855), German mathematician. In addition to his work in pure
mathematics, he made major contributions to theoretical astronomy, geodesy, terrestrial magnetism and
electricity, and other branches of physics.
42 CHAPTER 3. BOUNDARY INTEGRAL EQUATION FORMULATIONS
Condition (3.28) states that the total ux of uid through the surface must equal
zero.
With the above denitions and conditions, the following theorems on the existence and
uniqueness of a solution to the BVP (3.1-3.2) can be formulated and proven:
Theorem 1 : Existence for the boundary value problem (3.1-3.2)
Let IR
3
be a simply connected domain, and let , S
i
, and
i
satisfy the above
mentioned conditions. Further, let at each point on either , or /n, or a linear
combination of and /n be specied. Then there is at least one scalar function
satisfying the boundary value problem (3.1-3.2), only if (3.28) holds.
Theorem 2 : Uniqueness for the boundary value problem (3.1-3.2)
Let IR
3
be a simply connected domain, and let , S
i
, and
i
satisfy the above
mentioned conditions. Further, let at each point on either , or /n, or a linear
combination of and /n be specied. Then there is at most one scalar function
satisfying the boundary value problem (3.1-3.2); this solution is determinate but for a
constant if the value of is nowhere specied on .
From the foregoing it appears that the solution to the boundary value problem (3.1-3.2)
can be constructed by means of integral equations (3.15-3.16). However, the use of this
set of integral equations again raises the question of the existence of a unique solution
. From the derivation of the integral identities (3.7-3.8) with (x) = 2 for all x on
it is clear that
satises (3.2) [ satises (3.1) satises (3.15-3.16) ] ,
which solves the question of existence and uniqueness, assuming that already satises
the given boundary conditions.
Next, a summary of the uniqueness results with regard to a solution for various
boundary conditions is given. A boundary condition is known as
a Dirichlet condition if is prescribed;
a Neumann condition if /n is prescribed.
The results on uniqueness for (3.7) and (3.8) with (x) = 2 are listed in table 3.1,
where non-unique means that the solution is determined but for a constant.
With these results, the uniqueness of the dierent singularity distributions can be
determined for various boundary conditions. Therefore, let us return to the ctitious
problem introduced in the previous section, where
2

= 0 in the exterior domain

.
From (3.13-3.14) it is evident that the boundary conditions on for

and

/n
determine the type of surface distribution for the real problem of
2
= 0 in with
certain boundary conditions. For the uniqueness of this distribution it is necessary that
unique solutions and

of the given boundary value problems can be found with


identities (3.7-3.8).
3.4. CHOICE OF INTEGRAL EQUATION METHOD 43
boundary condition /n
integral equation (3.7)
internal Dirichlet specied unique
external Dirichlet specied unique
internal Neumann non-unique specied
external Neumann unique specied
integral equation (3.8)
internal Dirichlet specied unique
external Dirichlet specied unique
internal Neumann non-unique specied
external Neumann non-unique specied
Table 3.1: Uniqueness for solutions of (3.7-3.8) with various boundary conditions.
boundary condition source-only dipole-only Greens
integral equation (3.7)
internal Dirichlet unique unique /n unique
external Dirichlet unique non-unique /n unique
internal Neumann non-unique
external Neumann unique
integral equation (3.8)
internal Dirichlet /n unique
external Dirichlet /n unique
internal Neumann unique non-unique non-unique
external Neumann unique non-unique non-unique
Table 3.2: Uniqueness for various surface distributions.
The results on uniqueness for a source-only, a dipole-only, and a Greens (mixed source-
dipole) distribution on are listed in table 3.2. These results
7
will be used in the next
section to justify the choice of the boundary integral equation and the corresponding
surface distributions to solve the BVP (3.1-3.2).
3.4 Choice of integral equation method
In general, analytical solutions to the boundary value problem (3.1-3.2) can not be found.
However, the integral equations derived in section 3.2 can be used to construct approxi-
mate solutions; for that purpose, integral equations are formulated for a nite number of
7
The discussions so far have referred to simply connected domains only. The degree of connectivity
is determined by the minimum number of dierent barriers needed to make each subregion simply
connected; if n 1 such barriers are needed, the region is said to be n-ply connected. In a doubly
connected region the solution is a multi-valued function of position, and therefore non-unique. The
insertion of a barrier which creates two simply connected regions renders single-valued, and the theory
described earlier for simply connected domains can be applied. Along the barrier the jump in i.e.
the circulation when crossing the barrier has to be specied. See also Batchelor (1967).
44 CHAPTER 3. BOUNDARY INTEGRAL EQUATION FORMULATIONS
(say, N) points on the bounding surface , so that N equations are obtained in terms
of boundary data and unknowns. In the following we shall indicate a way to transform
this set of integral equations into a linear system, where the number of equations equals
the number of unknowns.
Figure 3.2: Boundary consisting of two smooth subsurfaces S
1
and S
2
.
Without loss of generality it is assumed that the boundary consists of two smooth
subsurfaces, see Figure 3.2. Suppose that a Dirichlet boundary condition is imposed
on S
1
, while S
2
is a Neumann boundary.
In order to solve the boundary value problem numerically, is discretized with N
1
panels (surface elements) on S
1
and N
2
panels on S
2
. For each panel Taylor series
expansions are used to approximate the surface shape and the variables dened on the
boundary; the Taylor series of the surface distributions and are expressed in terms
of N = N
1
+N
2
singularity strengths at the N panel midpoints. The integral equations
are then replaced by a set of N linear equations involving N source strengths and/or N
dipole strengths. A point collocation method is chosen to impose the boundary conditions
on ; the panel midpoints serve as collocation points, where the proper boundary
conditions are applied. Thus, N data are substituted, leaving only N unknowns to be
solved from a system of N linear equations:
Ay =

b . (3.29)
The choice of integral equation method determines the properties of this linear system.
The actual order of approximation of the surface shape and the order of expansion in
the Taylor series for the singularity distributions determine the accuracy of the method.
Higher order approximations will reduce leakage and hence the error in the solution.
By preference, the choice of the surface distributions and the integral equations should
lead to Fredholm integral equations of the second kind. Then, the resulting system
matrix A has large coecients on the diagonal (due to the principal parts of the singular
integrals), and iterative methods can be used to solve (3.29). Integral equations of the
second kind are numerically stable, which is one of the main reasons for their widespread
use in the past. A major disadvantage of this type of equations is that special properties
of Laplaces equation, such as symmetry, in general are not preserved, see Hsiao (1987).
3.4. CHOICE OF INTEGRAL EQUATION METHOD 45
Fredholm integral equations of the rst kind do preserve the symmetry and coerciveness
properties, which can be exploited by special matrix solvers such as conjugate gradient
(CG-) methods. However, equations of the rst kind give an ill-conditioned matrix A
(to be more precise: without large entries on the main diagonal), which may give rise to
numerical instabilities see, for instance, Hsiao (1987).
Another important criterium in the choice of the surface distributions and the integral
equations is the computational eort to attain a certain level of accuracy. As Hunt (1978)
states, the choice of a mixed source-dipole distribution reduces leakage considerably and
leads to better results. For this reason a Greens formulation is favourable. Another
major plus-point of a mixed formulation is that the ctitious potential

equals zero
throughout

. Hence, it can not introduce singularities in the solution on the bounding


surface .
Lastly, we mention an advantage which should not be underestimated; with a Greens
formulation the chosen singularities = /n and = are of direct physical inter-
est. To illuminate this, consider the free surface boundary conditions (2.76), involving
and its gradient , and suppose that the free surface potential is known. Then /n
is calculated with a panel method using a Greens formulation, and the tangential deriva-
tives /s
1
and /s
2
(where s
1
and s
2
are tangential coordinates) can be computed
from , thus giving the necessary ingredients to compute the free surface velocity .
For the above (and other less important) reasons Romate (1989) decided to use a Greens
formulation for his higher order panel method. Thus we are left with integral equa-
tions (3.23-3.24) to solve the boundary value problem (3.1-3.2). From these two identi-
ties, the latter is more expensive for a given level of accuracy, due to the evaluation of
the second order normal derivative of the Greens function G. Therefore, only the rst
integral equation will be used to construct a numerical solution procedure:
1
2
(x) =

; x

G
n

; x

dS

(3.30)
46 CHAPTER 3. BOUNDARY INTEGRAL EQUATION FORMULATIONS
With known in all collocation points on , the tangential velocities are calculated
using nite dierences. Once /n is also known in all collocation points, the velocity
eld on the boundary can be determined. Finally, with identity (3.30) the potential
can be computed throughout the uid domain , and from these data the quantities of
physical interest such as the velocity and pressure elds can be obtained.
3.5 Bibliography
Batchelor, G.K. 1967. An Introduction to Fluid Dynamics. Cambridge University
Press.
Colton, D., and Kress, R. 1983. Integral Equation Methods in Scattering Theory.
John Wiley and Sons.
Courant, R., and Hilbert, D. 1962. Methods of Mathematical Physics II. Inter-
science Publishers.
Hadamard, J. 1923. Lectures on Cauchys Problem. Oxford University Press.
Hsiao, G.C. 1987. On the stability of boundary element methods for integral equations
of the rst kind. In Boundary Element Methods IX. Volume 1: Mathematical and com-
putational aspects. Springer-Verlag.
Hunt, B. 1978. The panel method for subsonic aerodynamic ows: a survey of mathe-
matical formulations and numerical models, and an outline of the new British Aerospace
scheme. VKI Lecture series 1978-4 on Computational Fluid Dynamics.
Jaswon, M.A., and Symm, G.T. 1977. Integral Equation Methods in Potential The-
ory and Elastostatics. Academic Press.
Kellogg, O.D. 1953. Foundations of Potential Theory. Dover Publications.
Romate, J.E. 1989. The Numerical Simulation of Nonlinear Gravity Waves in Three
Dimensions using a Higher Order Panel Method. Ph.D. thesis, University of Twente.
Enschede, The Netherlands.
Romate, J.E., and Zandbergen, P.J. 1989. Boundary integral equation formulations
for free-surface ow problems in two and three dimensions. Computational Mechanics
4:276-282.
3.5. BIBLIOGRAPHY 47
Thereupon a spirit of contention stirred each chieftain, who should be the
last to leave his oar. For all around the windless air smoothed the swirling
waves and lulled the sea to rest. And they, trusting in the calm, mightily
drove the ship forward; and as she sped through the salt sea, not even the
storm-footed steeds of Poseidon would have overtaken her. Nevertheless when
the sea was stirred by violent blasts which were just rising from the rivers
about evening, forspent with toil, they ceased. But Heracles by the might of
his arms pulled the weary rowers along all together, and made the strong-knit
timbers of the ship to quiver. But when, eager to reach the Mysian mainland,
they passed along in sight of the mouth of Rhyndacus and the great cairn
of Aegaeon, a little way from Phrygia, then Heracles, as he ploughed up the
furrows of the roughened surge, broke his oar in the middle. And one half he
held in both his hands as he fell sideways, the other the sea swept away with
its receding wave. And he sat up in silence glaring round; for his hands were
unaccustomed to lie idle . . .
But the son of Zeus
8
having duly enjoined on his comrades to prepare the
feast took his way into a wood, that he might rst fashion for himself an oar
to t his hand. Wandering about he found a pine not burdened with many
branches, nor too full of leaves, but like to the shaft of a tall poplar; so great
was it both in length and thickness to look at. And quickly he laid on the
ground his arrow-holding quiver together with his bow, and took o his lions
skin. And he loosened the pine from the ground with his bronze-tipped club
and grasped the trunk with both hands at the bottom, relying on his strength;
and he pressed it against his broad shoulder with legs wide apart; and clinging
close he raised it from the ground deep-routed though it was, together with
clods of earth. And as when unexpectedly, just at the time of the stormy
setting of baleful Orion, a swift gust of wind strikes down from above, and
wrenches a ships mast from its stays, wedges and all; so did Heracles lift the
pine. And at the same time he took up his bow and arrows, his lion skin and
club, and started on his return . . .
Argonautica, Book I, Verses 1153-1171 and 1187-1206.
8
i.e. Heracles
48 CHAPTER 3. BOUNDARY INTEGRAL EQUATION FORMULATIONS
Chapter 4
Algorithm for Wave-Body
Simulations
Meantime Hylas with pitcher of bronze in hand had gone apart from the
throng, seeking the sacred ow of a fountain, that he might be quick in drawing
water for the evening meal and actively make all things ready in due order
against his lords
1
return . . . And quickly Hylas came to the spring which the
people who dwell thereabouts call Pegae. And the dances of the nymphs were
just now being held there . . . But one, a water-nymph was just rising from
the fair-owing spring; and the boy she perceived close at hand with the rosy
ush of his beauty and sweet grace . . . But as soon as he dipped the pitcher
in the stream, leaning to one side, and the brimming water rang loud as it
poured against the sounding bronze, straightway she laid her left arm above
upon his neck yearning to kiss his tender mouth; and with her right hand she
drew down his elbow, and plunged him into the midst of the eddy . . .
Alone of his comrades the hero Polyphemus, son of Eilatus, as he went
forward on the path, heard the boys cry, for he expected the return of mighty
Heracles . . . And straightway he told the wretched calamity while his heart
laboured with his panting breath. My poor friend, I shall be the rst to bring
thee tidings of bitter woe. Hylas has gone to the well and has not returned
safe, but robbers have attacked and are carrying him o, or beasts are tearing
him to pieces; I heard his cry. Thus he spake; and when Heracles heard his
words, sweat in abundance poured down from his temples and the black blood
boiled beneath his heart. And in wrath he hurled the pine to the ground and
hurried along the path whither his feet bore on his impetuous soul . . .
Argonautica, Book I, Fragments from Verses 1207-1264.
1
i.e. Heracles
49
50 CHAPTER 4. ALGORITHM FOR WAVE-BODY SIMULATIONS
4.1 Introduction
In this chapter we present a boundary element method for numerical time domain sim-
ulations of wave-body interactions in two and three dimensions.
Section 4.2 is a review of numerical techniques for the solution of the nonlinear wave-
body problem dened in section 2.6. For a recent survey of numerical methods for linear
and nonlinear water wave simulations we refer to the reviews given by Romate (1989).
The basic algorithm for nonlinear water wave simulations is outlined in section 4.3.
The major features of the discrete approximation of the problem are briey touched in
this section; the reader is referred to the work of Romate (1988, 1989, 1990) for a detailed
discussion of the geometry approximation, the calculation of the inuence coecients,
and the applied time stepping scheme.
In section 4.4 this basic algorithm is made t for the simulation of nonlinear water
waves interacting with oating bodies at zero forward speed. The essential feature of this
extended algorithm is that the hydrodynamic equilibrium of the uid and the body is
preserved at each time level. A stable time stepping mechanism for a freely oating body
calls for the accurate computation of the hydrodynamic forces from pressure integrations
over the wetted body surface S. However, the Bernoulli pressure involves the unknown
partial time derivative of the velocity potential . The fact that
t
, besides , satises
Laplaces equation, makes it possible to derive an extra boundary integral equation,
connecting
t
and its normal derivative
tn
; integration is over the uid domain boundary,
including S. But, if the body is oating freely, both
t
and
tn
are unknown along S; to
provide uniqueness, the hydrodynamic equations of motion for the body are transformed
into a boundary integral equation, connecting
t
and
tn
on S only. The discretized
versions of the extra integral equation and the transformed equations of motion together
yield a system of linear equations, such that the number of equations and the number of
unknowns are equal. The solution of the resulting matrix equation provides the
t
-term
in the Bernoulli pressure on S.
4.2 Review of methods for nonlinear ship motions
It has already been mentioned in section 2.6 that a time domain solution procedure for
the nonlinear wave-body problem involves the solution of Laplaces equation at each time
level, and the marching in time by integration of the nonlinear free surface conditions and
the equations of motion for an unrestrained body. With very few exceptions only nite
dierence methods and integral equation methods have been used to solve this problem.
2
In two dimensions, integral equation methods have proven to be very suitable for the
accurate and ecient solution of the nonlinear water-wave problem. At times, rather
successful extensions to linear and nonlinear wave-body problems have been reported.
The rst successful method for the transient water-wave problem (in the absence of
a oating body) was developed by Longuet-Higgins and Cokelet (1976). They used a
Lagrangian description of the free surface, which under the assumption of periodicity
in horizontal (propagation) direction is transformed into a closed contour around
the origin in the complex plane. Applying Greens theorem to the transformed Laplace
2
Other solution techniques applied to problems involving water waves only are, for instance,
spectral methods; see Fenton and Rienecker (1982), and Dommermuth and Yue (1987b). An excellent
work on the use of spectral methods in uid dynamics is due to Canuto, Hussaini, Quarteroni, and Zang
(1988); see also Hussaini and Zang (1987).
4.2. REVIEW OF METHODS FOR NONLINEAR SHIP MOTIONS 51
equation, they arrive at Fredholm integral equations of the rst kind. Despite the very
accurate high order approximation, smoothing techniques are needed to suppress the
development of non-physical wiggle instabilities.
Another successful method was initiated by Vinje and Brevig (1981a). This method
also for two-dimensional problems is based on Cauchys integral theorem
3
in the
complex plane. With z = x + iy, where y is taken vertically upwards, the complex
potential is put as (z; t) = (z; t) + i (z; t), where is the (real) potential and the
imaginary part is the stream function. Again, horizontal periodicity is assumed, but
now Laplaces equation is solved directly in the z-plane, allowing the extension to non-
periodic problems. Expressing the governing equations in terms of complex variables,
the following two identities are used:

0
(z
0
) + Re

(z)
z z
0
dz

= 0 , (4.1)
when is given in z
0
, and

0
(z
0
) Im

(z)
z z
0
dz

= 0 , (4.2)
when is given in z
0
. In (4.1-4.2) the interior angle at z
0
is denoted by
0
, which equals
if z
0
is on a smooth part of the uid contour C. In this way Fredholm integral equations
of the second kind are obtained in all cases. The resulting time domain algorithm is
very stable
4
, and smoothing is needed only for very large and steep waves. The same
technique has also been applied to the nonlinear wave-body problem in two dimensions;
in their ship motion computations, Vinje and Brevig (1981bc) encountered numerical
instabilities originating from the point where the body intersects the free surface. Another
major drawback of their method is that extension to three-dimensional problems is not
possible.
For a more complete literature survey on integral equation methods for transient non-
linear water-wave problems without oating bodies we refer to Romates doctoral thesis
(1989). The number of papers describing numerical results on nonlinear wave-body prob-
lems is still comparatively small. Using extensions of the integral equation techniques
described above, sometimes promising results have been obtained in two and three di-
mensions; a number of publications has been collected in table 4.1.
3
Cauchy, Augustin-Louis (1789-1857), French mathematician, who made important contributions to
the theory of complex functions.
4
Lin, Newman, and Yue (1984) use this method in their study of the singular behaviour of the free
surface near the intersection with an impulsively started vertical wavemaker; they suggest a special
treatment of the intersection point. This problem is discussed in detail in chapter 5.
52 CHAPTER 4. ALGORITHM FOR WAVE-BODY SIMULATIONS
author(s) / year method dim. description time-step
Faltinsen (1977) BEM-R 2D Lagrangian RK-4
Haussling/Coleman (1977/1979) FDM 2D Lagrangian mod.Euler
Chan/Chan (1980) FDM 3D Eulerian implicit
Vinje/Brevig (1981bc) BEM-C 2D Lagrangian Hamming
Isaacson (1982) BEM-R 3D Eulerian AB-2
Vinje/Maogang/Brevig (1982) BEM-C 2D Lagrangian Hamming
Greenhow/Lin (1985) BEM-C 2D Lagrangian Hamming
Telste (1985) FDM 2D Lagrangian mod.Euler
Yim (1985) BEM-C 2D Lagrangian Hamming
Greenhow (1987) BEM-C 2D Lagrangian Hamming
Dommermuth/Yue (1987a) BEM-R 2D mixed E-L RK-4
Wang/Spaulding (1988) FDM 2D Lagrangian mod.Euler
Yeung/Wu (1989) FDM 2D Lagrangian FTS
Cointe et al (1990) BEM-R 3D mixed E-L RK-4
Kang/Gong (1990) BEM-R 3D mixed E-L RK-4
Tanizawa/Sawada (1990) BEM-R 2D mixed E-L AB-1
Ehlers (1991) BEM-C 2D Lagrangian ABM-4
Saubestre (1991) BEM-R 2D mixed E-L RK-4
Yeung/Ananthakrishnan (1992) FDM 2D mixed E-L FTS
BEM = boundary element method
-R = in real domain (Greens second identity)
-C = in complex domain (Cauchys integral)
FDM = nite dierence method
RK = Runge-Kutta
FTS = fractional time-stepping
AB = Adams-Bashforth
ABM = Adams-Bashforth-Moulton
An appended number denotes the order of the time stepping method.
Table 4.1: Numerical methods for nonlinear ship motion simulations.
4.3 Basic algorithm for nonlinear water waves
In this section we shall briey discuss the basic features of our numerical algorithm
for nonlinear gravity wave simulations. For the record, it is stressed that the present
method for two-dimensional simulations was derived from Romates (1989) higher order
panel method for three-dimensional nonlinear gravity waves; once more, we refer to his
doctoral thesis for a detailed description of the original algorithm. Recently, Zandbergen,
Broeze, and van Daalen (1992) reported on a number of improvements on his method, and
presented very promising results on various types of free surface waves, such as solitary
waves, highly nonlinear waves, and plunging breakers. An account of the progress made
in three-dimensional wave simulations will be given by Broeze (1993).
It has been demonstrated in chapter 2 that under the assumptions of an irrotational
ow of an ideal uid the complete set of governing equations for nonlinear free surface
4.3. BASIC ALGORITHM FOR NONLINEAR WATER WAVES 53
ow under the action of gravity is given by:

2
= 0 in the uid domain , (4.3)
D
Dt
=
1
2
( ) gz on the free surface , (4.4)
Dx
F
Dt
= on the free surface , (4.5)

n
= 0 on the bottom , (4.6)
i.e. Laplaces equation throughout the uid, the dynamic and kinematic conditions on
the free surface, and the zero-ux condition on the bottom.
The above set of equations constitutes an elliptic spatial problem, characterized
by (4.3), with time dependent boundary conditions (4.4-4.5). In most numerical pro-
cedures, the water-wave problem is treated as such; the elliptic boundary value problem
is solved at a certain time level, and then the boundary conditions are updated to the
new time level, and so forth. This so-called step-by-step method is very popular for its
simplicity; it will be used here to develop a numerical algorithm for water wave simula-
tions.
In the rst place we have to solve the spatial problem, characterized by Laplaces
equation (4.3) for the velocity potential. With Greens second identity this eld equation
is transformed into a Fredholm equation of the second kind, see chapter 3:
1
2
(x) =

; x

G
n

; x

dS

(4.7)
where x is the boundary point considered, and integration is over the uid domain
boundary . This boundary is divided into a number of smooth subsurfaces, and each
of them is projected onto a grid in the computational domain with cubic splines. From
these spline data, each subsurface is approximated by curved panels. For each panel one
collocation point is determined from its central counterpart in the computational domain.
The boundary integral equation (abbreviated: BIE) is discretized with a linear source
distribution and a quadratic dipole distribution. In compact notation, the discretized
BIE, applied in collocation point i, reads
1
2

i
=
N

j=1

C
ij
s

j
n
+C
ij
d

j

(4.8)
where C
ij
s
and C
ij
d
are the source and dipole coecients respectively, and summation
is over all N collocation points. Substitution of for Dirichlet boundaries and
n
for
Neumann boundaries yields a system of linear equations, which is solved by Gaussian
elimination or with a conjugate gradients squared (CGS-) method. The structure of this
matrix equation is shown in Figure 4.1, where for convenience all Neumann boundaries
have been gathered (the same holds for all Dirichlet boundaries). The right-hand side
vector

contains contributions from the (substituted) known variables. The solution
of this matrix equation then provides
n
for Dirichlet boundaries and for Neumann
boundaries.
Next we have to solve the time dependent part of the problem, especially for the evolution
of the free surface. The new positions of the collocation points on moving boundaries
54 CHAPTER 4. ALGORITHM FOR WAVE-BODY SIMULATIONS
C
ij
s
C
ij
d

1
2

ij
Dirichlet Neumann

j
n

j
=

i
Figure 4.1: Structure of matrix equation.
are determined by integrating the kinematic boundary conditions in time. The same
procedure is followed for (or
n
) in all collocation points.
Suppose, for instance, that at a certain time level t
n
the following set of free surface
variables is known:
the free surface shape, i.e. the position of all free surface collocation points,
the free surface potential, i.e. the value of in these points, and
the free surface normal velocity, i.e. the normal derivative of in these points.
The rst step is the determination of the tangential velocities using a nite dierence
scheme. Next, the free surface velocity is computed from the normal and tangential
velocities. Then, the positions of all collocation points and the corresponding free surface
potential at the next time level t
n+1
= t
n
+ t are obtained by time integration of the
free surface conditions (4.4-4.5):
(t
n+1
) = (t
n
) + t
D
Dt
(t
n
) +O

t
2

= (t
n
) + t

1
2
( ) gz

(t
n
) +O

t
2

, (4.9)
x
F
(t
n+1
) = x
F
(t
n
) + t
Dx
F
Dt
(t
n
) +O

t
2

= x
F
(t
n
) + t (t
n
) +O

t
2

. (4.10)
This time integration is carried out with a fourth order Runge-Kutta method, which
implies that three intermediate time levels are used for each time step t. Once the
new positions of the collocation points are known, the free surface shape is updated in
terms of new spline data. Finally, the solution of integral equation (4.7) provides the
new values of the normal velocity i.e. /n in all updated free surface collocation
points.
Summarizing, the step-by-step method for nonlinear gravity wave simulations consists of
a repetition of the following sequence of operations:
1. calculation of the source and dipole coecients at time t
n
2. substitution of for Dirichlet boundaries and
n
for Neumann boundaries at time
t
n
4.4. EXTENSION TO NONLINEAR SHIP MOTIONS 55
3. solution of the BIE at time t
n
, giving
n
for Dirichlet boundaries and for Neumann
boundaries
4. calculation of the tangential velocities at time t
n
from using nite dierences
5. calculation of the boundary velocity at time t
n
from the tangential and normal
velocities
6. integration of the time dependent boundary conditions, giving the collocation point
positions and for Dirichlet boundaries and
n
for Neumann boundaries at time
t
n+1
7. spline approximation of the geometry at time t
n+1
In the next section we demonstrate how this basic scheme can be extended to nonlinear
interactions of water waves with oating bodies in either forced or free motion.
4.4 Extension to nonlinear ship motions
In section 2.5 it was stated that the presence of a oating body either partly or
totally submerged implies that extra equations are needed for a correct mathematical
description of the uid-body interaction. The problem of nonlinear ship motions at
zero forward speed is usually described with two orthonormal coordinate systems. One
system is xed in space and set up by the unit vectors e
1
, e
2
, e
3
, where e
3
is vertical
and pointing upwards. The origin of this system is initially located at the centre of mass
of the body, denoted by G. The other system is attached to the body; its origin is located
at G and it is set up by the unit vectors e
1

, e
2

, e
3

along the body principal axes of


inertia. The position and orientation of the body are specied by the three-component
vectors x
G
(corresponding to surge, sway, and heave motions) and

G
(corresponding to
roll, pitch, and yaw motions) respectively. Figure 4.2 illustrates these denitions.
Figure 4.2: Body motion and geometry denitions.
56 CHAPTER 4. ALGORITHM FOR WAVE-BODY SIMULATIONS
The motion of the uid at the wetted (submerged) part of the body surface, denoted by
S, is determined by the Neumann boundary condition

n
= n =

x
G
n + (r n)

G
, (4.11)
where n is the unit normal vector on S, and r = x
S
x
G
, with x
S
on S. In case of a
oating body under forced motion,

x
G
and

G
are known, and boundary condition (4.11)
completes the set of governing equations. The body position and orientation can be
integrated in time as follows:
x
G
(t
n+1
) = x
G
(t
n
) + t

x
G
(t
n
) +O

t
2

, (4.12)

G
(t
n+1
) =

G
(t
n
) + t

G
(t
n
) +O

t
2

. (4.13)
However, if the body is oating freely, then the translational and rotational velocities of
the body are not known a priori and have to be determined as part of the solution. In
this case, the mathematical model is completed with (4.11) and the equations of motion
for a rigid body with, say, mass M and moment of inertia

I about its principal axes:
M

x
G
=

pndS Mge
3
,

I

G
=

p (r n) dS , (4.14)
where denes a component-wise product of two vectors; see section 2.5 for a denition.
The pressure along S is obtained from the nonlinear Bernoulli equation:
p =

t
+
1
2
( ) +gz

. (4.15)
Since can be computed from the solution of Laplaces equation for , the calculation of
the body accelerations reduces to the computation of
t
along the wetted body surface S.
An obvious approach towards the calculation of
t
is to use a nite (backward) dierence
method. Due to the motion of the body,
t
then has to be determined from the material
derivative of :

t
=
D
Dt
. (4.16)
This principle can be used in combination with a predictor-corrector method, or other
complex numerical schemes. With this choice however, it is inevitable that one makes
use of data from previous time levels, which may give rise to numerical instabilities
see, for instance, Isaacson (1982).
A less obvious way to calculate
t
along S, is to use the fact that besides , its partial
time derivative
t
satises Laplaces equation as well. This enables us to derive an extra
boundary integral equation for
t
, similar to (4.7):
1
2

t
(x) =

; x

G
n

; x

dS

(4.17)
The discretized version of this second BIE, applied in collocation point i, reads:
1
2

i
t
=
N

j=1

C
ij
s

j
tn
+C
ij
d

j
t

(4.18)
4.4. EXTENSION TO NONLINEAR SHIP MOTIONS 57
Analogous to the solution of the rst BIE for , substitution of
t
for Dirichlet boundaries
and
tn
for Neumann boundaries into (4.18) yields a system of N linear equations.
However, since the body is oating freely, both
t
and
tn
are unknown along S. Hence,
a complementary set of equations is needed to ensure that the number of equations equals
the number of unknowns.
Up to now, the equations of motion (4.14) have not been used; in Appendix B we show
that these equations can be transformed into a boundary integral equation involving
t
and
tn
on S only:

tn
(x) +

x,

dS

= (x) (4.19)
After discretization, a set of linear equations connecting
t
and
tn
on S is obtained:

i
tn
+
N
S

j=1
C
ij
k

j
t
=
i
(4.20)
where collocation point i is on S and summation is over all N
S
collocation points on
S. We shall refer to this system of N
S
equations as the discretized equations of motion
henceforth.
C
ij
s
C
ij
d

1
2

ij
C
ij
s
0 C
ij
k

ij
Dirichlet Neumann Extra BIE
Floating Body

j
tn

j
t

j
tn
=

i
Figure 4.3: Structure of extra matrix equation.
The discretized extra BIE (4.18) and the discretized equations of motion (4.20) together
form a system of N +N
S
linear equations involving
t
and
tn
on all boundaries, includ-
ing the free surface. Substitution of
t
for Dirichlet boundaries and
tn
for Neumann
boundaries (excepting S) yields a linear system, where the number of equations equals
the number of unknowns.
The structure of this matrix equation is shown in Figure 4.3. Note that this matrix
system has N
S
extra equations compared to the standard matrix equation see Fig-
ures 4.1 and 4.3 due to the double number (i.e. 2N
S
) of unknowns on S.
The extra inuence coecients C
ij
k
have some favourable properties; rst of all, they
58 CHAPTER 4. ALGORITHM FOR WAVE-BODY SIMULATIONS
depend upon the geometry of the body only. More important however, is that the coef-
cient expressions are regular and symmetric, making their calculation straightforward
(see Appendix B). If the collocation points on S have xed positions, the coecients C
ij
k
have to be computed only once; in general, a redistribution of these collocation points
will be necessary from time to time, thus requiring an update of these coecients.
The extra components
i
in the right-hand side vector are computed from known
body variables, such as the potential and its spatial derivatives up to second order on
S (see Appendix B). Since the calculation of the source coecients C
ij
s
and the dipole
coecients C
ij
d
being the most expensive part of the computations has already
been carried out for the rst boundary integral equation (4.7), we arrive at the conclu-
sion that the extra computational eort is relatively small; it is mainly caused by the
solution of the extra matrix equation. Thus a direct, accurate, and ecient method for
the calculation of
t
on S is developed.
Once
t
is known on S, the hydrodynamic forces can be computed directly from pres-
sure integrations. The resulting accelerations can then be used to determine the body
velocities at the next time level:

x
G
(t
n+1
) =

x
G
(t
n
) + t

x
G
(t
n
) +O

t
2

, (4.21)

G
(t
n+1
) =

G
(t
n
) + t

G
(t
n
) +O

t
2

. (4.22)
Naturally, the aforementioned Runge-Kutta time stepping scheme is applied here. Fi-
nally, the Neumann contact boundary condition (4.11) at time t
n+1
is determined from
these updated body velocities.
Summarizing, the proposed step-by-step method for nonlinear wave-body simulations
consists of a repetition of the following sequence of operations:
1. calculation of the source and dipole coecients at time t
n
2. substitution of for Dirichlet boundaries and
n
for Neumann boundaries at time
t
n
3. solution of the rst BIE at time t
n
, giving
n
for Dirichlet boundaries and for all
Neumann boundaries, i.e. including the body surface S
4. calculation of the tangential velocities at time t
n
from
5. calculation of the boundary velocity at time t
n
from the tangential and normal
velocities
6. calculation of the extra inuence coecients at time t
n
7. calculation of the second order tangential derivatives of and the rst order tan-
gential derivatives of
n
on S at time t
n
8. substitution of
t
for Dirichlet boundaries and
tn
for Neumann boundaries, ex-
cepting S, at time t
n
9. solution of the second BIE at time t
n
, giving
tn
for Dirichlet boundaries and
t
for all Neumann boundaries, i.e. including S
4.4. EXTENSION TO NONLINEAR SHIP MOTIONS 59
10. calculation of the body accelerations from the hydrodynamic equations of motion
and pressure integrations over S
11. integration of the time dependent boundary conditions and the body position and
orientation, giving the collocation point positions and for Dirichlet boundaries
and
n
for all Neumann boundaries at time t
n+1
12. integration of the body velocities, giving the Neumann condition on S at time t
n+1
13. spline approximation of the geometry at time t
n+1
The approach outlined above resembles the somewhat earlier work of Cointe (1989), who
developed a panel method for two-dimensional linear wave-body interactions. However,
his computations for an extincting circular cylinder show poor agreement with the ana-
lytical linear predictions. See also chapter 7, where our results concerning freely oating
bodies are presented.
Recently in the course of writing this thesis and after completion of the calculations
described herein it has come to the authors attention that Tanizawa and Sawada
(1990) have successfully used the above technique in their boundary element method
for two-dimensional nonlinear wave-body interactions. Unfortunately, their work is less
accessible since it was written in Japanese.
We close this chapter with a few remarks on the numerical algorithm:
The essential feature of the algorithm proposed here is that apart from numerical
spatial discretization errors the dynamic equilibrium of the uid and the body
is preserved at all times. By using an extra BIE to compute
t
on the wetted
body surface, unnecessary time discretization errors are avoided, thus preserving
the stability and accuracy of the algorithm.
Another important feature of this algorithm is that it is not limited to two dimen-
sions; the derivations in Appendices A and B contain no two-dimensional elements
or assumptions.
The above approach permits more than one freely oating body; a second body
merely introduces an extra set of equations of motion which can be transformed into
a boundary integral equation over the wetted body surface. After discretization,
the two systems of linear equations (one for body I, and one for body II) are added
to the discretized extra BIE for
t
.
Note that in case of a body under forced motion, the hydrodynamic forces acting on
the body can be computed likewise;
tn
on S can be computed from the prescribed
body velocities and accelerations, and then
t
is solved from the extra BIE. In this
case, the discretized equations of motion should not be added to the extra BIE. The
(minor) advantage of this approach is that the hydrodynamic forces and moments
can be computed during the computations, and postprocessing is not necessary
to compute
t
. Kang and Gong (1990) employ this technique for this particular
purpose.
Restoring forces and moments acting on the body may also be incorporated in the
equations of motion (4.14), thus allowing for the numerical simulation of water
waves interacting with moored objects.
60 CHAPTER 4. ALGORITHM FOR WAVE-BODY SIMULATIONS
4.5 Bibliography
Broeze, J. 1993. Analysis of an Eective Method for 3-D Flow Computations with a
Nonlinear Free Surface. Ph.D. thesis, University of Twente. Enschede, The Netherlands.
Canuto, C., Hussaini, M.Y., Quarteroni, A., and Zang, T.A. 1988. Spectral
Methods in Fluid Dynamics. Springer-Verlag.
Chan, R.K.-C., and Chan, F.W.-K. 1980. Numerical solution of transient and
steady free-surface ows about a ship of general hull shape. Proceedings of the Thir-
teenth Symposium on Naval Hydrodynamics, Tokyo, Japan.
Cointe, R. 1989. Quelques Aspects de la Simulation Numerique dun Canal `a Houle.
Ph.D. thesis, Ecole des Ponts et Chaussees. Paris, France. (in French)
Cointe, R., Geyer, P., King, B., Molin, B., and Tramoni, M. 1990. Nonlinear
and linear motions of a rectangular barge in a perfect uid. Proceedings of the Eigh-
teenth Symposium on Naval Hydrodynamics, Ann Arbor, Michigan.
Dommermuth, D.G., and Yue, D.K.P. 1987a. Numerical simulations of nonlinear
axisymmetric ows with a free surface. Journal of Fluid Mechanics 178:195-219.
Dommermuth, D.G., and Yue, D.K.P. 1987b. A high-order spectral method for the
study of nonlinear gravity waves. Journal of Fluid Mechanics 184:267-288.
Ehlers, J. 1991. Numerische Simulation in Schwerewellen schwimmender Korper mit
einem Randelement-Zeitschrittverfahren. Masters thesis, Technische Universitat Hamburg-
Harburg, Arbeitsbereich Meerestechnik II - Strukturmechanik. Hamburg, Germany. (in
German)
Faltinsen, O.M. 1977. Numerical solutions of transient nonlinear free surface motion
outside or inside moving bodies. Proceedings of the Second International Conference on
Numerical Ship Hydrodynamics, Berkeley, California.
Fenton, J.D., and Rienecker, M.M. 1982. A Fourier method for solving nonlinear
water-wave problems; application to solitary-wave interactions. Journal of Fluid Mechan-
ics 118:411-443.
Greenhow, M. 1987. Wedge entry into initially calm water. Applied Ocean Research
9(4):214-223.
Greenhow, M., and Lin, W.-M. 1985. Numerical simulation of nonlinear free sur-
face ows generated by wedge entry and wavemaker motion. Proceedings of the Fourth
International Conference on Numerical Ship Hydrodynamics, Washington, DC.
Haussling, H.J., and Coleman, R.M. 1977. Finite-dierence computations using
boundary-tted coordinates for free-surface potential ows generated by submerged bodies.
Proceedings of the Second International Conference on Numerical Ship Hydrodynamics,
Berkeley, California.
Haussling, H.J., and Coleman, R.M. 1979. Nonlinear water waves generated by an
accelerated circular cylinder. Journal of Fluid Mechanics 92(4):757-781.
Hussaini, M.Y., and Zang, T.A. 1987. Spectral methods in uid dynamics. Annual
Review of Fluid Mechanics 19:339-367.
Isaacson, M. de St.Q. 1982. Nonlinear-wave eects on xed and oating bodies. Jour-
nal of Fluid Mechanics 120:267-281.
4.5. BIBLIOGRAPHY 61
Kang, C.-G., and Gong, I.-Y. 1990. A numerical solution method for three-
dimensional nonlinear free surface problems. Proceedings of the Eighteenth Symposium
on Naval Hydrodynamics, Ann Arbor, Michigan.
Lin, W.-M., Newman, J.N., and Yue, D.K.P. 1984. Nonlinear forced motions
of oating bodies. Proceedings of the Fifteenth Symposium on Naval Hydrodynamics,
Hamburg, Germany.
Longuet-Higgins, M.S., and Cokelet, E.D. 1976. The deformation of steep sur-
face waves on water. I. A numerical method of computation. Proceedings of the Royal
Society of London, Series A, 350:1-26.
Romate, J.E. 1988. Local error analysis in 3-D panel methods. Journal of Engineering
Mathematics 22:123-142.
Romate, J.E. 1989. The Numerical Simulation of Nonlinear Gravity Waves in Three
Dimensions using a Higher Order Panel Method. Ph.D. thesis, University of Twente.
Enschede, The Netherlands.
Romate, J.E. 1990. Local error analysis of three-dimensional panel methods in terms
of curvilinear surface coordinates. SIAM Journal on Numerical Analysis 27(2):529-542.
Saubestre, V. 1991. Numerical simulation of transient nonlinear free-surface ows
with body interaction. OMAE - Volume I-A, Oshore Technology ASME.
Tanizawa, K., and Sawada, H. 1990. A numerical method for nonlinear simulation
of 2-D body motions in waves by means of B.E.M. Journal of the Society of Naval Ar-
chitects of Japan 168:221-226.
Telste, J.G. 1985. Calculation of uid motion resulting from large-amplitude forced
heave motion of a two-dimensional cylinder in a free surface. Proceedings of the Fourth
International Conference on Numerical Ship Hydrodynamics, Washington, DC.
Vinje, T., and Brevig, P. 1981a. Breaking waves on nite water depths. A numerical
study. Norwegian Hydrodynamic Laboratories, Report R-111.81, Trondheim, Norway.
Vinje, T., and Brevig, P. 1981b. Nonlinear, two-dimensional ship motions. Norwe-
gian Hydrodynamic Laboratories, Report R-112.81, Trondheim, Norway.
Vinje, T., and Brevig, P. 1981c. Nonlinear ship motions. Proceedings of the Third
International Conference on Numerical Ship Hydrodynamics, Paris, France.
Vinje, T., Maogang, X., and Brevig, P. 1982. A numerical approach to nonlinear
ship motion. Proceedings of the Fourteenth Symposium on Naval Hydrodynamics, Ann
Arbor, Michigan.
Wang, X.M., and Spaulding, M.L. 1988. A two-dimensional potential ow model
of the wave eld generated by a semisubmerged body in heaving motion. Journal of Ship
Research 32(2):83-91.
Yeung, R.W., and Ananthakrishnan, P. 1992. Oscillation of a oating body in a
viscous uid. Journal of Engineering Mathematics 26:211-230.
Yeung, R.W., and Wu, C.-F. 1989. Nonlinear wave-body motion in a closed domain.
Computers and Fluids 17(2):351-370.
Yim, B. 1985. Numerical solution for two-dimensional wedge slamming with a nonlinear
free-surface condition. Proceedings of the Fourth International Conference on Numerical
Ship Hydrodynamics, Washington, DC.
62 CHAPTER 4. ALGORITHM FOR WAVE-BODY SIMULATIONS
Zandbergen, P.J., Broeze, J., and van Daalen, E.F.G. 1992. A panel method
for the simulation of nonlinear gravity waves and ship motions. In Advances in Boundary
Element Techniques. Springer-Verlag.
4.5. BIBLIOGRAPHY 63
But straightway the morning star rose above the topmost peaks and the
breeze swept down; and quickly did Tiphys urge them to go aboard and avail
themselves of the wind. And they embarked eagerly forthwith; and they drew
up the ships anchors and hauled the ropes astern. And the sails were bellied
out by the wind, and far from the coast were they joyfully borne past the
Posideian headland. But at the hour when gladsome dawn shines from heaven,
rising from the east, and the paths stand out clearly, and the dewy plains
shine with a bright gleam, then at length they were aware that unwittingly
they had abandoned those men.
5
And a erce quarrel fell upon them, and
violent tumult, for that they had sailed and left behind the bravest of their
comrades . . .
But to them appeared Glaucus from the depths of the sea, the wise inter-
preter of divine Nereus, and raising aloft his shaggy head and chest from his
waist below, with sturdy hand he seized the ships keel, and then cried to the
eager crew:
Why against the counsel of mighty Zeus do ye purpose to lead bold Her-
acles to the city of Aeetes? At Argos it is his fate to labour for insolent
Eurystheus and to accomplish full twelve toils and dwell with the immortals,
if so be that he bring to fullment a few more yet; wherefore let there be no
vain regret for him. Likewise it is destined for Polyphemus to found a glorious
city at the mouth of Cius among the Mysians and to ll up the measure of
his fate in the vast land of the Chalybes. But a goddess-nymph through love
has made Hylas her husband, on whose account those two wandered and were
left behind.
He spake, and with a plunge wrapped him about with the restless wave;
and round him the dark water foamed in seething eddies and dashed against
the hollow ship as it moved through the sea . . .
Argonautica, Book I, Verses 1273-1286 and 1310-1328.
5
i.e. Hylas, Heracles, and Polyphemus
64 CHAPTER 4. ALGORITHM FOR WAVE-BODY SIMULATIONS
Part II
Numerical Results
65
Chapter 5
Impulsive Wavemaker Motion
Here were the oxstalls and farm of Amycus, the haughty king of the Be-
brycians, whom once a nymph, Bithynian Melie, united to Poseidon Geneth-
lius, bare the most arrogant of men; for even for strangers he laid down an
insulting ordinance, that none should depart till they had made trial of him
in boxing; and he had slain many of the neighbours. And at that time too he
went down to the ship and in his insolence scorned to ask them the occasion
of their voyage, and who they were, but at once spake out among them all:
Listen, ye wanderers by sea, to what it bets you to know. It is the rule
that no stranger who comes to the Bebrycians should depart till he has raised
his hands in battle against mine. Wherefore select your bravest warrior from
the host and set him here on the spot to contend with me in boxing. But if
ye pay no heed and trample my decrees under foot, assuredly to your sorrow
will stern necessity come upon you.
Thus he spake in his pride, but erce anger seized them when they heard
it, and the challenge smote Polydeuces most of all. And quickly he stood forth
his comrades champion, and cried:
Hold now, and display not to us thy brutal violence, whoever thou art; for
we will obey thy rules, as thou sayest. Willingly now do I myself undertake
to meet thee.
Argonautica, Book II, Verses 1-24.
67
68 CHAPTER 5. IMPULSIVE WAVEMAKER MOTION
5.1 Introduction
In the study of water waves associated with surface piercing bodies, a number of linear
and nonlinear analyses has indicated that the potential solution may become singular at
the point where the oating body intersects the free surface. Physically, this singularity
is plausible in view of the splashing which occurs when, for instance, the bow of a ship
slams the water plane. From a mathematical point of view, this singular behaviour is
due to a conuence of the boundary conditions on the free surface and the oating body.
This singular behaviour is of great interest in the numerical simulation of ship motions
in water waves. In a discrete sense the boundary conditions must be satised on exact
boundaries, but the position of both the free surface and the oating body are not known
in advance; this explains the diculty in determining the position of the intersection point
and the local uid ow. Without detailed knowledge of the singular behaviour of the
displacement of, and the velocity potential at the free surface, it is hardly possible to
develop a stable and accurate numerical algorithm.
To reduce the complexity of this hydrodynamic problem, let us concentrate only on
the local region near the intersection point. As a simplied model, the following situation
is considered; in a semi-innite wave tank, a piston-type wavemaker starts to move from
rest at time t = 0 at a specied horizontal velocity, and the subsequent motion of the free
surface is sought. Assumptions on the wavemaker motion and the uid ow are made
later on.
The major diculty in the nonlinear analysis of this wavemaker problem is that both
the kinematic condition and the dynamic condition must be satised at the free surface.
In the Eulerian description of the wave motion, the free surface position is unknown, and
usually series expansions are used to overcome this diculty. For instance, Peregrine
(1972) used perturbation series in small time t; employing a moving coordinate system
attached to the wavemaker, he revealed a singularity in the free surface elevation at
the intersection point, which behaves like t log x. In a similar way, Chwang (1983)
developed a nonlinear theory to determine the hydrodynamic pressure on an accelerating
vertical plate.
1
Using a xed coordinate system, he suggested that the singularity lies
outside the uid domain and, hence, is of no physical interest. However, the singular
behaviour in the small time solution was conrmed by Lin (1984), who employed a
Lagrangian description for the wavemaker problem.
The absence of gravity and the singular behaviour of the small time solutions men-
tioned above, induced Roberts (1987) to carry out a rigorous analysis of free surface
waves generated by a wavemaker, using both small time and small Froude number ex-
pansions. For power-law displacement of a vertical plate, Roberts showed that the ow
develops a disperging wavetrain which emanates instantaneously from the intersection
point. He also found that the solution varies considerably close to the intersection point
and presented a self-similar formulation to describe this wiggling behaviour.
To avoid the articial singularity at the intersection point introduced by small expan-
sions, Joo, Schultz, and Messiter (1990) developed a Fourier-integral method for small
Froude number, which also allows the study of capillary eects. In the absence of surface
tension, their method reveals small-scale wiggles near the wavemaker, in agreement with
Roberts local solution for small time. In addition, it was found that the contact angle
1
Similar studies have also been carried out by Chwang and Wang (1984) for an accelerating rectangular
or circular container, and by Wang and Chwang (1989) for an accelerating vertical surface-piercing
cylinder.
5.2. NONLINEAR ANALYSIS 69
has a jump at t = 0; surface tension suppresses these wiggles and maintains the contact
angle at its initial value. Systematic studies of capillary eects along similar lines have
been carried out by Hocking and Mahdmina (1991) and Miles (1991).
In the past decade, the nonlinear wavemaker problem
2
has also been addressed numer-
ically. In particular, Lins (1984) doctoral thesis was devoted to this subject; earlier,
Vinje and Brevig (1981) reported numerical instabilities originating from the intersec-
tion of the free surface and a oating body in their ship motion computations. Following
their approach, Lin developed a boundary element method based on Cauchys integral
theorem, see section 4.2 for two-dimensional free surface ow simulations. To Lins
opinion, the numerical diculties could be solved by specifying both the velocity poten-
tial and the stream function at the intersection point. This approach was applied to the
problem of an impulsively started wavemaker; the numerical results show that accurate
global calculations of the potential along the wavemaker are obtained even when rela-
tively coarse grids are used. The local behaviour of the free surface is properly described
and the singularity is accommodated.
In view of the extension of Romates panel method for water waves to nonlinear ship
motions, it was felt necessary to test our computer code for the impulsive wavemaker
problem. The present method diers from Lins approach in that our collocation points
coincide with the panel midpoints. Hence, in our method there is no direct conuence
of boundary conditions in the intersection point, and it will appear that no special local
treatment is needed. In this study, attention is focused on the ability of the code to
predict the initial behaviour of the potential along the wavemaker and the free surface
elevation for small time, since these variables determine the hydrodynamic forces exerted
by the uid. We do not aim to reproduce the above-mentioned small-scale wiggles or
capillary eects.
The present chapter is an elaborated study, based on a recent paper written by van
Daalen and Huijsmans (1991). In section 5.2 we derive analytical solutions for the initial
behaviour of the potential and the tangential velocities on the wavemaker and the bot-
tom, using small time expansions; series expressions for these solutions are transformed
into compact integral expressions which are more suitable for numerical evaluation. An
analytical expression for the small time free surface elevation which conrms the log-
arithmic singularity is also derived. These solutions allow a direct verication of the
numerical results obtained with the TIPHYS-code for two-dimensional water-wave simu-
lations; this part is presented in section 5.3. Concluding remarks are given in section 5.4.
5.2 Nonlinear analysis
Consider the conguration depicted in Figure 5.1, where the two-dimensional uid do-
main extends to innity to the right and is bounded in vertical direction by a horizontal
bottom and a free surface; at the left end of the wave tank a piston-type wavemaker is
situated. The origin O of the coordinate system x, z is initially located at the intersec-
2
Some related problems are, for example: a two-dimensional cylinder in a current, see Grosenbaugh
and Yeung (1988); the oblique water entry of a two-dimensional prole, see Korobkin (1988); the vertical
water entry of a horizontal circular cylinder see Greenhow (1988) or a spherical projectile see
Miloh (1981, 1991a). The problem of an oblique water entry of a rigid sphere has also been discussed by
Miloh (1991b). Finally, we refer to the review article on water impact problems, written by Korobkin
and Pukhnachov (1988).
70 CHAPTER 5. IMPULSIVE WAVEMAKER MOTION
tion of the wavemaker and the free surface. At time t = 0, the wavemaker impulsively
starts to move towards the uid with a constant velocity U.
Figure 5.1: Impulsive wavemaker problem: geometry denition.
5.2.1 Governing equations and small time expansions
Firstly, the governing equations for this particular problem are derived; under the as-
sumptions of an inviscid, incompressible uid and an irrotational ow, the velocity po-
tential satises

2
= 0 for x > Ut , h < z < (x; t) (5.1)
i.e. Laplaces equation throughout the transient uid domain.
Due to the motion of the wavemaker, the free surface is changed from its undisturbed
level z = 0 to a new position z = (x; t), where the nonlinear free surface conditions
read

t
+
1
2
( ) +gz = 0

t
+
x

z
= 0

at z = (x; t) (5.2)
where a sux denotes partial dierentiation, i.e.
t
= /t et cetera.
At the bottom the vertical velocity must vanish; this is expressed by

z
= 0 at z = h (5.3)
The boundary condition on the moving wavemaker reads

x
= U at x = Ut (5.4)
which completes the set of governing equations for the impulsive wavemaker problem; it
is assumed that and vanish for x such that a well-posed problem is obtained.
5.2. NONLINEAR ANALYSIS 71
In order to obtain small time solutions for the velocity potential and the vertical deviation
of the free surface, and are written as power series in t:
(x, z; t) =

m=0

m
(x, z) t
m
(5.5)
(x; t) =

m=0

m
(x) t
m
(5.6)
The governing equations for the
m
s and
m
s are obtained by substitution of (5.5-5.6)
into the eld equation (5.1) and the boundary conditions (5.2-5.4).
Obviously, each
m
is a harmonic function, i.e. it satises Laplaces equation (5.1):

m
= 0 , m = 0, 1, 2, . . . (5.7)
Similarly, each
m
meets the zero-ux bottom condition (5.3), that is

m
z
= 0 at z = h , m = 0, 1, 2, . . . (5.8)
Substitution of (5.5) into the wavemaker condition (5.4) and developing Taylor series
round x = 0 up to rst order gives
U =

0
x
at x = 0 , (5.9)
U

2

0
x
2
=

1
x
at x = 0 . (5.10)
In a similar way, substitution of (5.5-5.6) into the nonlinear free surface conditions (5.2)
and developing Taylor series round z = 0 yields
0 =
0
at z = 0 , (5.11)

1
2

0
z

2
=
1
at z = 0 , (5.12)
and
0 =
0
at z = 0 , (5.13)

0
z
=
1
at z = 0 , (5.14)
1
2

1
z
+
1
2

0
z
2
=
2
at z = 0 . (5.15)
In the next paragraph we deduce leading order solutions for and , satisfying (5.7-5.15).
5.2.2 Leading order solutions
The leading (zeroth) order potential
0
, satisfying (5.7-5.9) and (5.11) is obtained in a
straightforward manner by the Fourier series method:

0
(x, z) =
2U
h

n=1
(1)
n
k
2
n
e
k
n
x
cos k
n
(z +h) (5.16)
72 CHAPTER 5. IMPULSIVE WAVEMAKER MOTION
where the coecients k
n
are dened as
k
n
=
(2n 1)
2h
, n = 1, 2, 3, . . . (5.17)
The next (rst order) term in the series expansion for the potential,
1
, satises (5.7-5.8),
(5.10), and (5.12). Using the zeroth order result (5.16), and employing the method of
Fourier cosine transformation, we obtain

1
(x, z) =
2U
2
h

n=1
(1)
n
k
n
e
k
n
x
cos k
n
(z +h)
4U
2

n=2
A
n
B
n
(x, z)
(5.18)
where the coecients A
n
and the functions B
n
(x, z) are dened as
A
n
= h
2
n1

l=1
(k
l
k
nl
)
1
, n = 2, 3, 4, . . . (5.19)
B
n
(x, z) =

2
n
+
2
cosh (z +h)
cosh h
cos xd , n = 2, 3, 4, . . . (5.20)
with

n
=
(n 1)
h
, n = 2, 3, 4, . . . (5.21)
The zeroth order term in the free surface elevation is, naturally, given by

0
(x) = 0 (5.22)
The rst non-trivial term is obtained from substitution of (5.16) into (5.14):

1
(x) =
2U
h

n=1
e
k
n
x
k
n
. (5.23)
Dierentiation with respect to x gives
d
1
dx
(x) =
2U
h

n=1
e
k
n
x
=
U
h
sinh
1

x
2h

, (5.24)
where we used the identity see Gradshteyn and Ryzhik (1980)

n=1
e
(2n1)y
=
1
2
sinh
1
y for y > 0 . (5.25)
Integration with respect to x yields the well-known result

1
(x) =
2U

log

tanh

x
4h

(5.26)
5.2. NONLINEAR ANALYSIS 73
The second order term in the series expansion for the free surface elevation is obtained
by substitution of (5.16), (5.18), and (5.26) into (5.15):

2
(x) =
U
2
h

n=1
e
k
n
x

2U
2

n=2
A
n

2
n
+
2
tanh hcos xd . (5.27)
When we use identity (5.25) and the following denition:
C
n
(x) =
B
n
z
(x, 0) =

2
n
+
2
tanh hcos xd , n = 2, 3, 4, . . .
(5.28)
we arrive at a more compact expression for
2
, namely

2
(x) =
U
2
2h
sinh
1

x
2h

2U
2

n=2
A
n
C
n
(x) (5.29)
The above set of leading order solutions for and will be used in the next paragraph to
derive expressions for the initial and small time behaviour of the ow along the wavemaker
and the bottom, and the free surface elevation.
5.2.3 Initial and small time behaviour
From (5.16) the potential on the wavemaker at t = 0 is found to be

0
(0, z) =
2U
h

n=1
(1)
n
k
2
n
cos k
n
(z +h) . (5.30)
Dierentiation with respect to z gives the initial vertical velocity along the wavemaker:

0
z
(0, z) =
2U
h

n=1
(1)
n
k
n
sin k
n
(z +h) . (5.31)
Using the identity see Gradshteyn and Ryzhik (1980)

n=1
(1)
n1
sin (2n 1) y
2n 1
=
1
2
log

tan

4
+
y
2

for [y[ <



2
, (5.32)
gives an expression which is more suitable for numerical evaluation:

0
z
(0, z) =
2U

log

tan

4

2 +
z
h

(5.33)
It is easy to see that the initial vertical velocity at the intersection of the wavemaker and
the bottom vanishes:

0
z
(0, h) = 0 (5.34)
74 CHAPTER 5. IMPULSIVE WAVEMAKER MOTION
where the initial vertical velocity at the intersection of the wavemaker and the free surface
tends to innity:

0
z
(0, z) for z 0 (5.35)
An integral expression for
0
(0, z) is obtained from integration of (5.33) with respect to
z and using
0
(0, 0) = 0:

0
(0, z) =
2U

0
log

tan

4

2 +

h

d . (5.36)
Substituting z = h in (5.30), we nd by denition, see Gradshteyn and Ryzhik
(1980)

0
(0, 0) =
8UhG

2
(5.37)
where G 0.916 is Catalans constant. This leads us to an alternative integral expression
for the initial potential distribution on the wavemaker:

0
(0, z) =
2U

h
log

tan

4

2 +

h

d
4hG

(5.38)
From (5.16) it follows that the analytical solution for the initial potential distribution on
the bottom z = h is given by

0
(x, h) =
2U
h

n=1
(1)
n
k
2
n
e
k
n
x
. (5.39)
Dierentiating this expression twice with respect to x, we obtain

0
x
2
(x, h) =
2U
h

n=1
(1)
n
e
k
n
x
=
U
h
cosh
1

x
2h

, (5.40)
where we used the identity see Gradshteyn and Ryzhik (1980)

n=1
(1)
n1
e
(2n1)y
=
1
2
cosh
1
y for y > 0 . (5.41)
Integrating (5.40) with respect to x and using

0
x
(x, h) 0 for x (5.42)
gives

0
x
(x, h) =
4U

arctan e
x/2h
(5.43)
5.2. NONLINEAR ANALYSIS 75
Integration with respect to x, using (5.37), yields

0
(x, h) =
8Uh

e
x/2h
arctan

d G

(5.44)
The initial vertical velocity of the free surface is easily deduced from (5.14) and (5.26):

0
z
(x, 0) =
2U

log

tanh

x
4h

(5.45)
In order to calculate the hydrodynamic force on the wavemaker at t = 0, we expand the
Bernoulli pressure p in the time t:
p (x, z; t) =

t
+
1
2
( ) +gz

m=0
p
m
(x, z) t
m
. (5.46)
Substituting (5.5) and retaining zeroth order terms only, gives
p
0
(x, z) =

1
+
1
2
(
0

0
) +gz

(5.47)
The rst order potential solution
1
along the wavemaker reads

1
(0, z) =
2U
2
h

n=1
(1)
n
k
n
cos k
n
(z +h)
4U
2

n=2
A
n
B
n
(0, z) . (5.48)
Using the identity see Gradshteyn and Ryzhik (1980)

n=1
(1)
n1
cos (2n 1) y
2n 1
=

4
for 0 < y <

2
, (5.49)
the following expression for
1
(0, z) is obtained:

1
(0, z) = U
2

1 +
4

n=2
A
n

2
n
+
2
cosh (z +h)
cosh h
d

(5.50)
With
0
/x = U on the wavemaker, and expression (5.33) for
0
/z at x = 0, the
initial pressure distribution on the wavemaker can be computed analytically (and hence,
the initial force acting on the wavemaker).
Finally, from (5.22) and (5.26) it follows that to rst order in time
(x; t) =
Ut

log

tanh

x
4h

(5.51)
76 CHAPTER 5. IMPULSIVE WAVEMAKER MOTION
At the wavemaker we have x = Ut. Then, using tanh y = O(y), we nd
lim
t0
(x = Ut; t) = lim
t0

Ut

log

Ut
4h

= 0 , (5.52)
since
lim
y0
y log y = 0 . (5.53)
5.3 Numerical results
As stated in the introduction, the present numerical study is concerned with the impulsive
motion of a vertical wavemaker; at t = 0, the horizontal velocity is a step function and
as a consequence the wavemaker acceleration is innite. Analytically, a logarithmic
singularity is expected at the intersection of the free surface and the wavemaker. From
a numerical point of view this case is rather critical because the high gradients near the
intersection point may ruin the stability of the computational scheme.
For the actual runs, the length of the wave tank is L = 10 m, the water depth is
h = 1 m, the wavemaker velocity is U = 1 m/s, and the time t is from zero to 0.2 s.
Comparative tests with a doubled tank length show that during this time interval there
is no signicant reection from the vertical wall downstream.
3
The boundary integral equation is solved using a wide range of elements which are dis-
tributed over the free surface, the wavemaker, the bottom, and the right lateral boundary
in accordance with the distribution shown in Figure 5.2.
s s s s s s s s s s s s s s s s s s s s s s s s s
s s s s s s s s s s s s s s s s s s s s s s s s s
s
s
s
s
s
s
s
s
s
s
s
s
s
s
s
s
N
B
N
F
N
W
N
R
N
F
: N
W
: N
B
: N
R
= 25 : 8 : 25 : 8
Figure 5.2: Element distributions for impulsive wavemaker problem.
For the wavemaker and the free surface special care is taken in the distribution of the
collocation points. For the wavemaker, the actual grid is obtained from an equidistant
distribution x = 0, z
i
= (i 1/2) z, with z = 1/N
W
, through the power trans-
formation z z
3/2
. The initial grid on the free surface is similarly generated from a
uniform distribution.
The choice of the time step t is rather critical; it is anticipated that shortly after
the impulsive start-up a very thin uid lm is formed along the wavemaker face. Hence,
3
In the shallow water limit the long (high-speed) wave components travel at a phase velocity c
f
=
(gh)
1/2
3.1 m/s.
5.3. NUMERICAL RESULTS 77
t must be suciently small in order to prevent free surface collocation points from
penetrating the advancing wavemaker. In the actual computations a time step t =
0.0025 is used. Finally, we remark that the position of the intersection point is based on
cubic spline extrapolations from the outer collocation points on the adjacent boundaries.
In the following paragraphs we discuss the computed results for
the initial potential and pressure distributions on the wavemaker, and the vertical
velocity,
the initial potential distribution on the bottom, and the horizontal velocity,
the initial vertical velocity of the free surface, and
the free surface elevation for small time.
5.3.1 Initial behaviour on the wavemaker
A plot of the velocity potential distribution on the wavemaker at t = 0 is presented
in Figure 5.3. It is observed that even with the smallest number of elements (eight)
on the wavemaker, the computed potential agrees remarkably well with the analytical
result (5.38).
4
The analytical solution (5.33) suggests that the vertical uid velocity near the inter-
section point tends to innity at t = 0, see also (5.35). This is conrmed by the computed
results for /z at t = 0, shown in Figure 5.4. With each doubling of the number of
elements, the numerical approximation is improved signicantly, especially in the vicinity
of the intersection point. The small deviations in /z that can be observed close to
the intersection point, are apparently due to the nite dierence approximation, since
the computed potential is in exact agreement with the analytical result over the whole
wavemaker range, see Figure 5.3.
Since the local behaviour of the potential distribution near the intersection point
can not be seen clearly from Figure 5.3, the computations have been repeated for larger
numbers of elements, corresponding to N
W
= 48 and N
W
= 64. Figure 5.5 gives an
enlarged view of the upper wavemaker region and the local distribution of the potential.
For the case N
W
= 64, the distance of the upper collocation point to the free surface
is 7 10
4
; it is remarkable that the numerical solution still agrees very well with the
analytical prediction (5.38). From Figure 5.6 it can be observed that the computed
vertical velocity /z is also improved substantially when more elements are used.
The initial pressure distribution on the wavemaker is shown in Figure 5.7, where the
dashed line represents the hydrostatic pressure. It can be seen that the initial dynamic
pressure is constant over the entire wavemaker, apart from the region near the intersec-
tion point. Globally, the computed pressure is in good agreement with the analytical
result (5.47).
5.3.2 Initial behaviour on the bottom
A plot of the initial potential distribution on the bottom is presented in Figure 5.8. Note
that the results are shown for 0 x 5 only, i.e. for the left half of the wave tank.
4
For completeness it is remarked that analytical solutions are represented by solid lines in Figures 5.3-
5.13.
78 CHAPTER 5. IMPULSIVE WAVEMAKER MOTION
The analytical solution (5.44) indicates a regular behaviour which is conrmed by the
computations. Excellent agreement is also observed for the horizontal velocity along the
bottom, shown in Figure 5.9, where the analytical predictions from (5.43) are used for
reference.
5.3.3 Initial behaviour on the free surface
The numerical results for the initial vertical velocity of the free surface are shown in
Figure 5.10. Note that the results have been plotted for 0 x 2.5 only, that is for the
region close to the wavemaker. Even for the smallest number of elements (twenty-ve)
on the free surface, the computed vertical velocity is in excellent agreement with the
analytical prediction (5.45). This is also the case for the ner grid distributions with
fty and one-hundred free-surface elements.
In order to make sure that our method is stable and suciently accurate, the number
of elements is increased; Figure 5.11 shows the computed vertical velocity in the
region 0 x 0.25 for up to two-hundred free-surface elements. Since the analytical
leading order solutions for the vertical velocity and the free surface elevation are equal
see (5.14) we have a rm condence in the ability of our method to predict the free
surface elevation, at least for small time; these results are discussed next.
5.3.4 Free surface elevation for small time
Next, the transient shape of the free surface is calculated up to t = 0.2 s. Figures 5.12
and 5.13 show the computed results in the region 0 x 2 compared to the ana-
lytical predictions from (5.26). The highest symbol in each curve represents the position
of the upper collocation point. For small time, say for t < 0.1, excellent agreement is
observed; for larger time, the computed free surface shape deviates somewhat from the
analytical result (which, indeed, is valid for small time only).
Taking into account the uniform horizontal motion of the wavemaker, it is clear
from Figure 5.13 that a very thin layer (or lm) of uid is formed along the advancing
wavemaker. This results in extremely small distances between free surface nodes and
wavemaker nodes, which destabilizes our algorithm; in the end, the computations break
down.
5.3. NUMERICAL RESULTS 79
THIS PAGE INTENTIONALLY LEFT BLANK
80 CHAPTER 5. IMPULSIVE WAVEMAKER MOTION
5.4 Concluding remarks
In this chapter we have made an attempt towards the numerical solution of the impulsive
wavemaker problem. From both physical and mathematical considerations it can be
expected that the potential solution becomes singular at the intersection of the advancing
wavemaker and the free surface. This singularity is conrmed in the small time analysis
given in section 5.2, which shows that the free surface elevation behaves like t log x
close to the wavemaker (which is located at x = Ut). The computed results support
the analytical expressions for the initial distributions of the potential and the tangential
velocities along the wavemaker and the bottom. Excellent agreement is also observed for
the analytical and numerical results for the initial vertical free surface velocity, and for
the free surface elevation for small time.
Contrary to Lins approach, no special treatment of the intersection point is needed
in our method to control the local behaviour of the initial potential distribution on
the wavemaker and the free surface elevation for small time. Therefore, it seems that
Lins suggestion to maintain a collocation point at the critical intersection point is not
applicable to all integral equation methods.
Although the method of small time expansions yields singular expressions for the free
surface elevation and the vertical uid velocity at the intersection point, the leading order
solutions appear to be very suitable for a direct verication of the numerically computed
values. Even extremely close to the wavemaker, the predicted logarithmic singularity in
the elevation is conrmed by the computations; capillary eects may be found by adding
small-scale physical factors such as surface tension to our numerical model. Such
eects may be important locally, but will not aect the global (integral) behaviour of the
solution.
5.5 Bibliography
Chwang, A.T. 1983. Nonlinear hydrodynamic pressure on an accelerating plate. The
Physics of Fluids 26(2):383-387.
Chwang, A.T., and Wang, K.-H. 1984. Nonlinear impulsive force on an accelerating
container. Journal of Fluids Engineering 106:233-240.
van Daalen, E.F.G., and Huijsmans, R.H.M. 1991. On the impulsive motion of a
wavemaker. Sixth International Workshop on Water Waves and Floating Bodies, Woods
Hole, Massachusetts.
Gradshteyn, I.S., and Ryzhik, I.M. 1980. Table of Integrals, Series, and Products.
Academic Press.
Greenhow, M. 1988. Water-entry and -exit of a horizontal circular cylinder. Applied
Ocean Research 10(4):191-198.
Greenhow, M., and Lin, W.-M. 1985. Numerical simulation of nonlinear free sur-
face ows generated by wedge entry and wavemaker motion. Proceedings of the Fourth
International Conference on Numerical Ship Hydrodynamics, Washington, DC.
Grosenbaugh, M.A., and Yeung, R.W. 1988. Nonlinear bow ows An experimen-
tal and theoretical investigation. Proceedings of the Seventeenth Symposium on Naval
Hydrodynamics, The Hague, The Netherlands.
5.5. BIBLIOGRAPHY 81
Hocking, L.M., and Mahdmina, D. 1991. Capillary-gravity waves produced by a
wavemaker. Journal of Fluid Mechanics 224:217-226.
Joo, S.W., Schultz, W.W., and Messiter, A.F. 1990. An analysis of the initial-
value wavemaker problem. Journal of Fluid Mechanics 214:161-183.
Korobkin, A.A. 1988. Inclined entry of a blunt prole into an ideal uid. Fluid Dy-
namics 23:443-447.
Korobkin, A.A., and Pukhnachov, V.V. 1988. Initial stage of water impact. An-
nual Review of Fluid Mechanics 20:159-185.
Lin, W.-M. 1984. Nonlinear Motion of the Free Surface near a Moving Body. Ph.D.
thesis, Massachusetts Institute of Technology Department of Ocean Engineering.
Lin, W.-M., Newman, J.N., and Yue, D.K.P. 1984. Nonlinear forced motions
of oating bodies. Proceedings of the Fifteenth Symposium on Naval Hydrodynamics,
Hamburg, Germany.
Miles, J. 1991. On the initial-value problem for a wavemaker. Journal of Fluid Me-
chanics 229:589-601.
Miloh, T. 1981. Wave slam on a sphere penetrating a free surface. Journal of Engi-
neering Mathematics 15(3):221-240.
Miloh, T. 1991a. On the initial-stage slamming of a rigid sphere in a vertical water
entry. Applied Ocean Research 13(1):43-48.
Miloh, T. 1991b. On the oblique water-entry of a rigid sphere. Journal of Engineering
Mathematics 25(1):77-92.
Peregrine, D.H. 1972. Flow due to vertical plate moving in a channel. Unpublished
note.
Roberts, A.J. 1987. Transient free-surface ows generated by a moving vertical plate.
Quarterly Journal of Mechanics and Applied Mathematics 40(1):129-158.
Vinje, T., and Brevig, P. 1981. Nonlinear ship motion. Proceedings of the Third
International Conference on Numerical Ship Hydrodynamics, Paris, France.
Wang, K.-H., and Chwang, A.T. 1989. Nonlinear free surface ow around an
impulsively moving cylinder. Journal of Ship Research 33(3):194-202.
82 CHAPTER 5. IMPULSIVE WAVEMAKER MOTION
Now when they stood apart and were ready with their gauntlets, straightway
in front of their faces they raised their heavy hands and matched their might
in deadly strife. Hereupon the Bebrycian king
5
even as a erce wave of
the sea rises in a crest against a swift ship, but she by the skill of the crafty
pilot just escapes the shock when the billow is eager to break over the bulwark
so he followed up the son of Tyndareus
6
, trying to daunt him, and gave
him no respite. But the hero, ever unwounded, by his skill baed the rush
of his foe, and he quickly noted the brutal play of his sts to see where he
was invincible in strength, and where inferior, and stood unceasingly and
returned blow for blow. And as when shipwrights with their hammers smite
ships timbers to meet the sharp clamps, xing layer upon layer; and the blows
resound one after another; so cheeks and jaws crashed on both sides, and a
huge clattering of teeth arose, nor did they cease ever from striking their
blows until laboured gasping overcame both. And standing a little apart they
wiped from their foreheads sweat in abundance, wearily panting for breath.
Then back they rushed together again, as two bulls ght in furious rivalry for
a grazing heifer. Next Amycus rising on tiptoe, like one who slays an ox,
sprung to his full height and swung his heavy hand down upon his rival; but
the hero swerved aside from the rush, turning his head, and just received the
arm on his shoulder; and coming near and slipping his knee past the kings,
with a rush he struck him above the ear, and broke the bones inside, and the
king in agony fell upon his knees; and the Minyan heroes
7
shouted for joy;
and his life was poured forth all at once . . .
Argonautica, Book II, Verses 67-97.
5
i.e. Amycus
6
i.e. Polydeuces
7
i.e. the Argonauts
5.5. BIBLIOGRAPHY 83
Figure 5.3: Initial potential distribution on the wavemaker.
84 CHAPTER 5. IMPULSIVE WAVEMAKER MOTION
Figure 5.4: Initial vertical velocity along the wavemaker.
5.5. BIBLIOGRAPHY 85
Figure 5.5: Initial potential distribution close to the intersection point.
86 CHAPTER 5. IMPULSIVE WAVEMAKER MOTION
Figure 5.6: Initial vertical velocity close to the intersection point.
5.5. BIBLIOGRAPHY 87
Figure 5.7: Initial pressure distribution on the wavemaker.
88 CHAPTER 5. IMPULSIVE WAVEMAKER MOTION
Figure 5.8: Initial potential distribution on the bottom.
5.5. BIBLIOGRAPHY 89
Figure 5.9: Initial horizontal velocity along the bottom.
90 CHAPTER 5. IMPULSIVE WAVEMAKER MOTION
Figure 5.10: Initial vertical velocity of the free surface.
5.5. BIBLIOGRAPHY 91
Figure 5.11: Initial vertical free surface velocity close to the intersection point.
92 CHAPTER 5. IMPULSIVE WAVEMAKER MOTION
Figure 5.12: Free surface elevation at times t = 0.0025 (0.0025) 0.0200 s.
5.5. BIBLIOGRAPHY 93
Figure 5.13: Free surface elevation at times t = 0.025 (0.025) 0.200 s.
94 CHAPTER 5. IMPULSIVE WAVEMAKER MOTION
Chapter 6
Hydrodynamic Mass and
Damping
But when the sun rising from far lands lighted up the dewy hills and wak-
ened the shepherds, then they loosed their hawsers from the stem of the baytree
and put on board all the spoil they had need to take; and with a favour-
ing wind they steered through the eddying Bosporus. Hereupon a wave like a
steep mountain rose aloft in front as though rushing upon them, ever upheaved
above the clouds; nor would you say that they could escape grim death, for in
its fury it hangs over the middle of the ship, like a cloud, yet it sinks away
into calm if it meets with a skilful helmsman. So they by the steering-craft of
Tiphys escaped, unhurt but sore dismayed. And on the next day they fastened
the hawsers to the coast opposite the Bithynian land.
There Phineus, son of Agenor, had his home by the sea, Phineus who
above all men endured most bitter woes because of the gift of prophecy which
Letos son had granted him aforetime. And he reverenced not a whit even Zeus
himself, for he foretold unerringly to men his sacred will. Wherefore Zeus sent
upon him a lingering old age, and took from his eyes the pleasant light, and
suered him not to have joy of the dainties untold that the dwellers around
ever brought to his house, when they came to enquire the will of heaven. But
on a sudden, swooping through the clouds, the Harpies with their crooked beaks
incessantly snatched the food away from his mouth and hands. And at times
not a morsel of food was left, at others but a little, in order that he might live
and be tormented. And they poured forth over all a loathsome stench; and
no one dared not merely to carry food to his mouth but even to stand at a
distance; so foully reeked the remnants of the meal . . .
Argonautica, Book II, Verses 164-193.
95
96 CHAPTER 6. HYDRODYNAMIC MASS AND DAMPING
6.1 Introduction
A solution for the ship motion problem at sea requires the determination of the hydro-
dynamic equilibrium of forces and moments. It is generally accepted that for the uid
forces the inuence of viscosity and surface tension are of minor importance compared
to pressure and wave eects. As far as ship motions are concerned, this proposition has
not been disproved by model or full scale experiments apart from the manoeuvring
problem. It has further been supposed that the whole motion problem can be regarded
as linear; up to the present state of development this has been conrmed substantially,
in any case for small amplitude motion and apart from very special objects.
By these circumstances the determination of the hydrodynamic forces acting on the
ships hull can be put as a linear boundary value problem in potential theory. The
superposition principle holds and the actual phenomenon is split up into the sum of
harmonic oscillations of the ship in still water and waves coming in on the restrained ship.
The two elds can be investigated entirely separately. Considering only the rst eld,
the problem can be stated as the oscillation of a rigid body moving with a certain speed
in the surface of a heavy and ideal uid. The solution supplies the six transfer functions
of the ship, which are composed of both rigid body characteristics and hydrodynamic
quantities.
Ursell (1949ab) made the rst contribution to the solution of this problem. He considered
a circular cylinder which oscillates harmonically with small amplitude, while the mean
position of the cylinder axis coincides with the mean surface of the uid. A solution to
the corresponding linear potential problem was found by superimposing suitably chosen
functions, such that each separate function satises Laplaces equation and the linearized
free surface conditions, while a combination of these functions fullls the remaining
boundary conditions.
Grim (1953, 1955/1956) and Tasai (1959, 1961) extended this principle from the
circular to elliptic cylinders and so-called Lewis-forms, while Porter (1960) formulated
the solution for heaving of an arbitrarily shaped cylinder. The extension of Ursells
theory for circular cylinders to nite constant depth was given by Yu and Ursell (1961).
Now in principle the way was free to investigate the inuence of form, of frequency of
motion, and of the coupling eects between sway and roll in detail. But rst the validity
of the theoretical approach had to be established by experiment; naturally, this was
rst tried for the most simple case of heaving. Tasai (1960) measured the wave heights
produced by forced heaving cylinders, and Porter (1960) measured the total vertical force
on a heaving circular cylinder and the pressure in a number of points along the contour.
Paulling and Richardson (1962) carried out extensive experiments recording the vertical
force and the pressure both in magnitude and in phase for dierent sections; wave
heights were measured as well. The results of these experiments were such that the
theoretical predictions were conrmed substantially. Few years later, the attention was
shifted towards rolling and swaying by Vugts (1968), who systematically carried out force
measurements as to amplitude and phase, for heaving, swaying, and rolling cylinders of
various cross sections.
For completeness, we mention the extension of Ursells techniques to three dimensions
by Havelock (1955) and Barakat (1962) for a heaving sphere. Kim (1965) computed added
mass and damping coecients for spheroids in surge (or sway), heave, and pitch (or roll)
motion, for ellipsoids in surge, heave, and pitch motion, and for cylinders with elliptic
cross section in all three modes of motion. Hulme (1982) computed added mass and
6.2. MATHEMATICAL MODEL FOR TWO-DIMENSIONAL MOTION 97
damping coecients for heaving and surging hemispheres.
Here we report on the numerical determination of the hydrodynamic (added) mass and
damping coecients for circular cylinders in forced harmonic heaving and swaying mo-
tion. The computed results are compared to the measurements of Vugts (1968) and to
the analytical predictions of de Jong (1973), who in turn based his calculations on Ursells
(1949ab) linear theory. In section 6.2 we present a mathematical model for the motion
of innitely long cylinders in a free surface; for sinusoidal motion with small amplitude,
the hydrodynamic coecients can be expressed in terms of the frequency and amplitude
of the body motion, and the amplitude and phase of the forces. The numerical results
are discussed in section 6.3, and concluding remarks are given in section 6.4.
6.2 Mathematical model for two-dimensional motion
Let x, z be a coordinate system which is xed in space. The x-axis coincides with the
still water level and its positive part is directed towards the right. The z-axis is vertical
and positive upwards. The origin O is the intersection of the centreline of the cylinder
section which is of arbitrary shape and the waterline, see Figure 6.1. Suppose
now that the centre of mass G of the cylinder is situated in O. The cylinder position
and orientation are denoted by (x
G
, z
G
) and
G
respectively. The most general way to
describe the cylinder motion as a linear system is a set of three linear coupled equations
of motion. Following Vugts (1968), this set can be put as
F
x
= (M +a
xx
) x
G
+b
xx
x
G
+c
xx
x
G
+ a
xz
z
G
+b
xz
z
G
+c
xz
z
G
+a
x

G
+b
x

G
+c
x

G
, (6.1)
F
z
= (M +a
zz
) z
G
+b
zz
z
G
+c
zz
z
G
+ a
z

G
+b
z

G
+c
z

G
+a
zx
x
G
+b
zx
x
G
+c
zx
x
G
, (6.2)
L

= (I +a

G
+b

G
+c

G
+ a
x
x
G
+b
x
x
G
+c
x
x
G
+a
z
z
G
+b
z
z
G
+c
z
z
G
, (6.3)
98 CHAPTER 6. HYDRODYNAMIC MASS AND DAMPING
Figure 6.1: Two-dimensional cylinder oating on a free surface.
where a dot denotes dierentiation with respect to time, and
M is the mass of the cylinder section,
I is the moment of inertia about G,
a
ii
is the hydrodynamic mass (or moment of inertia) coecient in the i-mode of
motion,
a
ij
is the mass coupling coecient in the i-mode equation by j-mode motion,
b
ii
is the hydrodynamic damping coecient against i-mode motion,
b
ij
is the damping coupling coecient in the i-mode equation by j-mode motion,
c
ii
is the hydrostatic restoring coecient against an i-mode displacement,
c
ij
is the restoring coupling coecient in the i-mode equation against a j-mode
displacement,
F
x
is the horizontal external force,
F
z
is the vertical external force,
L

is the external moment about G.


The coecients c
ii
and c
ij
can be determined by pure hydrostatics. For example, if
a cylinder is given a downward displacement z, Archimedes principle
1
states that the
restoring force equals the weight of the displaced water volume. If the cross section width
at the waterline is denoted by B, we have c
zz
z = gBz, and hence c
zz
= gB.
By simple reasoning the above set of equations can be simplied considerably. The
horizontal force is not opposed by any restoring force, so c
xx
= c
zx
= c
x
= 0. The
vertical motion is symmetric with respect to the z-axis and can not produce any lateral
forces or moments; therefore a
xz
= b
xz
= c
xz
= 0, and a
z
= b
z
= c
z
= 0. A static
1
Archimedes (287-212 B.C.), mathematician of Syracuse. He discovered the ratio of the radius of a
circle to its circumference and the formulas for the surface and volume of a cylinder and of a sphere, and
demonstrated the hydrostatic law known as Archimedes principle.
6.2. MATHEMATICAL MODEL FOR TWO-DIMENSIONAL MOTION 99
roll displacement does not generate a horizontal force, which implies c
x
= 0. Thus, the
mathematical model is reduced to
F
x
= (M +a
xx
) x
G
+b
xx
x
G
+a
x

G
+b
x

G
, (6.4)
F
z
= (M +a
zz
) z
G
+b
zz
z
G
+c
zz
z
G
+ a
z

G
+b
z

G
+c
z

G
+a
zx
x
G
+b
zx
x
G
, (6.5)
L

= (I +a

G
+b

G
+c

G
+a
x
x
G
+b
x
x
G
. (6.6)
From (6.4) and (6.6) it is seen that heave does not inuence the coupled sway-roll motion.
However, the reverse need not be true, see (6.5). In a potential ow the sway and roll
problem is asymmetric with respect to the z-axis and the corresponding contributions
in (6.5) vanish. Thus, heaving becomes an uncoupled motion with one degree of freedom.
The cylinders are harmonically oscillated in one of the three modes of motion:
x
G
= a
x
sin t and z = = 0 , (6.7)
z
G
= a
z
sin t and = x = 0 , (6.8)

G
= a

sin t and x = z = 0 , (6.9)


while F
x
, F
z
, and L

are computed in amplitude and phase:


F
x
=

S
p n
x
ds = A
x
sin (t +
x
) , (6.10)
F
z
=

S
p n
z
ds = A
z
sin (t +
z
) , (6.11)
L

S
p (r
x
n
z
r
z
n
x
) ds = A

sin (t +

) . (6.12)
Substitution of (6.7) and (6.10) into (6.4) shows that the hydrodynamic mass and damp-
ing coecients for sway motion read
a
xx
=
A
x
cos
x

2
a
x
M , b
xx
=
A
x
sin
x
a
x
(6.13)
Similarly, the coecients for heave motion are found from substitution of (6.8) and (6.11)
into (6.5):
a
zz
=
a
z
c
zz
A
z
cos
z

2
a
z
M , b
zz
=
A
z
sin
z
a
z
(6.14)
Following the same procedure for roll, it is found that the hydrodynamic mass and
damping coecients are given by
a

=
A

cos

2
a

I , b

=
A

sin

(6.15)
These expressions will be used in the next section to compare the computed results with
the analytical predictions and experimental measurements.
100 CHAPTER 6. HYDRODYNAMIC MASS AND DAMPING
6.3 Numerical results
Here we discuss the numerical results for added mass and damping for two-dimensional
circular cylinders in forced harmonic heaving and swaying motion.
The sinusoidal motions x
G
(i.e. sway) and z
G
(i.e. heave) are given in (6.7) and (6.8)
respectively. The forces F
x
and F
z
are computed from pressure integrations over the wet-
ted part of the cylinder surface, see (6.10-6.11). The imposed motions and the computed
forces serve as input signals for a harmonic analysis.
2
The main output usually consists
of the amplitude and phase leads of the leading (and, if possible, higher) frequency com-
ponents, i.e. A
x
and
x
for sway, and A
z
and
z
for heave. These data can then be used
to compute the hydrodynamic mass and damping coecients from (6.13) for sway, and
from (6.14) for heave.
Finally, we note that the forces M x
G
and M z
G
account for the accelerations of the
cylinder. Since these contributions are not computed in the numerical tests discussed
hereafter, they should be removed from (6.4-6.5). Consequently, the mass M is set equal
to zero in (6.13-6.14).
6.3.1 Circular cylinder in heaving motion
First we discuss the computed results for circular cylinders undergoing a sinusoidal heav-
ing motion. The tests have been carried out in a numerical wave tank with length
L = 60 m and water depth h = 10 m; the cylinder beam is B = 4 m. The imposed cylin-
der motion and the ensuing wave motion have been simulated for twelve dierent fre-
quencies ranging from = 0.5 rad/s to = 6.0 rad/s, with increment = 0.5 rad/s
and the calculations have been sustained for about six cycles of motion with amplitude
a
z
= 0.1 m.
The results for added mass and damping are listed in Table 6.1. The frequency and
the hydrodynamic coecients are made nondimensional by
=

B
2g

1/2
, a
zz
=
a
zz

,

b
zz
=
b
zz

B
2g

1/2
(6.16)
where = 1000 kg/m
3
is the water density, and =
1
8
B
2
is the displaced water volume
when the cylinder is in its equilibrium position.
2
We used the harmonic analysis program HARMAN, which has been developed at MARIN.
6.3. NUMERICAL RESULTS 101
A
z

z
a
zz
a
zz
b
zz

b
zz

10
3

(deg)

10
3

10
3

0.5 0.226 3.759 6.17 7.471 1.189 8.080 0.581


1.0 0.452 3.537 13.26 4.813 0.766 8.113 0.584
1.5 0.677 3.243 21.29 4.010 0.638 7.850 0.565
2.0 0.903 2.697 29.53 3.943 0.628 6.646 0.478
2.5 1.129 1.749 40.34 4.145 0.660 4.529 0.326
3.0 1.355 0.908 104.79 4.618 0.735 2.926 0.211
3.5 1.580 2.336 163.43 5.031 0.801 1.903 0.137
4.0 1.806 4.628 173.98 5.329 0.848 1.213 0.087
4.5 2.032 7.390 177.77 5.584 0.889 0.639 0.046
5.0 2.258 10.62 178.97 5.817 0.926 0.382 0.027
5.5 2.483 14.15 179.31 5.975 0.951 0.313 0.022
6.0 2.709 17.97 179.39 6.081 0.968 0.319 0.023
Table 6.1: Added mass and damping coecients for heaving circular cylinder.
Figures 6.2 and 6.3 present the numerical results for added mass and damping re-
spectively, and the comparison with the measurements of Vugts and the analytical
predictions
3
by de Jong.
4
Over the whole frequency range excellent agreement is observed
for both added mass and damping, except for the lower frequencies; for = 0.226 and
= 0.452 the computed added mass is too low, and for the lowest frequency = 0.226
the damping is too high.
It is very likely that these eects are caused by reections from the lateral out-
ow boundaries situated at both ends of the wave tank. For low frequency motion,
the generated long wave components radiate at a high velocity towards these so-called
Sommerfeld/Orlanski boundaries, where they are partially reected. Once the reected
waves approach the heaving cylinder, the force registration is disturbed; this may spoil
the numerical results.
To conrm this supposition, the computations for = 0.452 have been repeated in a
numerical wave tank with a doubled length L = 120 m. The results are listed in Table
6.2; indeed, it can be observed that the computed added mass is somewhat improved
that is, a small increase can be observed. However, the damping has also increased;
evidently, there are other circumstances (i.e. physical parameters) which inuence the
numerical computations.
3
For completeness it is remarked that analytical solutions are represented by solid lines in Figures 6.2-
6.7.
4
Note that the results for 4.5 rad/s are not shown in these gures; the computations for these
high frequencies have been performed just to verify that in this case for the added mass
and damping tend to unity and zero respectively.
102 CHAPTER 6. HYDRODYNAMIC MASS AND DAMPING
L A
z

z
a
zz
a
zz
b
zz

b
zz

10
3

(deg)

10
3

10
3

60 3.537 13.26 4.813 0.766 8.113 0.584


120 3.528 14.48 5.081 0.809 8.821 0.635
Table 6.2: Added mass and damping coecients for heaving circular cylinder: eects
due to longer wave tank for = 0.452.
Another important parameter in this problem is the uniform water depth. Since the
above-mentioned improvement in the added mass is but marginal, it can be expected that
both the wave tank length and the water depth must be chosen considerably larger to
obtain fair agreement with the analytical predictions for both added mass and damping.
This option however, is rather inecient since the number of panels will increase, and
the computational costs will rise dramatically.
A more ecient way to improve the computed results is to choose a smaller cylinder
diameter; by decreasing the beam B and increasing the frequency such that the
nondimensional frequency remains constant in eect a larger wave tank is obtained.
5
The computed results for three combinations of , B, and a
z
are listed in Table 6.3.
/B/a
z
A
z

z
a
zz
a
zz
b
zz

b
zz

10
3

(deg)

10
3

10
3

1.0/4.0/0.10 0.452 3.537 13.26 4.813 0.766 8.113 0.584


2.0/1.0/0.05 0.452 0.424 15.41 0.409 1.041 1.127 0.648
2.0/0.25/0.025 0.226 0.056 5.81 0.051 2.095 0.057 0.526
Table 6.3: Added mass and damping coecients for heaving circular cylinder: eects
due to smaller cylinders.
Figures 6.4 and 6.5 show the overall results, including the improved results for the lower
frequency range. Excellent agreement with linear theory is observed, whereas Vugts
measurements are too low and scattered
6
for low-frequency added mass.
5
In this respect it should be mentioned that Vugts (1968) used cylinders with beam B = 0.3 m in an
experimental wave tank with length L = 142 m and mean water depth h = 2 m.
6
Probably, this scattering of experimental results for the lower frequency range is also due to the
limited accuracy of the measurement equipment used by Vugts.
6.3. NUMERICAL RESULTS 103
6.3.2 Circular cylinder in swaying motion
Finally, we discuss the computed results for circular cylinders in forced harmonic swaying
motion. The tank dimensions are given by L = 60 m and h = 10 m; the cylinder beam is
B = 4 m. The imposed sinusoidal cylinder motion with amplitude a
x
= 0.1 m and
the ensuing wave motion are computed for four dierent frequencies, and up to six motion
cycles. The frequency and the hydrodynamic coecients are made nondimensional by
=

B
2g

1/2
, a
xx
=
a
xx

,

b
xx
=
b
xx

B
2g

1/2
(6.17)
The results for added mass and damping are listed in Table 6.4, where both dimensional
and nondimensional values are given.
A
x

x
a
xx
a
xx
b
xx

b
xx

10
3

(deg)

10
3

10
3

1.0 0.452 0.800 167.7 7.816 1.244 1.704 0.123


1.5 0.677 2.072 143.3 7.383 1.175 8.255 0.594
2.0 0.903 2.572 121.8 3.388 0.539 10.93 0.786
3.0 1.355 2.883 112.0 1.200 0.191 8.910 0.641
Table 6.4: Added mass and damping coecients for swaying circular cylinder.
Figures 6.6 and 6.7 compare our results with the experimental results from Vugts and
the analytical linear predictions. Over the whole frequency range fair agreement can
be observed. Unfortunately, the limited amount of time available did not allow us to
compute added mass and damping in sway for the higher and lower frequency range, nor
did we have time to investigate the eects of the tank dimensions for the case of swaying
motion.
104 CHAPTER 6. HYDRODYNAMIC MASS AND DAMPING
6.4 Concluding remarks
In this chapter we have focused our attention on the hydrodynamic mass and damping
coecients for circular cylinders in harmonic heaving and swaying motion. In section 6.2
the mathematical model for linearized two-dimensional motion has been presented, in-
cluding expressions for the added mass and damping coecients in terms of the body
amplitude and frequency, and the force amplitude and phase lead.
At rst sight, the numerically computed results for heave seem to be in excellent
agreement with both analytical and experimental results. However, in the lower frequency
range one should be careful; the reected parts of the long (radiated) wave components
disturb the registered force signal, and hence the computed added mass and damping.
This problem can be by-passed by choosing a larger (numerical) wave tank that is,
both the tank length and the water depth have to be increased or by decreasing the
cylinder diameter. The latter option is more ecient, considering the number of panels
and the computational eort involved. Indeed, this technique leads to improved results
for both added mass and damping in heave.
The computed results for added mass and damping for a circular cylinder in swaying
motion are in fair agreement with the linear theory and Vugts experiments.
Of course we are well aware of the limited scope of this study. First of all, the approach
suggested above should also be applied to the higher frequency range in order to validate
the method. Secondly, for all modes of motion the wave-amplitude ratios should also be
computed. Further, we have to admit that the number of frequencies used in swaying
motion is rather small; unfortunately, the completion of this thesis did not permit us to
spend more time on additional computations.
On the other hand, the results obtained so far give us some condence in the ability
of the method to predict added mass and damping coecients for other cross sections, in
all three modes of motion. Moreover, our nonlinear method oers the unique opportunity
to investigate the inuence of amplitude which is likely to introduce nonlinear eects
on linear concepts as added mass and damping coecients.
6.5 Bibliography
Barakat, R. 1962. Vertical motion of a oating sphere in a sine-wave sea. Journal of
Fluid Mechanics 13:540-556.
Grim, O. 1953. Berechnung der durch Schwingungen eines Schikorpers erzeugten
hydrodynamischen Krafte. Jahrbuch der Schibautechnischen Gesellschaft.
Grim, O. 1955/1956. Die hydrodynamischen Krafte beim Rollversuch. Schistechnik,
Band 3.
Havelock, T. 1955. Waves due to a oating sphere making periodic heaving oscilla-
tions. Proceedings of the Royal Society of London, Series A, 231:1-7.
Hulme, A. 1982. The wave forces acting on a oating hemisphere undergoing forced
periodic oscillations. Journal of Fluid Mechanics 121:443-463.
de Jong, B. 1973. Computation of the hydrodynamic coecients of oscillating cylin-
ders. Report 145S, Netherlands Ship Research Centre TNO, Shipbuilding Department,
Delft, The Netherlands.
6.5. BIBLIOGRAPHY 105
Kim, W.D. 1965. On the harmonic oscillations of a rigid body on a free surface. Journal
of Fluid Mechanics 21(3):427-451.
Paulling, J.R., and Richardson, R.K. 1962. Measurement of pressures, forces
and radiating waves for cylinders oscillating in a free surface. University of California,
Institute of Engineering Research, Berkeley.
Porter, W.R. 1960. Pressure distribution, added mass and damping coecients for
cylinders oscillating in a free surface. University of California, Institute of Engineering
Research, Berkeley.
Tasai, F. 1959. On the damping force and added mass of ships heaving and pitching.
Reports of the Research Institute for Applied Mechanics, Kuyushu University, Volume
IX, Number 26.
Tasai, F. 1960. Measurement of the wave height produced by the forced heaving of the
cylinders. Reports of the Research Institute for Applied Mechanics, Kuyushu University,
Volume VIII, Number 29.
Tasai, F. 1961. Hydrodynamic force and moment produced by swaying and rolling os-
cillation of cylinders on the free surface. Reports of the Research Institute for Applied
Mechanics, Kuyushu University, Volume IX, Number 35.
Ursell, F. 1949a. On the heaving motion of a circular cylinder on the surface of a
uid. Quarterly Journal of Mechanics and Applied Mathematics 2(2):218-231.
Ursell, F. 1949b. On the rolling motion of cylinders in the free surface of a uid.
Quarterly Journal of Mechanics and Applied Mathematics 2(3):335-353.
Vugts, J.H. 1968. The hydrodynamic coecients for swaying, heaving and rolling cylin-
ders in a free surface. International Shipbuilding Progress 15(167):251-275.
Yu, S.C., and Ursell, F. 1961. Surface waves generated by an oscillating circular
cylinder on water of nite depth: theory and experiment. Journal of Fluid Mechanics
11:529-551.
106 CHAPTER 6. HYDRODYNAMIC MASS AND DAMPING
Meanwhile the chiefs carefully cleansed the old mans squalid skin and
with due selection sacriced sheep which they had borne away from the spoil
of Amycus. And when they had laid a huge supper in the hall, they sat down
and feasted, and with them feasted Phineus ravenously, delighting his soul,
as in a dream. And there, when they had taken their ll of food and drink,
they kept awake all night waiting for the sons of Boreas.
7
And the aged sire
himself sat in the midst, near the hearth, telling of the end of their voyage
and the completion of their journey:
Listen then. Not everything is it lawful for you to know clearly; but
whatever is heavens will, I will not hide. I was infatuated aforetime, when in
my folly I declared the will of Zeus in order and to the end. For he himself
wishes to deliver to men the utterances of the prophetic art incomplete, in
order that they may still have some need to know the will of heaven.
First of all, after leaving me, ye will see the twin Cyanean rocks where the
two seas
8
meet. No one, I ween, has won his escape between them. For they
are not rmly xed with roots beneath, but constantly clash against one an-
other to one point, and above a huge mass of salt water rises in a crest, boiling
up, and loudly dashes upon the hard beach. Wherefore now obey my counsel,
if indeed with prudent mind and reverencing the blessed gods ye pursue your
way; and perish not foolishly by a self-sought death, or rush on following the
guidance of youth. First entrust the attempt to a dove when ye have sent her
forth from the ship. And if she escapes safe with her wings between the rocks
to the open sea, then no more do ye refrain from the path, but grip your oars
well in your hands and cleave the seas narrow strait, for the light of safety
will be not so much in prayer as in strength of hands. Wherefore let all else
go and labour boldly with might and main, but ere then implore the gods as
ye will, I forbid you not. But if she ies onward and perishes midway, then
do ye turn back; for it is better to yield to the immortals. For ye could not
escape an evil doom from the rocks, not even if Argo were of iron.
Argonautica, Book II, Verses 308-340.
7
i.e. Zetes and Calais, who will chase away the Harpies
8
i.e. the Sea of Marmara and the Black Sea
6.5. BIBLIOGRAPHY 107
Figure 6.2: Added mass coecients for circular cylinder in heaving motion.
108 CHAPTER 6. HYDRODYNAMIC MASS AND DAMPING
Figure 6.3: Damping coecients for circular cylinder in heaving motion.
6.5. BIBLIOGRAPHY 109
Figure 6.4: Improved added mass coecients for circular cylinder in heaving motion.
110 CHAPTER 6. HYDRODYNAMIC MASS AND DAMPING
Figure 6.5: Improved damping coecients for circular cylinder in heaving motion.
6.5. BIBLIOGRAPHY 111
Figure 6.6: Added mass coecients for circular cylinder in swaying motion.
112 CHAPTER 6. HYDRODYNAMIC MASS AND DAMPING
Figure 6.7: Damping coecients for circular cylinder in swaying motion.
Chapter 7
Cylinders in Free Motion
Now when they reached the narrow strait of the winding passage, hemmed
in on both sides by rugged clis, while an eddying current from below was
washing against the ship as she moved on, they went forward sorely in dread;
and now the thud of the crashing rocks ceaselessly struck their ears, and the
sea-washed shores resounded, and then Euphemus grasped the dove in his
hand and started to mount the prow; and they, at the bidding of Tiphys, son
of Hagnias, rowed with good will to drive Argo between the rocks, trusting to
their strength. And as they rounded a bend they saw the rocks opening for the
last time of all. Their spirit melted within them; and Euphemus sent forth the
dove to dart forward in ight; and they all together raised their heads to look;
but she ew between them, and the rocks again rushed together and crashed
as they met face to face. And the foam leapt up in a mass like a cloud; awful
was the thunder of the sea; and all round them the mighty welkin roared.
The hollow caves beneath the rugged clis rumbled as the sea came surging
in; and the white foam of the dashing wave spurted high above the cli. Next
the current whirled the ship round. And the rocks shore away the end of the
doves tail-feathers; but away she ew unscathed. And the rowers gave a loud
cry; and Tiphys himself called to them to row with might and main. For the
rocks were again parting asunder. But as they rowed they trembled, until the
tide returning drove them back within the rocks. Then most awful fear seized
upon all; for over their head was destruction without escape . . .
Argonautica, Book II, Verses 549-578.
113
114 CHAPTER 7. CYLINDERS IN FREE MOTION
7.1 Introduction
In this chapter we report on the numerical calculations for two-dimensional cylinders
in free heaving and rolling motion. The computations have been performed with our
computer code TIPHYS, which is based on the panel method for nonlinear ship motions
in water waves. A complete description of the numerical algorithm is given in chapter 4,
where we suggested a special approach to compute the partial time derivative of the
velocity potential along the wetted part of the body. It is expected that the solution
of an extra boundary integral equation for
t
, involving the hydrodynamic equations
of motion for the body, provides a good approximation of the Bernoulli pressure along
the wetted body surface. In this way the stability of the step-by-step method for freely
oating bodies is ensured.
In all test cases discussed here, the cylinder motion starts either from an initial
displacement with zero initial velocity or with an initial velocity from the equilibrium
position. The ensuing motion consists of the cylinder motion and the motion of the
surrounding uid. If possible, we shall compare our results with analytical predictions.
7.2 Circular heaving cylinder
First, consider a circular cylinder oating on the free surface of a wave tank. The cylinder
radius is denoted by R, and its uniform density equals half the water density; in a state
of equilibrium the centre of mass G lies on the mean water level z = 0. The cylinder is
disturbed from its equilibrium position either by a vertical upward displacement z
G
(0) <
R or by an upward velocity z
G
(0) > 0. The ensuing free motion of this wave-body system
consists of the cylinder motion, the motion of the surrounding uid which contributes to
the eective cylinder mass i.e. the cylinder mass plus the virtual added mass and
a wave motion which extracts energy from the cylinder i.e. wave radiation damping.
The unsteady heaving motion of this particular cylinder type was rst addressed
analytically by Ursell (1949, 1964) who considered the problem by superposing har-
monic wave components. From the analytical behaviour of the force coecients in the
frequency domain, Ursell obtained large time asymptotics for the response of a half-
immersed cylinder. Few years later, the entire response range for initial displacements
and initial velocities was computed by Maskell and Ursell (1970). On the basis of their
linear theory, they found that the normalized and time-scaled displacement

G
()
z
G

g/R

z
G
(0)
(7.1)
7.2. CIRCULAR HEAVING CYLINDER 115
behaves for small time like

G
() = 1
1

2
+
2
9
2

4
+O

. (7.2)
For large , the following approximation was derived:

G
() = 0.9664e
0.1309
cos (0.9117 0.4805) . (7.3)
For the intermediate time range say, for [1.5, 11.0] Maskell and Ursell computed

G
() numerically. Though approximative, the data measured from their publications
are used here for reference.
To start with, our method is tested for a circular cylinder with radius R = 2 m in
three initial upward displacements: z
G
(0) =
1
8
R = 0.25 m, z
G
(0) =
1
4
R = 0.50 m,
and z
G
(0) =
1
2
R = 1.00 m; the initial cylinder positions relative to the still water level
have been sketched above. The tank length is L = 40 m, and the mean water depth is
h = 10 m. A Sommerfeld/Orlanski radiation condition (
t
= c
x
) is used at both ends of
the wave tank, with the shallow water approximation c = (gh)
1/2
9.90 m/s substituted
for the wave celerity. The number of panels is 72 at the free surface (left plus right),
30 at the wetted part of the cylinder contour, 20 at the lateral outow boundaries (left
plus right), and 20 at the bottom. The time step is t = 0.05 s, so that 200 steps are
needed for the rst 10 simulation seconds. Comparative tests with a doubled tank length
L = 80 m indicate that within this time interval the reected parts of the outgoing waves
do not aect the solution in the neighbourhood of the cylinder.
116 CHAPTER 7. CYLINDERS IN FREE MOTION
The computed transient heaving motion of the cylinder is presented in Figure 7.1, where
both the vertical displacements and the vertical velocities have been scaled with respect
to the respective initial displacements. From these plots it can be observed that the
motion is strongly damped, which indicates a high rate of energy (or momentum) transfer
from the cylinder to the uid. Of course this eect can be attributed to the special
circular shape, which implies a strong variance of the cross section width at the waterline.
Since the heave response lines nearly coincide within drawing accuracy, we arrive at the
conclusion that clear nonlinear eects are absent in these computations.
Additional tests are carried out for z
G
(0) =
5
8
R = 1.25 m, and z
G
(0) =
3
4
R = 1.50 m, see
also the above sketch. The results from these computations are compared with the results
for z
G
(0) =
1
2
R = 1.00 m in Figure 7.2. It is remarkable that clear nonlinear eects still
can not be observed, in spite of the fact that in the last case, where z
G
(0) = 1.50 m,
the cylinder motion nearly starts from a state of total emergence. However, it should
be remarked that for the whole range of initial displacements the computed results are
in excellent agreement with the analytical predictions from the aforementioned theory
of Ursell and Maskell; Figure 7.3 compares our results for z
G
(0) = 1.50 m with their
normalized and time-scaled results.
The transient heave responses for a unit initial displacement and a unit initial velocity
are presented in Figure 7.4. For cylinders of arbitrary shape, Yeung (1982) has shown
that the initial-velocity response equals the time derivative of the initial-displacement
response multiplied by one half of the innite-uid virtual mass of the cylinder. Apart
from nite-uid eects in our computations and scaling eects, this result is conrmed
by the plots in Figure 7.4; compare the solid line starting from zero with the dotted line
starting from unity.
7.3. RECTANGULAR HEAVING CYLINDER 117
7.3 Rectangular heaving cylinder
Next, consider a homogeneous rectangular cylinder oating on the free surface; the equi-
librium state corresponds to the situation where the centre of mass is on the mean water
level. In the cases discussed here, the cylinder is disturbed from its equilibrium by down-
ward displacements or velocities. The cylinder width is B = 4 m, the height is H = 2 m,
and the draft is T = 1 m. Since sharp edges may introduce singularities in the ow of
the surrounding uid, the cylinder corners are rounded with radius R = 0.25 m. In this
way the local uid motion is expected to meet the requirements for potential ow as well
as possible. Neither the tank dimensions nor the panel distributions have been altered.
Three initial downward displacements were chosen: z
G
(0) =
1
4
T = 0.25 m, z
G
(0) =

1
2
T = 0.50 m, and z
G
(0) =
3
4
T = 0.75 m. The above sketch shows the initial
positions of the cylinder relative to the still water level.
The computed results for the normalized heave displacements and velocities are shown
in Figure 7.5; as in the previous test series, the motion is strongly damped and appar-
ently without nonlinearities. In Figure 7.6 the heave responses for an initial displace-
ment z
G
(0) = 0.50 m and an initial velocity z
G
(0) = 0.50 m/s are presented. Again,
apart from nite-uid and scaling eects, the computations seem to support Yeungs the-
ory regarding the simple relationship between the initial-displacement response and the
initial-velocity response. Finally, it is remarked that the results presented here are in fair
agreement with the computations by Chapman (1979); a discussion of this comparison
will be given by van Daalen and Zandbergen (1993).
118 CHAPTER 7. CYLINDERS IN FREE MOTION
7.4 Rectangular rolling cylinder
The rectangular cylinder introduced in the previous section is now given an initial roll
displacement; the sketch below gives an impression of the cylinder orientation for
G
(0) =
0.05 rad 2.9 deg,
G
(0) = 0.15 rad 8.6 deg,
G
(0) = 0.25 rad 14.3 deg, and
G
(0) =
0.35 rad 20.1 deg. Neither the tank dimensions nor the panel distributions have been
changed.
The transient normalized roll responses have been gathered in Figure 7.7. Clearly, the
motion is but slightly damped for all four initial roll displacements. Small nonlinear
eects can be observed, since the roll period is but slightly decreased when the initial
angle of roll is increased.
7.5 Concluding remarks
The above results for two-dimensional freely oating cylinders seem to indicate that clear
nonlinear eects are absent in these simulations. This is not a surprising result for the
case of heaving circular and rectangular cylinders with small initial displacements, but
one might expect distinct nonlinearities in the case where the cylinders are near total
emergence or submergence. Also, some nonlinear eects might have been anticipated for
the case of a rectangular rolling cylinder.
So far, we did not succeed in nding a conclusive answer to the question why nonlinear
phenomena remain absent in our computations. Perhaps other simulations involving
special geometries for instance, where in the equilibrium position the larger part of
the body is submerged will reveal nonlinearities in the coupled wave-body motion.
For the time being, we have to close this discussion with the statement that these (and
other) problems will be the subject of future research.
7.6. BIBLIOGRAPHY 119
7.6 Bibliography
Chapman, R.B. 1979. Large-amplitude transient motion of two-dimensional oating
bodies. Journal of Ship Research 23(1):20-31.
van Daalen, E.F.G., and Zandbergen, P.J. 1993. A novel extension of a panel
method for nonlinear gravity waves to interactions with freely oating bodies. To appear
in the Journal of Engineering Mathematics.
Maskell, S.J., and Ursell, F. 1970. The transient motion of a oating body.
Journal of Fluid Mechanics 44(2):303-313.
Ursell, F. 1949. On the heaving motion of a circular cylinder on the surface of a uid.
Quarterly Journal of Mechanics and Applied Mathematics 2(2):218-231.
Ursell, F. 1964. The decay of the free motion of a oating body. Journal of Fluid
Mechanics 19(2):305-319.
Yeung, R.W. 1982. The transient heaving motion of oating cylinders. Journal of
Engineering Mathematics 16(2):97-119.
120 CHAPTER 7. CYLINDERS IN FREE MOTION
. . . And now to right and left broad Pontus was seen, when suddenly a huge
wave rose up before them, arched, like a steep rock; and at the sight they
bowed with bended heads. For it seemed about to leap down upon the ships
whole length and to overwhelm them. But Tiphys was quick to ease the ship
as she laboured with the oars; and in all its mass the wave rolled away beneath
the keel, and at the stern it raised Argo herself and drew her far away from
the rocks; and high in air was she borne. But Euphemus strode among all
his comrades and cried to them to bend to their oars with all their might;
and they with a shout smote the water. And as far as the ship yielded to the
rowers, twice as far did she leap back, and the oars were bent like curved bows
as the heroes used their strength.
Then a vaulted billow rushed upon them, and the ship like a cylinder ran
on the furious wave plunging through the hollow sea. And the eddying current
held her between the clashing rocks; and on each side they shook and thun-
dered; and the ships timbers were held fast. Then Athena with her left hand
thrust back one mighty rock and with her right pushed the ship through; and
she, like a winged arrow, sped through the air. Nevertheless the rocks, cease-
lessly clashing, shore o as she passed the extreme end of the stern-ornament.
But Athena soared up to Olympus, when they had escaped unscathed. And
the rocks in one spot at that moment were rooted fast for ever to each other,
which thing had been destined by the blessed gods, when a man in his ship
should have passed between them alive. And the heroes breathed again after
their chilling fear, beholding at the same time the sky and the expanse of sea
spreading far and wide. For they deemed that they were saved from Hades
1
;
and Tiphys rst of all began to speak:
It is my hope that we have safely escaped this peril we, and the ship;
and none other is the cause so much as Athena, who breathed into Argo
divine strength when Argus knitted her together with bolts; and she may not
be caught. Son of Aeson, no longer fear thou so much the hest of thy king,
since a god hath granted us escape between the rocks; for Phineus, Agenors
son, said that our toils hereafter would be lightly accomplished.
Argonautica, Book II, Verses 579-618.
1
i.e. the underworld
7.6. BIBLIOGRAPHY 121
Figure 7.1: Transient heaving motion of a circular cylinder due to moderate initial dis-
placements.
122 CHAPTER 7. CYLINDERS IN FREE MOTION
Figure 7.2: Transient heaving motion of a circular cylinder due to large initial displace-
ments.
7.6. BIBLIOGRAPHY 123
Figure 7.3: Transient heaving motion of a circular cylinder due to initial displacement.
Numerical versus analytical results.
124 CHAPTER 7. CYLINDERS IN FREE MOTION
Figure 7.4: Transient heaving motion of a circular cylinder due to initial unit displace-
ment and initial unit velocity.
7.6. BIBLIOGRAPHY 125
Figure 7.5: Transient heaving motion of a rectangular cylinder due to initial displace-
ments.
126 CHAPTER 7. CYLINDERS IN FREE MOTION
Figure 7.6: Transient heaving motion of a rectangular cylinder due to initial displacement
and initial velocity.
7.6. BIBLIOGRAPHY 127
Figure 7.7: Transient rolling motion of a rectangular cylinder due to initial displacements.
128 CHAPTER 7. CYLINDERS IN FREE MOTION
Part III
Variational Principles and
Hamiltonian Formulations
129
Chapter 8
Lagrangian and Hamiltonian
Formulations
And they two
1
by the pathway came to the sacred grove, seeking the huge
oak tree on which was hung the eece, like to a cloud that blushes red with
the ery beams of the rising sun. But right in front the serpent with his keen
sleepless eyes saw them coming, and stretched out his long neck and hissed in
awful wise . . . And as he writhed, the maiden came before his eyes, with sweet
voice calling to her aid Sleep, highest of gods, to charm the monster; and she
cried to the queen of the underworld, the night-wanderer, to be propitious to
her enterprise. And Aesons son followed in fear, but the serpent, already
charmed by her song, was relaxing the long ridge of his giant spine, and
lengthening out his myriad coils, like a dark wave, dumb and noiseless, rolling
over a sluggish sea; but still he raised aloft his grisly head, eager to enclose
them both in his murderous jaws. But she with a newly cut spray of juniper,
dipping and drawing untempered charms from her mystic brew, sprinkled his
eyes, while she chanted her song; and all around the potent scent of the charm
cast sleep; and on the very spot he let his jaw sink down . . .
Hereupon Jason snatched the golden eece from the oak, at the maidens
bidding . . . Heavy it was, thickly clustered with ocks; and as he moved along,
even beneath his feet the sheen rose up from the earth. And he strode on now
with the eece covering his left shoulder from the height of his neck to his feet,
and now again he gathered it up in his hands; for he feared exceedingly, lest
some god or man should meet him and deprive him thereof . . .
Argonautica, Book IV, Fragments from Verses 123-182.
1
i.e. Jason and Medea
131
132 CHAPTER 8. LAGRANGIAN AND HAMILTONIAN FORMULATIONS
8.1 Introduction
This chapter is concerned with special mathematical formulations for both the water-
wave problem and the wave-body problem. The governing equations are derived from
variational or Lagrangian
2
principles, and it is shown that the nonlinear free surface
conditions and the hydrodynamic equations of motion for a oating body describe so-
called Hamiltonian
3
systems.
The application of variational principles to problems in uid dynamics goes back to
Kelvins (1849) minimum energy theorem and to Clebschs (1859) paper on the inte-
gration of the equations of motion of an ideal uid. The formulation of a Hamiltons
principle for an Eulerian description of the motion of an ideal uid brought Seliger and
Whitham (1968) to the conclusion that the Lagrangian density is simply the pressure;
Luke (1967) discovered that in the special case of a homogeneous uid with a free surface,
the corresponding Euler-Lagrange equations comprise not only Laplaces equation in the
interior of the uid but also the boundary conditions on the free surface.
Zakharov (1968) was the rst to note that the exact equations for waves on a perfect
uid of innite depth constitute a dynamical system with a positive denite Hamiltonian
functional, which represents the total energy of the uid. The vertical displacement of,
and the velocity potential at the free surface are canonical variables in Hamiltons sense.
This formalism was independently obtained for water waves on a uid of nite depth
by Broer (1974). With explicit use of this special Hamiltonian structure, a systematic
account of the symmetries and the corresponding conservation laws for water waves was
given by Benjamin and Olver (1982); the extension to wave-body problems is presented
in chapter 9.
The reader who is unfamiliar with but interested in variational principles and
Hamiltonian formulations is referred to the many textbooks on these and other related
subjects. In chronological order we recommend: Courant and Hilbert (1953, 1962),
Gelfand and Fomin (1963), Lanczos (1970), Landau and Lifshitz (1976), Finlayson (1972),
Weinstock (1974), Arnold (1978), and Goldstein (1980).
In section 8.2 Lukes variational principle for the classical water-wave problem is
discussed briey. A Hamiltonian formulation for water waves, in the form as it was
discovered by Zakharov and Broer, is discussed and linked with Lukes principle via
a Legendre
4
transformation in section 8.3. Then, we generalize these principles and
formulations to the more complex problems of wave-body interactions in section 8.4,
which is based on a paper recently written by van Daalen and van Groesen (1993).
8.2 Lukes variation principle for water waves
In this section Lukes (1967) variational principle for the classical water-wave problem is
resumed. The two-dimensional problem is described in terms of a horizontal coordinate
x and a vertical coordinate z; extension to three dimensions is straightforward and will
2
Lagrange, Joseph-Louis, Comte de (1736-1813), French mathematician. His principal work was
in pure mathematics, including dierential equations and the calculus of variations, and in mechanics
(Mecanique analytique, 1788). He presided over the committee which introduced the metric system
(1793).
3
Hamilton, Sir William Rowen (1805-1865), Irish mathematician and astronomer. His wide researches
include the origination and development of the theory of quaternions.
4
Legendre, Adrien-Marie (1752-1833), French mathematician. He made important contributions to
the theory of numbers and the theory of elliptic functions.
8.2. LUKES VARIATION PRINCIPLE FOR WATER WAVES 133
therefore not be discussed. Only the irrotational case is considered, which allows the
introduction of a potential representing the velocity eld v, that is v = .
So, let (x, z; t) be the velocity potential of a uid lying between a (not necessarily
even) bottom z = h(x) and a free surface z = (x; t), with gravity acting in the
negative z-direction, see Figure 8.1.
Figure 8.1: Free surface potential ow.
The variational principle for water waves, as it was rst proposed by Luke
5
, reads
J =
t
2

t
1
x
2

x
1
Ldxdt = 0 (8.1)
with the Lagrangian density
L(, ) =
(x;t)

h(x)
p dz =
(x;t)

h(x)

t
+
1
2

2
x
+
2
z

+gz

dz (8.2)
where p denotes the Bernoulli pressure; the water density is taken as unity from now on.
In (8.1-8.2), the potential and the free surface elevation are allowed to vary, but are
subject to the restrictions
= 0 , = 0 at x = x
1,2
and t = t
1,2
(8.3)
i.e. the variations and must vanish at the end points of the spatial and time
intervals.
Following the usual procedure in the calculus of variations, (8.1-8.2) becomes
J =
t
2

t
1
x
2

x
1

t
+
1
2

2
x
+
2
z

+gz

z=

dxdt
5
From now on, Luke refers to his 1967-paper, unless stated otherwise.
134 CHAPTER 8. LAGRANGIAN AND HAMILTONIAN FORMULATIONS

t
2

t
1
x
2

x
1

(x;t)

h(x)
(
t
+
x

x
+
z

z
) dz

dxdt = 0 . (8.4)
The contributions from the second integral are rewritten to

(x;t)

h(x)

t
dz =

t
(x;t)

h(x)
dz +
t
[]
z=
, (8.5)

(x;t)

h(x)

x
dz =

x
(x;t)

h(x)

x
dz +
(x;t)

h(x)

xx
dz
+
x
[
x
]
z=
+h
x
[
x
]
z=h
, (8.6)

(x;t)

h(x)

z
dz =

z
(x;t)

h(x)

z
dz +
(x;t)

h(x)

zz
dz
= [
z
]
z=
+ [
z
]
z=h
+
(x;t)

h(x)

zz
dz . (8.7)
The rst term on the right-hand side of (8.5) integrates out to the end points of [t
1
, t
2
],
and hence vanishes. The rst term on the right-hand side of (8.6) integrates out to the
end points of [x
1
, x
2
], and therefore equals zero. Substitution of the remaining terms
into (8.4) yields
J =
t
2

t
1
x
2

x
1

(x;t)

h(x)

2
dz

t
+
1
2

2
x
+
2
z

+gz

z=

dxdt
+
t
2

t
1
x
2

x
1

[(
t
+
x

z
) ]
z=
+ [(h
x

x
+
z
) ]
z=h

dxdt
= 0 . (8.8)
First, choose = 0, []
z=h
= 0, and []
z=
= 0; since is arbitrary otherwise, it
follows that

2
= 0 for h(x) < z < (x; t) (8.9)
Then, since , []
z=
and []
z=h
may be given arbitrary independent values, we
obtain
p =

t
+
1
2

2
x
+
2
z

+gz

= 0 at z = (x; t) (8.10)
and

t
+
x

z
= 0 at z = (x; t) (8.11)
8.2. LUKES VARIATION PRINCIPLE FOR WATER WAVES 135
and
h
x

x
+
z
= 0 at z = h(x) (8.12)
respectively.
Evidently, (8.9) is Laplaces equation for the velocity potential, i.e. the continuity
equation for incompressible, inviscid and irrotational ow. Condition (8.10) states that
the Bernoulli pressure vanishes at the free surface, and it is known as the dynamic free
surface condition. Condition (8.11) expresses that the free surface is a material boundary
which means that uid particles can not leave it and it is known as the kinematic free
surface condition. Finally, (8.12) is a zero-ux condition expressing the impermeability
of the bottom.
Thus it has been shown that (8.1-8.2), being equivalent to the eld equation (8.9)
and the physical boundary conditions (8.10-8.12), is a proper variational principle for the
classical water-wave problem. In other words, Lukes variation principle describes the
irrotational motion of an inviscid and incompressible uid with a free surface under the
inuence of gravity.
Remark 1: The above variational principle is easily modied such that the eects of
surface tension are accounted for. With denoting the coecient of surface tension, the
net extra force per unit area is given by
T

() =

x
(1 +
2
x
)
1/2
. (8.13)
The appropriate variation principle is then given by (8.1), where the Lagrangian density
L is replaced by
L

(, ) = L(, ) +T

() =
(x;t)

h(x)
p dz

x
(1 +
2
x
)
1/2
. (8.14)
Variation of T

with respect to , and retaining terms up to rst order only, gives


T

() =

x
(1 +
2
x
)
1/2
. (8.15)
Then, integrating by parts with respect to x, and using the vanishing of at the end
points x = x
1,2
, one arrives at the following set of free surface conditions confer (8.10-
8.11)
p =

t
+
1
2

2
x
+
2
z

+gz

=

xx

1 +
2
x

3/2

t
+
x

z
= 0

at z = (x; t) ,
(8.16)
showing that surface tension, known to act like a stretched membrane on the free surface,
indeed results in a positive pressure at the crest of a wave, where
xx
< 0, and a negative
pressure at the trough of a wave, where
xx
> 0.
136 CHAPTER 8. LAGRANGIAN AND HAMILTONIAN FORMULATIONS
Linearization of (8.16) about the mean water level yields the following simplied set
of free surface conditions confer (2.35)

t
= g +
xx

t
=
z

at z = 0 . (8.17)
For the case of small amplitude waves
6
travelling over a horizontal bottom, analyti-
cal solutions have been presented in section 2.4; substitution of expressions (2.37-2.38)
into (8.17) yields the rst order
7
dispersion relation confer (2.39)

2
= (gk tanh kh)

1 +

g
k
2

. (8.18)
For water waves under gravity, we have g = 9.81 m/s
2
, = 1000 kg/m
3
, and =
0.073 N/m, so that the minimum wavelength is given by

m
= 2

1/2
= 1.71 cm , (8.19)
and surface tension eects become negligible for wavelengths several times greater than
this minimum value.
The special choice of the integrated Bernoulli pressure for the Lagrangian density en-
abled Luke to derive the complete set of governing equations for the classical water-wave
problem from a single variational principle. Being aware of the arbitrariness in the choice
of the Lagrangian density, he noticed that this particular choice is more productive than
the traditional form of the Lagrangian density, being
L

(, ) = /(, ) { () =
(x;t)

h(x)
1
2

2
x
+
2
z

dz
1
2
g

2
h
2

, (8.20)
i.e. kinetic minus potential energy. In Lukes own words, the key to the dierence appears
to be conservation of mass; at the outset of (8.9) and (8.11-8.12), variation of L

with
respect to leads to the dynamic free surface condition (8.10). The surprising thing
is that Luke, by raising this point, undoubtedly has been very close to a dierent, but
related, description of the system under consideration; a Hamiltonian formulation for the
water-wave problem.
8.3 Zakharov & Broers Hamiltonian formulation
To the best of our knowledge, Zakharov (1968) was the rst to show that the governing
equations for waves on the surface of an innitely deep uid describe a so-called Hamil-
tonian system; the evolution equations can be expressed in terms of surface variables and
a suitably chosen Hamiltonian density only. Few years later, Broer (1974) independently
established the proof of an equivalent theorem for water waves travelling over a horizontal
6
That is, for waves with small amplitude compared to the wavelength and the mean water depth.
7
That is, to rst order in the wave amplitude.
8.3. ZAKHAROV & BROERS HAMILTONIAN FORMULATION 137
bottom. A corrected proof of the latter result, with applications towards approximative
wave equations and conservation laws, was given two years later by Broer, van Groesen,
and Timmers (1976). Other contributions with regard to Hamiltonian theories for water
waves are due to Miles (1976, 1977).
Many approximate model equations exhibit a Hamiltonian structure analogous to the
exact governing equations; examples are the shallow water equations, see Salmon (1983),
the Boussinesq equations, see Whitham (1965), and the Korteweg-de Vries equations, see
for instance Broer (1975), Gardner (1971), and van Groesen, van Beckum, and Valkering
(1990). Stability and bifurcation properties of water waves are directly related to the
Hamiltonian structure and the symmetries; see Mackay and Saman (1986), Saman
(1985, 1988), and Zuria (1987). Some applications of Lagrangian and Hamiltonian
formulations for stratied uids have been given by Henyey (1983) and Miles (1986ab).
In most discussions of Hamiltonian formulations for free surface waves, no reference is
made to the corresponding Lagrangian (variational) principles. It is well known that for
discrete systems the transition from a Lagrangian to a Hamiltonian formulation is easily
established through a so-called Legendre transformation; in the following intermezzo this
concept is illustrated for a simple n-particle system. Next, it will be shown that Lukes
Lagrangian principle and the Hamiltonian formulation due to Zakharov and Broer are
linked through the Legendre transformation for continuous systems.
8
Intermezzo: Consider a one-dimensional n-particle system under the action of a central
force eld
F (q
i
) =
V
q
i
, (8.21)
where q
i
is the position of particle i; the mass of each particle is denoted by m
i
.
The evolution of this system with n degrees of freedom is described by n ordinary
dierential equations of the form
d
dt

L
q
i

L
q
i
= 0 , (8.22)
which are familiar as the Euler-Lagrange equations. In (8.22) L is the system Lagrangian,
dened as kinetic minus potential energy:
L = K P =
n

i=1

1
2
m
i
q
2
i
V (q
i
)

. (8.23)
By substitution of (8.23) into (8.22), the explicit equations of motion are obtained:
m
i
q
i
= F (q
i
) . (8.24)
Since all equations are of second order, the motion of this particular system is determined
for all time only when 2n initial values are specied. For instance, one can specify the
n q
i
s and the n q
i
s at some time t
1
, or the n q
i
s at two times t
1
and t
2
. From the
Lagrangian point of view the problem is dened in terms of n independent variables
q
i
(t), and q
i
appears only as shorthand for the time derivative of q
i
.
8
This connection between Lukes variation principle and Zakharovs Hamiltons principle was already
noticed by van Groesen (1978).
138 CHAPTER 8. LAGRANGIAN AND HAMILTONIAN FORMULATIONS
In a Hamiltonian formulation the evolution of this n-particle system is described in terms
of rst-order ordinary dierential equations. Since the number of conditions determining
the motion is still 2n, there must be 2n independent rst order equations expressed
in 2n independent variables. In thus doubling the set of independents, it is natural
to choose half of them to be the generalized coordinates q
i
. The formulation becomes
nearly symmetric if we choose the other half of the set to be the generalized conjugate
momenta p
i
, dened by the so-called Legendre transformation
p
i
=
L
q
i
= m
i
q
i
. (8.25)
The quantities (q
i
, p
i
) are known as the canonical variables. The Hamiltonian density H
is dened as
H =
n

i=1
p
i
q
i
L =
n

i=1

1
2
m
i
q
2
i
+V (q
i
)

= K +P , (8.26)
i.e. the sum of kinetic and potential energy.
The canonical equations read
d
dt

p
i
q
i

0 1
1 0

H/p
i
H/q
i

, (8.27)
being equivalent to (8.22).
From the Hamiltonian viewpoint the problem is dened in terms of 2n independent
variables, namely n q
i
s and n p
i
s. From a purely mathematical point of view, the
transition from a Lagrangian to a Hamiltonian formulation corresponds to changing the
variables following the Legendre transformation (8.25).
Finally, we note that H, as a function of the generalized coordinates and momenta,
is a constant of the motion; this follows directly from
dH
dt
=
n

i=1

H
p
i
p
i
+
H
q
i
q
i

(8.28)
and substitution of the canonical equations (8.27).
For a continuous system with generalized coordinate , the transition from a Lagrangian
to a Hamiltonian formulation is represented by
L(,
t
) H(, ) =
t
L(, ) , (8.29)
where the generalized momentum is dened by
=
L

t
. (8.30)
The innite-dimensional quantities and are then known as the canonical conjugate
variables of the system.
Let us return to the classical water-wave problem, for which Luke dened the Lagrangian
density as
L(, ) =
(x;t)

h(x)

t
+
1
2

2
x
+
2
z

+gz

dz . (8.31)
8.3. ZAKHAROV & BROERS HAMILTONIAN FORMULATION 139
With the free surface elevation in the natural role of generalized coordinate, the gen-
eralized momentum (conjugate to ) follows from

(x;t)

h(x)

t
dz =

t
(x;t)

h(x)
dz +
t
[]
z=
. (8.32)
Clearly, application of (8.30) yields
L

t
= []
z=
, (8.33)
i.e. the restriction of the velocity potential to the free surface; this variable is denoted
by from now on.
Then, with (8.31-8.32), the Hamiltonian density follows from the Legendre transfor-
mation (8.29):
H(, ) = /(, ) +{ () =
(x;t)

h(x)
1
2

2
x
+
2
z

dz +
1
2
g

2
h
2

(8.34)
i.e. the sum of the kinetic and potential energy densities. In this particular application
of the Legendre transformation it has been assumed that the rst term on the right-hand
side of (8.32) i.e. the partial time derivative of the integrated potential integrates
out to the end points of the time interval under consideration and vanishes there.
With the above denitions, and with satisfying the boundary value problem

2
= 0 for h(x) < z < (x; t) , (8.35)

n
= 0 at z = h(x) , (8.36)
(x, z; t) = (x; t) at z = (x; t) , (8.37)
the following theorem can be deduced:
Theorem 3 : Hamiltonian formulation for the water-wave problem
(Zakharov, 1968 / Broer, 1974)
The equations of motion for gravity driven water waves describe an innite-dimensional
Hamiltonian system in the canonically conjugate variables and and with the energy
density H(, ) as Hamiltonian density; the canonical equations

0 1
1 0

(8.38)
are equivalent with the nonlinear free surface conditions (8.10-8.11).
The proof of the above theorem is based on the following lemma:
140 CHAPTER 8. LAGRANGIAN AND HAMILTONIAN FORMULATIONS
Lemma 1 : Variational derivatives of the kinetic energy density
The variational derivatives of the kinetic energy density /(, ) are given by

/(, ) =

1 +

1/2

z=
, (8.39)

/(, ) =

1
2
( )

/(, )

z

z=
. (8.40)
Proof: First we prove (8.39); keep xed and let vary such that its variation
corresponds to a change in the free surface potential . Then the rst variation in
/(, ) reads
/(, ; ) =
(x;t)

h(x)
dz =
(x;t)

h(x)

()
2

dz . (8.41)
With Gauss divergence theorem and (8.35-8.36) there results
/(, ; ) = [( n) ds ]
z=
, (8.42)
where ds denotes the local arclength of the free surface contour corresponding to an in-
nitesimal horizontal distance dx; in terms of the generalized coordinate this arclength
reads
ds =

1 +

1/2
. (8.43)
The last two results lead directly to
/(, ; ) =

1 +

1/2

z=
, (8.44)
from which (8.39) follows.
Next, vary and assume that the solution of the boundary value problem (8.35-
8.37) is correspondingly modied for the varied uid domain. At the modied free surface
we have to lowest order:
(x, +; t) = (x, ; t) +

z=
(x; t) +

(x; t) , (8.45)
where the variation

corresponding to is given by

z=
. (8.46)
With the above result the total eect of a variation in /(, ) is found to be
/(, ; ) +/(, ;

) =

1
2
( )

z=
, (8.47)
8.4. WAVE-BODY FORMULATIONS 141
and hence, to lowest order

/(, ) +

/(, )

z=
=

1
2
( )

z=
, (8.48)
from which (8.40) follows.
Proof of theorem 3: The expressions for the variational derivatives of the potential
energy { () read

{ () = 0 , (8.49)

{ () = g . (8.50)
For the partial time derivative of the free surface potential we write

t
=

t
+

t

z=
. (8.51)
In terms of the generalized coordinate , the unit normal vector along the free surface
reads
n =
(
x
, 1)
T
(1 +
2
x
)
1/2
. (8.52)
Then, with (8.39) and (8.49) the second canonical equation
t
=

H is easily shown to
be equivalent with
[
t
+
x

z
]
z=
= 0 , (8.53)
i.e. the kinematic free surface condition (8.11).
Then, with (8.40) and (8.50-8.53), the rst canonical equation
t
=

H yields

t
+
1
2
( ) +gz

z=
= 0 , (8.54)
i.e. the dynamic free surface condition (8.10).
Thus it has been proven that with the free surface elevation and the free surface
potential as canonical coordinates, with the energy density as Hamiltonian density
H(, ) and with satisfying the boundary value problem (8.35-8.37), the Hamiltonian
equations (8.38) are equivalent with the free surface conditions (8.53-8.54).
8.4 Wave-body formulations
In this section both Lukes variation (Lagrangian) principle and the Hamiltonian for-
mulation for water waves due to Zakharov and Broer are extended to the wave-body
problem. The complete set of governing equations for the three-dimensional problem of
nonlinear water waves in hydrodynamic interaction with bodies oating in or below the
free surface is derived from a single variational principle. In addition, we show that the
nonlinear free surface conditions and the hydrodynamic equations of motion for a freely
142 CHAPTER 8. LAGRANGIAN AND HAMILTONIAN FORMULATIONS
Figure 8.2: Wave-body system.
oating body constitute an innite-dimensional Hamiltonian system in terms of properly
chosen canonical coordinates and the total energy as Hamiltonian.
The system under consideration consists of a uid, bounded by the impermeable bottom
B (which is not necessarily even), the free surface F, and the wetted surface S of a
oating body (either partially or totally submerged), see Figure 8.2. In the horizontal
directions x and y, the uid domain is cut o by a cylindrical vertical surface
R
of
innite radius R, extending from the bottom to the free surface. The transient uid
domain is denoted by (t).
The body mass and moments about its principal axes of inertia are denoted by M
and

I = (I
1
, I
2
, I
3
)
T
respectively. The position of the body is specied by its centre of
gravity x
G
= (x
1
, x
2
, x
3
)
T
, and the body orientation is denoted by

G
= (
1
,
2
,
3
)
T
.
8.4.1 A variation principle for wave-body interactions
The Lagrangian for the uid is given by
L
f
=

(t)

p d =

(t)

t
+
1
2
( ) +gz

d (8.55)
i.e. the Bernoulli pressure integrated over the uid domain; the uid density is taken as
unity.
The Lagrangian for the body is dened as the kinetic energy minus the potential
energy:
L
b
= /
b
{
b
(8.56)
8.4. WAVE-BODY FORMULATIONS 143
where
/
b
=
1
2
M

x
G


x
G
+
1
2

G
(8.57)
{
b
= Mge
3
x
G
(8.58)
In (8.57) the symbol is used to dene the component-wise product of two vectors:
a

b (a
1
b
1
, a
2
b
2
, a
3
b
3
)
T
.
The Lagrangian for the total system, consisting of the uid and the body, is then
simply given as the sum of the separate Lagrangians:
L
s
= L
f
+L
b
(8.59)
With (8.55-8.59) the proposed variational principle reads
J =
t
2

t
1
L
s
dt = 0 (8.60)
for all variations in the free surface elevation , the velocity potential , and the body
position x
G
and orientation

G
. These variations are subject to the restrictions that they
vanish at the end points of the time interval (i.e. at times t = t
1
and t = t
2
) and on the
vertical boundary at innity (i.e. on
R
).
Following the standard procedure in the calculus of variations, (8.55-8.60) yields
J =
t
2

t
1

p dS +

p (x
S
n) dS

(t)

(
t
+ ) d

dt
+
t
2

t
1

M

x
G

x
G
Mge
3
x
G
+

dt = 0 . (8.61)
Here x
S
denotes the position of a point on the wetted body surface S; the change in x
S
due to variations in x
G
and

G
is given by
x
S
= x
G
+

G
r , (8.62)
where r = x
S
x
G
has been introduced to dene the position of x
S
relative to x
G
, see
Figure 8.2.
Taking into account the motion of (t) we may write

(t)


t
d =

t

(t)


t
dxdy

x
S
n

dS , (8.63)
where n is the unit normal vector along S. The rst term on the right-hand side
of (8.63) vanishes due to the restriction = 0 at times t = t
1
and t = t
2
.
With Greens rst identity we obtain

(t)

d =


n
dS

(t)


2
d . (8.64)
144 CHAPTER 8. LAGRANGIAN AND HAMILTONIAN FORMULATIONS
Due to the restriction = 0 on
R
the corresponding contribution to the surface
integral on the right-hand side of (8.64) vanishes.
Integrating by parts and using the restrictions x
G
=

0 and

G
=

0 at times t = t
1
and t = t
2
gives
t
2

t
1
M

x
G

x
G
dt =
t
2

t
1
M

x
G
x
G
dt , (8.65)
t
2

t
1

G
dt =
t
2

t
1

G
dt . (8.66)
On the free surface F, the area dS corresponding to innitesimal horizontal distances dx
and dy is given by
dS = dxdy

1 +

2
+

1/2
, (8.67)
and the unit normal vector on F can be expressed as
n =
(
x
,
y
, 1)
T

1 +
2
x
+
2
y

1/2
. (8.68)
With (8.61-8.68), the proposed variational principle reads
J =
t
2

t
1

p dS +

(t)


2
d

dt

t
2

t
1

x


y

y
+

z


t

dxdy

dt

t
2

t
1

n


x
S
n

dS +


n
dS

dt
+
t
2

t
1

pndS Mge
3
M

x
G

x
G

dt
+
t
2

t
1

p (r n) dS

dt = 0 . (8.69)
From this it is clear that invariance of J with respect to a variation in the free surface
elevation yields the dynamic free surface condition
p =

t
+
1
2
( ) +gz

= 0 on F . (8.70)
8.4. WAVE-BODY FORMULATIONS 145
Similarly, invariance of J with respect to a variation in the velocity potential yields
the eld equation

2
= 0 in (t) , (8.71)
the kinematic condition on the free surface

t
+

x

x
+

y

y
=

z
on F , (8.72)
the contact condition on the wetted body surface

n
=

x
S
n on S , (8.73)
and the impermeability (zero-ux) condition at the bottom

n
= 0 on B . (8.74)
Finally, invariance of J with respect to variations in the body position x
G
and orientation

G
yields the equations of motion for the body:

pndS Mge
3
= M

x
G
, (8.75)

p (r n) dS =

I

G
. (8.76)
Thus it has been proven that (8.60) with (8.55-8.59) is a proper variation principle for
nonlinear gravity waves interacting with a body oating freely in or below the free surface.
For completeness, it is noted that the extension to multiple independently oating
bodies is straightforward. We shall close this paragraph with a discussion of other
less obvious extensions which are signicant from a practical point of view.
Remark 2: The above formulation is easily extended such that the eects of restoring
forces and moments acting on the body are accounted for. Suppose that the
conservative forces and moments are given by

F
r
=
1
x
G
,

L
r
=
1

G
, (8.77)
where 1

x
G
,

denotes the restoring potential. Then, the potential energy of the


body is given by
{
b
= Mge
3
x
G
+1

x
G
,

. (8.78)
Applying the same procedure as demonstrated above, we arrive at the following set of
modied equations of motion:

pndS Mge
3
+

F
r
= M

x
G
, (8.79)

p (r n) dS +

L
r
=

I

G
. (8.80)
146 CHAPTER 8. LAGRANGIAN AND HAMILTONIAN FORMULATIONS
Remark 3: Non-conservative external forces and moments acting on the body may be
incorporated in the above variational description as well. The virtual work done by such
forces and moments (denoted by

F
e
and

L
e
respectively) during the time interval [t
1
, t
2
]
is given by
J(t
1
; t
2
) =
t
2

t
1
dJ =
t
2

t
1

F
e
dx
G
+

L
e
d

. (8.81)
Now the proposed variational principle reads
J =
t
2

t
1
L
s
dt +J(t
1
; t
2
) = 0 , (8.82)
for all variations in , , x
G
, and

G
subject to the usual restrictions.
Following the standard procedure in the calculus of variations, we arrive at the mod-
ied equations of motion for the body:

pndS Mge
3
+

F
e
= M

x
G
, (8.83)

p (r n) dS +

L
e
=

I

G
. (8.84)
Remark 4: Finally, we discuss the extension of the above variation principle to the
description of phenomena as overturning waves. The key dierence is that the single-
valued free surface elevation is replaced by the three-component position vector x
F
to
describe the free surface prole. With (8.55-8.60) unaltered, the invariance of J with
respect to a variation in now yields the modied kinematic free surface condition

x
F
n =

n
on F , (8.85)
which includes phenomena as overturning waves, and permits a Lagrangian description
of the free surface:
Dx
F
Dt
= on F . (8.86)
8.4.2 A Hamiltonian formulation for wave-body interactions
Next, the Hamiltonian theory for the classical water-wave problem is extended to the
interaction with bodies oating in or below the free surface.
For the body we dene the coordinate vector

as the combination of the position
vector x
G
and the orientation vector

G
:

x
G

. (8.87)
Similarly, the normal vector along the wetted body surface S is dened as

n
r n

, (8.88)
8.4. WAVE-BODY FORMULATIONS 147
and the diagonal mass matrix is dened as
diag () [M, M, M, I
1
, I
2
, I
3
]
T
. (8.89)
With the above denitions the Lagrangian for the system can be written as con-
fer (8.55-8.59)
L
s
=

(t)

t
+
1
2
( ) +gz

d +
1
2

ge
3

(8.90)
Taking into account the evolution of the uid domain, and integrating by parts, we may
write
L
s
=

t

(t)

d +



t
dxdy +

dS

(t)

1
2
( ) +gz

d
1
2

ge
3

, (8.91)
where

is the normal velocity of a point on S. The rst term on the right-hand side
of (8.91) integrates out to the end points of the time interval considered.
Considering (8.91) we may put formally
L
s
(
s
,
s
) =
s

s
H
s
(
s
,
s
) (8.92)
with the following denitions: the canonical coordinates are given by

s
= (
f
,
b
) with
f
and
b

(8.93)
i.e. the free surface elevation and the body position and orientation.
The canonical conjugate momenta are given by

s
= (
f
,
b
) with
f
and
b

dS (8.94)
i.e. the free surface potential []
z=
and the rigid body impulse plus a momentum
contribution of the velocity potential over the wetted body surface.
9
The Hamiltonian is given by
H
s
=

(t)

1
2
( ) +gz

d +
1
2

+ge
3

(8.95)
9
In uid mechanics this integral contribution is known as the Kelvin impulse, see also Benjamin
(1987). A physical interpretation is given by Lamb (1932) in his discussion of the impulsive generation
of motion. As Lamb says, a sudden alteration in the motion of a uid may take place when a solid
immersed in the uid is suddenly set in motion. The reverse case is, of course, also true; a sudden
change in the uid motion causes a transfer of momentum from the uid to the solid. The resultant
impulsive pressure can be shown to satisfy Laplaces equation, and from its formal denition it is clear
that, for our problem, this quantity is exactly the velocity potential. The surface integral obviously
accounts for the transfer of momentum (or energy) from the uid to the oating body and vice versa.
See also Lambs discussion of the motion of solids through an innite liquid.
148 CHAPTER 8. LAGRANGIAN AND HAMILTONIAN FORMULATIONS
i.e. the total energy of the system.
Then, with satisfying the boundary value problem

2
= 0 in (t) , (8.96)

n
=

on S , (8.97)

n
= 0 on B , (8.98)
(x, y, z) = (x, y) on F , (8.99)
we can establish the following theorem:
Theorem 4 : Hamiltonian formulation for the wave-body problem
The equations of motion for gravity driven water waves interacting with bodies oating
in or below the free surface describe an innite dimensional Hamiltonian system in the
canonically conjugate variables
s
and
s
and with the total energy H
s
as Hamiltonian;
the canonical equations

0 1
1 0

s
H
s

s
H
s

(8.100)
are equivalent with the nonlinear free surface conditions (8.70) and (8.72) and the equa-
tions of motion for the body (8.75-8.76).
The rst canonical equation
t

s
=

s
H
s
is shorthand for

f
t
=
H
s

f
, (8.101)
and

b
t
=
H
s

b
. (8.102)
Similarly, the second canonical equation
t

s
=

s
H
s
is shorthand for

f
t
=
H
s

f
, (8.103)
and

b
t
=
H
s

b
. (8.104)
Proof of (8.103): Keeping xed and varying such that its variation corresponds
to a change in the free surface potential , the rst variation in /
f
reads
/
f
() =

(t)

d =


n
dS

(t)


2
d , (8.105)
from which we obtain

/
f
=

1 +

2
+

1/2

F
. (8.106)
8.4. WAVE-BODY FORMULATIONS 149
Then, with

{
f
= 0 and

H
b
= 0, and with (8.67) it is seen that (8.103) is equivalent
with the kinematic free surface condition

t
+

x

x
+

y

y
=

z
on F . (8.107)
Proof of (8.101): Here we must be aware of the fact that a variation in the free
surface elevation naturally leads to a variation in the free surface potential . To rst
order we have
( +) = () +

F
+

, (8.108)
where

F
. (8.109)
Then, the eect of a variation in /
f
is found to be
/
f
() +/
f
(

) =

1
2
( )

F
, (8.110)
hence

/
f
+

/
f

F
=

1
2
( )

F
. (8.111)
With

{
f
= g and

H
b
= 0, and the relation

t
=

t
+

z

F
, (8.112)
it follows that (8.101) is equivalent with the dynamic free surface condition

t
+
1
2
( ) +gz = 0 on F . (8.113)
Proof of (8.104): This is more complicated, due to the fact that the canonical momen-
tum of the body depends on both

and along S, see (8.94). Thus, a variation in


b
corresponds to variations

and according to

b
=

dS . (8.114)
Using (8.96), it can be shown that a variation in the potential along S yields the rst
variation in /
f
, given by
/
f
() =


n
dS . (8.115)
With (8.97) this can be written as
/
f
() =

dS . (8.116)
150 CHAPTER 8. LAGRANGIAN AND HAMILTONIAN FORMULATIONS
It is obvious that the rst variation of /
b
due to a variation in

is given by
/
b

. (8.117)
With the nullity of the variational derivatives of {
f
and {
b
with respect to and

, we
obtain
H
s
(
b
) =

dS

=
b

b
, (8.118)
which completes the proof of (8.104).
Proof of (8.102): Note that a variation in

leads to a variation in H
f
, according to
H
f

S

1
2
( ) +gz

dS . (8.119)
The rst variation in H
b
due to a variation in

is given by
H
b

= ge
3

. (8.120)
It is then easy to show that (8.102) is equivalent with the equations of motion for the
oating body:

p dS ge
3
=

. (8.121)
This completes the proof of theorem 4.
Remark 5: The left-hand side of (8.102) represents the change of momentum of the
body, i.e. it is the force related to the change of the body velocity. The right-hand
side of (8.102) represents the force due to a change in the position of the body. For
a single body immersed in an innite uid in the absence of external forces the
total energy H
s
is independent of the body position; for this special case the right-hand
side expression equals zero. This leads us to dAlemberts
10
paradox, stating that no
hydrodynamic force acts on a body moving with constant translational velocity in an
innite and inviscid uid which is free of vorticity.
Remark 6: In the canonical momentum for the body see (8.94) the integral over
the wetted body surface S accounts for the transfer of momentum from the uid to the
body (and vice versa) through the velocity potential. This transfer becomes more clear
10
Alembert, Jean le Rond d (1717-1783), French mathematician and philosopher. In his Traite de la
dynamique (1743) he proved that Newtons third law (to every action there is an equal and opposite
reaction) applies to moving as well as stationary bodies. This is known as dAlemberts principle.
8.4. WAVE-BODY FORMULATIONS 151
if we consider the product of the canonical body coordinate and the canonical body
momentum:

b
=

dS



n
dS , (8.122)
where we used (8.97). The rst term on the right-hand side equals twice the kinetic
energy of the body as if it moved autonomously. The second term represents the energy
transferred from the uid to the body. This contribution may be called added energy,
and obviously it is related to concepts like added mass and damping for a body oscillating
with small amplitude in a uid.
152 CHAPTER 8. LAGRANGIAN AND HAMILTONIAN FORMULATIONS
8.5 Bibliography
Several relevant publications which have not been cited in the text are also included here.
Abarbanel, H.D.I., Brown, R., and Yang, Y.M. 1988. Hamiltonian formulation
of inviscid ows with free boundaries. The Physics of Fluids 31:2802-2809.
Aranha, J.A.P., and Pesce, C.P. 1989. A variational method for water wave radi-
ation and diraction problems. Journal of Fluid Mechanics 204:135-157.
Arnold, V.I. 1978. Mathematical Methods of Classical Mechanics. Springer-Verlag.
Bateman, H. 1932. Partial Dierential Equations of Mathematical Physics. Cambridge
University Press.
Becker, J.M., and Miles, J.W. 1991. Standing radial cross-waves. Journal of Fluid
Mechanics 222:471-499.
Benjamin, T.B. 1984. Impulse, ow force and variational principles. IMA Journal of
Applied Mathematics 32:3-68.
Benjamin, T.B. 1987. Hamiltonian theory for motions of bubbles in an innite liquid.
Journal of Fluid Mechanics 181:349-379.
Benjamin, T.B., and Olver, P.J. 1982. Hamiltonian structure, symmetries and
conservation laws for water waves. Journal of Fluid Mechanics 125:137-185.
Bridges, T.J., and Dias, F. 1989. Group-theoretic considerations lead to new solu-
tions of the water wave problem. The Fourth International Workshop on Water Waves
and Floating Bodies, ystese, Norway.
Broer, L.J.F. 1974. On the Hamiltonian theory of surface waves. Applied Scientic
Research 29:430-446.
Broer, L.J.F. 1975. Approximate equations for long water waves. Applied Scientic
Research 31:377-395.
Broer, L.J.F., van Groesen, E.W.C., and Timmers, J.M.W. 1976. Stable model
equations for long water waves. Applied Scientic Research 32:619-636.
Clebsch, A. 1859.

Uber die Integration der hydrodynamischen Gleichungen. Journal
f ur Reine und Angewandte Mathematik 56:110.
Courant, R., and Hilbert, D. 1953. Methods of Mathematical Physics I. Inter-
science Publishers.
Courant, R., and Hilbert, D. 1960. Methods of Mathematical Physics II. Inter-
science Publishers.
Creamer, D.B., Henyey, F., Schult, R., and Wright, J. 1989. Improved linear
representation of ocean surface waves. Journal of Fluid Mechanics 205:135-161.
van Daalen, E.F.G., and van Groesen, E.W.C. 1993. A Hamiltonian formulation
for water waves and oating bodies. Submitted to the Journal of Ship Research.
Dingemans, M.W. 1993. Water Wave Propagation over Uneven Bottoms. World Sci-
entic.
Finlayson, B.A. 1972. The Method of Weighted Residuals and Variational Principles.
Academic Press.
8.5. BIBLIOGRAPHY 153
Gardner, C.S. 1971. Korteweg-de Vries equation and generalizations IV. The Korteweg-
de Vries equation as a Hamiltonian system. Journal of Mathematical Physics 12(8):1548-
1551.
Gelfand, I.M., and Fomin, S.V. 1963. Calculus of Variations. Prentice-Hall.
Goldstein, H. 1980. Classical Mechanics. Addison-Wesley.
van Groesen, E.W.C. 1978. Variational Methods in Mathematical Physics. Ph.D.
thesis, Eindhoven University. Eindhoven, The Netherlands.
van Groesen, E.W.C., van Beckum, F.P.H., and Valkering, T.P. 1990. Decay
of travelling waves in dispersive Poisson systems. Journal of Applied Mathematics and
Physics (ZAMP) 41:501-523.
Henyey, F.S. 1983. Hamiltonian description of stratied uid dynamics. The Physics
of Fluids 26:40-47.
Katopodes, N.D., and Dingemans, M.W. 1989. Hamiltonian approach to surface
wave models. HYDROCOMP89, Dubrovnik, Yugoslavia. Elsevier.
Kelvin, Lord. 1849. On the vis-viva of a liquid in motion. Cambridge and Dublin
Mathematical Journal.
Lamb, Sir H. 1932. Hydrodynamics. Cambridge University Press.
Lanczos, C. 1970. The Variational Principles of Mechanics. University of Toronto
Press.
Landau, L.D., and Lifshitz, E.M. 1976. Mechanics. Volume 1 of Course of Theo-
retical Physics. Pergamon Press.
Luke, J.C. 1967. A variational principle for a uid with a free surface. Journal of Fluid
Mechanics 27(2):395-397.
Mackay, R.S., and Saffman, P.G. 1986. Stability of water waves. Proceedings of
the Royal Society of London, Series A, 406:115-125.
McIntyre, M.E., and Shepherd, T.G. 1987. An exact local conservation theorem for
nite-amplitude disturbances to non-parallel shear ows, with remarks on Hamiltonian
structure and on Arnolds stability theorems. Journal of Fluid Mechanics 181:527-565.
Milder, D.M. 1977. A note regarding On Hamiltons principle for surface waves.
Journal of Fluid Mechanics 83(1):159-161.
Milder, D.M. 1982. Hamiltonian dynamics of internal waves. Journal of Fluid Me-
chanics 119:269-282.
Milder, D.M. 1990. The eect of truncation on surface-wave Hamiltonians. Journal
of Fluid Mechanics 217:249-262.
Miles, J.W. 1976. Nonlinear surface waves in closed basins. Journal of Fluid Mechanics
75(3):419-448.
Miles, J.W. 1977. On Hamiltons principle for surface waves. Journal of Fluid Me-
chanics 83:153-158, with a note by D.M. Milder (1977).
Miles, J.W. 1981. Hamiltonian formulations for surface waves. Applied Scientic Re-
search 37:103-110.
Miles, J.W. 1986a. Weakly nonlinear waves in a stratied uid: a variational formu-
154 CHAPTER 8. LAGRANGIAN AND HAMILTONIAN FORMULATIONS
lation. Journal of Fluid Mechanics 172:499-512.
Miles, J.W. 1986b. Weakly nonlinear Kelvin-Helmholtz waves. Journal of Fluid Me-
chanics 172:513-529.
Miles, J.W., and Henderson, D. 1990. Parametrically forced surface waves. Annual
Review of Fluid Mechanics 22:143-165.
Miloh, T. 1984. Hamiltons principle, Lagranges method and ship motion theory. Jour-
nal of Ship Research 28(4):229-237.
Olver, P.J. 1982. A nonlinear Hamiltonian structure for the Euler equations. Journal
of Mathematical Analysis and Applications 89:233-250.
Radder, A.C. 1992. An explicit Hamiltonian formulation of surface waves in water of
nite depth. Journal of Fluid Mechanics 237:435-455.
Saffman, P.G. 1985. The superharmonic instability of nite-amplitude water waves.
Journal of Fluid Mechanics 159:169-174.
Saffman, P.G. 1987. Application of Hamiltonian methods to the structure and stabil-
ity of water waves of permanent form. IUTAM Symposium on Nonlinear Water Waves,
Tokyo, Japan. Springer-Verlag.
Salmon, R. 1983. Practical use of Hamiltons principle. Journal of Fluid Mechanics
132:431-444.
Salmon, R. 1988. Hamiltonian uid mechanics. Annual Review of Fluid Mechanics
20:225-256.
Seliger, R.L., and Whitham, G.B. 1968. Variational principles in continuum me-
chanics. Proceedings of the Royal Society of London, Series A, 305:1-25.
Weinstock, R. 1974. Calculus of Variations. Dover Publications.
Whitham, G.B. 1965. A general approach to linear and nonlinear dispersive waves us-
ing a Lagrangian. Journal of Fluid Mechanics 22:273-283.
Whitham, G.B. 1967. Variational methods and applications to water waves. Proceed-
ings of the Royal Society of London, Series A, 299:6-25.
Zakharov, V.E. 1968. Stability of periodic waves of nite amplitude on the surface of
a deep uid. Journal of Applied Mechanics and Technical Physics 2:190-194.
Zufiria, J.A. 1987a. Weakly nonlinear non-symmetric gravity waves on water of nite
depth. Journal of Fluid Mechanics 180:371-385.
Zufiria, J.A. 1987b. Non-symmetric gravity waves on water of innite depth. Journal
of Fluid Mechanics 181:17-39.
Zufiria, J.A. 1987c. Symmetry breaking in periodic and solitary gravity-capillary waves
on water of nite depth. Journal of Fluid Mechanics 184:183-206.
Zufiria, J.A. 1988. Oscillatory spatially periodic weakly nonlinear gravity waves on
deep water. Journal of Fluid Mechanics 191:341-372.
Zufiria, J.A., and Saffman, P.G. 1986a. An example of stability exchange in a
Hamiltonian wave system. Studies in Applied Mathematics 74:85-91.
Zufiria, J.A., and Saffman, P.G. 1986b. The superharmonic instability of nite-
amplitude surface waves on water of nite depth. Studies in Applied Mathematics 74:259-
8.5. BIBLIOGRAPHY 155
266.
Zwartkruis, T.J.G. 1991. Computation of solitary wave proles described by a Hamil-
tonian model for surface waves. Part 1: Final report; Part 2: Appendices. ECMI-report,
Eindhoven University of Technology, Eindhoven, The Netherlands.
156 CHAPTER 8. LAGRANGIAN AND HAMILTONIAN FORMULATIONS
Dawn was spreading over the earth when they reached the throng of heroes;
and the youths marvelled to behold the mighty eece, which gleamed like the
lightning of Zeus. And each one started up eager to touch it and clasp it in his
hands. But the son of Aeson restrained them all, and threw over it a mantle
newly-woven; and he led the maiden to the stern and seated her there, and
spake to them all as follows:
No longer now, my friends, forbear to return to your fatherland. For now
the task for which we dared this grievous voyage, toiling with bitter sorrow
of heart, has been lightly fullled by the maidens counsels. Her for such
is her will I will bring home to be my wedded wife; do ye preserve her,
the glorious saviour of all Achaea and of yourselves. For of a surety, I ween,
will Aeetes come with his host to bar our passage from the river into the sea.
But do some of you toil at the oars in turn, sitting man by man; and half
of you raise your shields of oxhide, a ready defence against the darts of the
enemy, and guard our return. And now in our hands we hold the fate of our
children and dear country and of our aged parents; and on our venture all
Hellas depends, to reap either the shame of failure or great renown.
Thus he spake, and donned his armour of war; and they cried aloud,
wondrously eager. And he drew his sword from the sheath and cut the hawsers
at the stern. And near the maiden he took his stand ready armed by the
steersman Ancaeus, and with their rowing the ship sped on as they strained
desperately to drive her clear of the river.
Argonautica, Book IV, Verses 183-211.
Chapter 9
Symmetries and Conservation
Laws
Thence they entered the deep stream of Rhodanus which ows into Eri-
danus; and where they meet there is a roar of mingling waters. Now that
river, rising from the ends of the earth, where are the portals and mansions
of Night, on one side bursts forth upon the beach of Ocean, at another pours
into the Ionian sea, and on the third through seven mouths sends its stream
to the Sardinian sea and its limitless bay
1
. And from Rhodanus they entered
stormy lakes, which spread throughout the Celtic mainland of wondrous size;
and there they would have met with an inglorious calamity; for a certain
branch of the river was bearing them towards a gulf of Ocean which in ig-
norance they were about to enter, and never would they have returned from
there in safety. But Hera leaping forth from heaven pealed her cry from the
Hercynian rock; and all together were shaken with fear of her cry; for ter-
ribly crashed the mighty rmament. And backward they turned by reason of
the goddess, and noted the path by which their return was ordained. And
after a long while they came to the beach of the surging sea by the devising
of Hera, passing unharmed through countless tribes of the Celts and Ligyans.
For round them the goddess poured a dread mist day by day as they fared on.
Argonautica, Book IV, Verses 627-648.
1
Apollonius seems to have thought that the Po, the Rhone, and the Rhine are all connected together.
157
158 CHAPTER 9. SYMMETRIES AND CONSERVATION LAWS
9.1 Introduction
This chapter is devoted to conservation laws for the problem of gravity driven water
waves and bodies oating freely in or below the free surface. Our main objective is to
nd constants of the motion for the nonlinear wave-body problem. The present treatment
is an elaborated and more detailed version of a paper written recently by van Daalen
and van Groesen (1993).
The rst systematic account of symmetries
2
and conservation laws for the water-wave
problem was given ten years ago in an excellent paper written by Benjamin and Olver
(1982). In their quest for constants of the motion, the Hamiltonian structure of wa-
ter waves
3
is exploited successfully; the techniques are based on some results of Olver
(1980b, 1986) for the possible sets of integral invariants of evolution equations under the
transformations of a Lie group of symmetries.
4
Many of the results thus obtained were already well-known; see, for instance, Lax
(1968), Miura, Gardner, and Kruskal (1968), Benjamin and Mahony (1971), Benjamin
(1974), Longuet-Higgins (1974, 1975, 1980), Longuet-Higgins and Fenton (1974), and
Broer, van Groesen and Timmers (1976). Considering free surface waves on a layer
of water of innite depth, Benjamin and Olver
5
proved the existence of exactly eight
conserved densities i.e. quantities that are not exact dierentials but whose integrals
over any horizontal domain depend on boundary values only for two-dimensional ow
6
without surface tension.
7
Some of these densities are related to well-known quantities,
for example the mass and energy; others correspond to less familiar quantities, such as
the angular momentum and another similar-looking quantity.
Because of the generality of Benjamin and Olvers analysis and the length of their
derivations, the particular results mentioned above were deduced in an alternative, more
simple, and direct way by Longuet-Higgins (1983) for two-dimensional, inviscid, irrota-
tional ow under gravity. The method employed was already referred to in Benjamin
and Olvers paper, but full credit must be given to Longuet-Higgins for presenting a
clear and simple proof of the existence of eight invariants for this particular problem. In
addition, most conservation laws were generalized to rotational motion or vorticity (i.e.
no potential) ow.
The primary aim of the present treatment is to generalize the aforementioned results
for water waves only to the more complex problem of wave-body interactions. In
particular, an attempt is made towards the extension of the water-wave conservation
laws to freely oating bodies in two-dimensional free surface potential ow.
For future reference, the fundamental results from Benjamin and Olvers analysis are
summarized in section 9.2. For the deeper question of completeness in the sense that
all basic conservation laws for the water-wave problem are covered indeed the reader
2
Throughout this chapter, the word symmetry is used to denote an invariance of a given system
under certain transformations. For example, if a system is completely self-contained, with only internal
forces between the particles, then the system can be moved as a rigid whole without aecting the forces
or subsequent motion; the system is said to be invariant or symmetric under a rigid displacement.
3
This special structure of water waves is discussed in chapter 8.
4
For a more recent application of these techniques to the Boussinesq model for wave motions we refer
to Benjamin (1986a).
5
From now on, Benjamin and Olver refers to their 1982-paper, unless stated otherwise.
6
For the three-dimensional case twelve conserved densities were found.
7
When surface tension is operative, the number of conserved densities is reduced by one.
9.2. CONSERVED DENSITIES FOR WATER WAVES 159
will of course have to study the arguments of Benjamin and Olver, who in turn refer to
a subsequent paper written by Olver (1983).
In section 9.3 Longuet-Higgins approach
8
is adopted in the treatment of the wave-
body problem; it appears that seven out of eight uid quantities have a direct counterpart
for a rigid body. In the absence of external forces and moments, one arrives at extended
conservation laws by a straightforward method; explicit proofs, based on the equations
of motion of the body and the boundary condition on the wetted body surface, are
presented.
The remaining eighth uid quantity and its body counterpart are extensively dealt
with in section 9.4. The main diculty in this discussion is the lack of an unambigu-
ous physical interpretation of this quantity, which seems to be related to the angular
momentum.
9
Here it is surmised (and argued) that the presence of a rigid body in ef-
fect breaks the underlying symmetry and, as a result, prohibits the generalization of the
corresponding conservation law to the wave-body problem.
The key assumption in Longuet-Higgins proofs of the conservation laws for water
waves is the vanishing of the Bernoulli pressure along the boundary of the uid domain.
This implies the supposition that the uid extends to innity in downward direction,
which is consistent with Benjamin and Olvers analysis for the innite depth case. For
practical purposes however, this is not a reasonable assumption; therefore the eects of
impermeable xed boundaries are discussed in section 9.5.
Finally, in section 9.6 we demonstrate that the validity of the theory is supported by
numerical results obtained with our panel method for nonlinear wave-body interactions.
The theoretical predictions with respect to the constants of the motion including the
symmetry-breaking eect of an impermeable bottom are conrmed by the computa-
tions for a special, yet simple, test conguration.
The reader who is primarily interested in practical applications is advised to skip sec-
tion 9.2 and proceed directly to section 9.3. Further, it is remarked that the approach in
section 9.4 is rather theoretical, contrary to the analyses in sections 9.5 and 9.6, where
matters of a more practical nature are discussed.
9.2 Conserved densities for water waves
In this section we present a summary of Benjamin and Olvers treatise on symmetries
and conservation laws for the water-wave problem. Only the two-dimensional innite
depth case is considered here; let x and z be rectangular coordinates in the plane of
motion, with z vertically upwards, see Figure 9.1.
As usual, let denote the velocity potential, which is introduced under the assumptions
of an ideal uid and an irrotational ow. The well-known governing equations for free
surface ow under the inuence of gravity read

2
= 0 for z < (x; t) , (9.1)

t
+
x

x
=
z
at z = (x; t) , (9.2)

t
+
1
2
( ) +gz = at z = (x; t) , (9.3)
|| 0 for r =

x
2
+z
2

1/2
, (9.4)
8
From now on, Longuet-Higgins refers to his 1983-paper, unless stated otherwise.
9
Benjamin and Olver refer to the corresponding density as the virial density.
160 CHAPTER 9. SYMMETRIES AND CONSERVATION LAWS
Figure 9.1: Two-dimensional water waves: the innite depth case.
where suxes denote partial dierentiation. It is assumed that the free surface elevation
is a single-valued function of x for all times t. Surface tension is included in the dynamic
free surface condition (9.3) through , the coecient of surface tension, and the surface
curvature , given by
=

x


x
(1 +
2
x
)
1/2

=

xx
(1 +
2
x
)
3/2
. (9.5)
The following two theorems due to Benjamin and Olver state that the full symmetry
group for the two-dimensional water-wave problem is generated by nine one-parameter
subgroups. This means that any symmetry with a continuous connection to the identity
map can be constructed by applying a limited number of the component symmetries in
succession.
10
In the rst theorem the full symmetry group or the Lie algebra of symmetries is
dened as the linear space spanned by nine vector elds; each vector eld may be regarded
as a rst order dierential operator acting on smooth functions f (x, z, t, ) : IR
4
IR
4
.
Theorem 5 : Symmetry-generating vector elds
(Benjamin and Olver, 1982)
The Lie algebra of innitesimal symmetries for the two-dimensional water-wave problem
in the absence of surface tension i.e. (9.1-9.4) with = 0 is spanned by the
following nine vector elds:
v
1
=
x
, (9.6)
v
2
=
t
, (9.7)
v
3
=

, (9.8)
10
This may be compared to the group of rotations in IR
3
being generated by the three one-parameter
subgroups of rotations about three cartesian axes. Similarly, the group of translations in IR
3
is generated
by the three one-parameter subgroups of translations along three cartesian axes.
9.2. CONSERVED DENSITIES FOR WATER WAVES 161
v
4
=
z
gt

, (9.9)
v
5
= t
x
+x

, (9.10)
v
6
= t
z
+

z
1
2
gt
2

, (9.11)
v
7
= gt
2

z
t
t
+

+ 2gtz
1
3
g
2
t
3

, (9.12)
v
8
=

z +
1
2
gt
2

x
x
z
+gtx

, (9.13)
v
9
= x
x
+z
z
+
1
2
t
t
+
3
2

. (9.14)
Each of the symmetries generated by these vector elds is unaected by the additional sur-
face tension term in the dynamic free surface condition (9.3); hence, no extra symmetry
is induced when surface tension is operative.
The above vector elds are the innitesimal generators of one-parameter groups of dif-
feomorphisms, i.e. continuous maps G
i
: IR
4
IR
4
transforming solutions (x, z, t, )
into new solutions to the water-wave problem (9.1-9.4). The dieomorphisms are obtain-
able by integrating a system of ordinary dierential equations involving the dependent
variables x, z, t, and , and their derivatives with respect to the group parameters.
An alternative method introduced by Benjamin and Olver to establish symmetry
groups for the present problem is based on the so-called prolongation theory; each vector
eld v
j
generates a one-parameter symmetry group G
j
, as stated in the next theorem.
Theorem 6 : Lie algebra of symmetries (Benjamin and Olver, 1982)
The full symmetry group for the two-dimensional water-wave problem in the absence of
surface tension i.e. (9.1-9.4) with = 0 is generated by the following nine one-
parameter subgroups:
Horizontal translation:
G
1
: (x +
1
, z, t, ) , (9.15)
Time translation:
G
2
: (x, z, t +
2
, ) , (9.16)
Variation of base-level for potential:
G
3
: (x, z, t, +
3
) , (9.17)
Vertical translation:
G
4
: (x, z +
4
, t,
4
gt) , (9.18)
Horizontal Galilean boost:
G
5
:

x +
5
t, z, t, +
5
x +
1
2

2
5
t

, (9.19)
162 CHAPTER 9. SYMMETRIES AND CONSERVATION LAWS
Vertical Galilean boost:
G
6
:

x, z +
6
t, t, +
6

z
1
2
gt
2

+
1
2

2
6
t

, (9.20)
Vertical acceleration:
G
7
:

x, z +
1
2
gt
2

1
2
7

,
1
7
t,

+gtz

1
2
7

+
1
6
g
2
t
3

1 3
2
7
+ 2
4
7

, (9.21)
Gravity-compensated rotation:
G
8
:

xcos
8
+

z +
1
2
gt
2

sin
8
,
xsin
8
+

z +
1
2
gt
2

cos
8

1
2
gt
2
, t,
+gt

xsin
8
+

z +
1
2
gt
2

(1 cos
8
)

, (9.22)
Scaling:
G
9
:

9
x,
9
z,
1/2
9
t,
3/2
9

. (9.23)
When surface tension is operative ( > 0), all the above groups except G
7
and G
9
remain
symmetries; in this case, the following combination of G
7
and G
9
in eect,

G
7
=
G
9
G
7
with
7
=

1
7
=
1
9
remains a symmetry:
Scaled acceleration:

G
7
:

7
x,

z +
1
2
gt
2

2
7

3/2
7
t,

1/2
7

+gtz

2
7

+
1
6
g
2
t
3

1 3

2
7
+ 2

4
7

. (9.24)
In the above expressions, each
i
IR denotes an additive group parameter, and each

j
= e

j
> 0 denotes a multiplicative group parameter.
The given subgroups may be considered to act geometrically on the four-dimensional
space with coordinates (x, z, t, ) for the two-dimensional problem. Each subgroup in-
duces a transformation in the space of solutions, in eect transforming the graphs of
the free surface and the velocity potential. The key point is that transforming a given
solution by any of the symmetries produces a continuous family of other solutions. For
instance, the subgroup G
1
induces a horizontal translation of the water-wave problem:
(x; t) , (x, z; t) (x
1
; t) , (x
1
, z; t) . (9.25)
Similarly, G
2
represents a time shift t t
2
; G
3
has a bearing on the fact that the
potential is determined up to an additive constant; G
4
induces a vertical translation of
the water-wave problem. The Galilean boosts G
5
and G
6
represent the eects caused by
frames of reference moving at a constant speed in the horizontal and vertical direction
respectively. The group G
7
represents the eects due to a frame that is accelerating
9.2. CONSERVED DENSITIES FOR WATER WAVES 163
uniformly in vertical direction, thus modifying the eective gravity constant.
11
. To un-
derstand G
8
, consider the special case where g = 0; since there is no preferred direction
then, any solution will remain a solution after being rotated about an arbitrary point in
the (x, z)-plane. Finally, the transformations in the scaling group G
9
are evident from
dimensional considerations.
Noethers theorem
12
(1918) states that every one-parameter group of symmetries for a
variational problem
13
determines a conservation law satised by solutions of the corre-
sponding Euler-Lagrange equations; a more comprehensive description of this fundamen-
tal result can be found in, for instance, Goldstein (1980). This principle, in an adapted
form for free-boundary problems due to Olver (1980a), was used by Benjamin and Olver
to obtain the conserved densities listed in the following theorem.
Theorem 7 : Conserved densities (Benjamin and Olver, 1982)
The two-dimensional water-wave problem in the absence of surface tension i.e. (9.1-
9.4) with = 0 has the following eight conserved densities:
T
1
=
x
, (9.26)
T
2
= H =
1
2

t
+
1
2
g
2
, (9.27)
T
3
= , (9.28)
T
4
= +gtT
3
, (9.29)
T
5
= x tT
1
, (9.30)
T
6
=
1
2

2
tT
4
+
1
2
gt
2
T
3
, (9.31)
T
7
= ( x
x
) t (4T
2
7gT
6
) +
7
2
gt
2
T
4

7
6
g
2
t
3
T
3
, (9.32)
T
8
= (x +
x
) +gtT
5
+
1
2
gt
2
T
1
, (9.33)
where (x; t) (x, z = (x; t) ; t) is the restriction of the velocity potential to the free
surface, and H is the Hamiltonian density.
Each density T
i
is generated by the corresponding vector eld v
i
, apart from T
7
, which
is induced by v
9
(7/2) v
7
. When surface tension is operative, i.e. > 0, then v
9
is no
longer a symmetry, and consequently the density T
7
is no longer conserved. Furthermore,
the appropriate term is then added to the Hamiltonian density:
T
2
= H =
1
2

t
+
1
2
g
2
+

1 +
2
x

1/2
1

. (9.34)
Apart from these two changes, the results remain as above.
The rst density is easily associated with the horizontal momentum, by noting that
+

T
1
dx = []
+

+
+

x
dx =
+

x
dx , (9.35)
11
Hence, for the appropriate choice of
7
, the transformed graphs of (x; t) and (x, z; t) show the
evolution of (ctitious) water waves on the moon!
12
Noether, Emmy (1882-1935), German mathematician. She was one of the leading mathematicians
of this century, and has been properly described as the greatest of women mathematicians.
13
Note that both the water-wave problem and the wave-body problem can be described by a variational
principle, see chapter 8.
164 CHAPTER 9. SYMMETRIES AND CONSERVATION LAWS
where it is assumed that the free surface elevation vanishes for [x[ .
Clearly, T
2
corresponds to the energy, and T
3
is a density of mass. Further physical
connotations due to Benjamin and Olver are as follows: T
4
is to be interpreted as the
vertical momentum density; T
5
and T
6
correspond to the horizontal and vertical position
of the mass centroid; T
8
is an angular momentum density, and T
7
is a related density
corresponding to an unfamiliar quantity, also named a virial density.
9.3 Conservation laws for the wave-body problem
The above results for water waves were deduced in a more direct and simple way by
Longuet-Higgins; in the next subsections his approach is followed to construct conserva-
tion laws for the more complex wave-body problem.
9.3.1 Invariants for the two-dimensional case
The notation used here diers but slightly from the notation used by Longuet-Higgins.
The subscripts f, b, and s refer to the uid, the oating body, and the uid-body system
14
respectively, see Figure 9.2.
Figure 9.2: Two-dimensional wave-body system.
As in the previous section, let x and z be rectangular coordinates in the plane of motion,
with z vertically upwards, and let denote the velocity potential, which is introduced
under the assumptions of an ideal uid and an irrotational ow. If C denotes any simple
closed contour bounding a domain D
f
of the uid see Figure 9.2 then, using Greens
14
From now on, the system consisting of the uid and the oating body is referred to as the wave-body
system or (simply) the system.
9.3. CONSERVATION LAWS FOR THE WAVE-BODY PROBLEM 165
theorem
15

D
f

(P
z
Q
x
) dxdz =

C
(P dx +Qdz) , (9.36)
where P
z
= P/z and Q
x
= Q/x, Longuet-Higgins dened the following eight inte-
gral quantities for the uid:

f
=

D
f

dxdz =

C
z dx , (9.37)

f
x
f
=

D
f

xdxdz =

C
xz dx , (9.38)

f
z
f
=

D
f

z dxdz =

C
1
2
z
2
dx , (9.39)
1
f
=

D
f


x
dxdz =

C
dz , (9.40)
.
f
=

D
f


z
dxdz =

C
dx , (9.41)
/
f
=

D
f

[(x)
z
(z)
x
] dxdz =

C
(xdx +z dz) , (9.42)
B
f
=

D
f

[(x)
x
+ (z)
z
] dxdz =

C
(z dx xdz) , (9.43)
H
f
= /
f
+{
f
, (9.44)
with /
f
and {
f
dened as
/
f
=

D
f

1
2

2
x
+
2
z

dxdz =

C
1
2
(
z
dx
x
dz) , (9.45)
{
f
=

D
f

gz dxdz =

C
1
2
gz
2
dx =
f
g z
f
. (9.46)
Thus,
f
is familiar as the uid mass (the uid density
f
being taken as unity); x
f
and z
f
are the coordinates of the centre of uid mass; 1
f
and .
f
are the two components
of the uid momentum; /
f
is the angular uid momentum; B
f
an unfamiliar uid
quantity, looking similar to /
f
; and H
f
is the uid energy, being the sum of the kinetic
uid energy /
f
and the potential uid energy {
f
. In (9.45) Laplaces equation for the
velocity potential is used, by continuity.
15
Green, George (1793-1841), British mathematician. In 1828 he published for private circulation
an essay on electricity and magnetism that contained the important theorem bearing his name. This
theorem, or its analogue in three dimensions, is also known as the divergence theorem or Gauss theorem,
for Greens results were largely overlooked until rediscovered by Lord Kelvin in 1846. The theorem
meanwhile had been discovered by Michel Ostrogradski (1801-1861), and in Russia it bears his name to
this day.
166 CHAPTER 9. SYMMETRIES AND CONSERVATION LAWS
Most of the above uid quantities have a direct analogue for a rigid body oating in or
below the free surface. The body mass and moment of inertia about its centre of mass
or centre of gravity ( x
b
, z
b
) are denoted by
b
and ^
b
respectively. The body angle
of roll is denoted by

b
, the bar indicating that the rolling motion is considered with
respect to the centre of gravity (see Figure 9.2). The wetted part of the body surface is
denoted by S.
With these denitions, the following quantities for the body are dened:
1
b
=
b

x
b
, (9.47)
.
b
=
b

z
b
, (9.48)
/
b
= ^
b

b
+
b
( x
b

z
b
z
b

x
b
) , (9.49)
H
b
= /
b
+{
b
, (9.50)
with /
b
and {
b
dened as:
/
b
=
1
2

x
2
b
+

z
2
b

+
1
2
^
b

2
b
, (9.51)
{
b
=
b
g z
b
. (9.52)
Thus, 1
b
and .
b
are the two components of the body momentum; /
b
is the angular body
momentum with respect to the origin in appendix C we derive expression (9.49); and
H
b
is the body energy, being the sum of the kinetic body energy /
b
and the potential
body energy {
b
.
With the uid quantities (9.37-9.42) and (9.44-9.46), and with the body quantities (9.47-
9.52), the quantities for the wave-body system are dened as

s
=
f
+
b
, (9.53)

s
x
s
=
f
x
f
+
b
x
b
, (9.54)

s
z
s
=
f
z
f
+
b
z
b
, (9.55)
1
s
= 1
f
+1
b
, (9.56)
.
s
= .
f
+.
b
, (9.57)
/
s
= /
f
+/
b
, (9.58)
H
s
= H
f
+H
b
. (9.59)
Thus,
s
is the system mass; x
s
and z
s
are the coordinates of the centre of system
mass; 1
s
and .
s
are the two components of the system momentum; /
s
is the angular
system momentum; and H
s
is the system energy, being the sum of the kinetic system
energy /
s
= /
f
+/
b
and the potential system energy {
s
= {
f
+{
b
. The discussion of
B
b
and B
s
(the analogues of B
f
for the body and the system respectively) is deferred to
section 9.4.
Let p denote the pressure given by Bernoullis equation
p +
t
+
1
2

2
x
+
2
z

+gz = 0 . (9.60)
9.3. CONSERVATION LAWS FOR THE WAVE-BODY PROBLEM 167
Then, assuming that the contour C moves with the uid, Longuet-Higgins deduced the
following identities:
d
f
dt
= 0 , (9.61)

f
d x
f
dt
= 1
f
, (9.62)

f
d z
f
dt
= .
f
, (9.63)
d1
f
dt
=

C
p dz , (9.64)
d.
f
dt
=

C
p dx
f
g , (9.65)
d/
f
dt
=

C
p (xdx +z dz)
f
g x
f
, (9.66)
dH
f
dt
=

C
p (
z
dx
x
dz) . (9.67)
Proof: Note that since C moves with the uid,
16
which is incompressible, we have in
general
d
dt

D
f

f dxdz =

D
f

D
Dt
[f] dxdz , (9.68)
where D/Dt denotes material dierentiation i.e. following the motion and f is a
function of position and time dened in D
f
.
The proof of (9.61-9.63) follows directly from
D
Dt
[1] = 0 , (9.69)
D
Dt
[x] =
x
, (9.70)
D
Dt
[z] =
z
. (9.71)
To prove (9.64) and (9.65), note that from (9.60)
D
Dt
[
x
] = p
x
, (9.72)
D
Dt
[
z
] = p
z
g . (9.73)
To prove (9.66), note that from (9.60)
D
Dt
[] =
t
+
2
x
+
2
z
= (p +gz) +
1
2

2
x
+
2
z

. (9.74)
16
This means that C is a Lagrangian boundary; its transient shape is determined by the trajectories
of the outmost uid particles.
168 CHAPTER 9. SYMMETRIES AND CONSERVATION LAWS
Hence, using (9.70-9.73), it is found that
D
Dt
[(x)
z
(z)
x
] =
D
Dt
(x
z
z
x
)
=

Dx
Dt

Dz
Dt

x
D
z
Dt
z
D
x
Dt

= (
x

x
) x(p
z
+g) +zp
x
= (xp)
z
+ (zp)
x
gx . (9.75)
Lastly, to prove (9.67), note that since
2
= 0
D
Dt

1
2

2
x
+
2
z

+gz

x
D
x
Dt
+
z
D
z
Dt

+g
Dz
Dt
=
x
p
x

z
(p
z
+g) +g
z
= (p
x
)
x
(p
z
)
z
. (9.76)
With (9.47-9.59) and (9.61-9.67), the following identities for the wave-body system can
be proven:
d
s
dt
= 0 , (9.77)

s
d x
s
dt
= 1
s
, (9.78)

s
d z
s
dt
= .
s
, (9.79)
d1
s
dt
=

C
p dz +
b

x
b
, (9.80)
d.
s
dt
=

C
p dx
f
g +
b

z
b
, (9.81)
d/
s
dt
=

C
p (xdx +z dz)
f
g x
f
+^
b

b
+
b
( x
b

z
b
z
b

x
b
) , (9.82)
dH
s
dt
=

C
p (
z
dx
x
dz)
+
b
(

x
b

x
b
+

z
b

z
b
) +^
b

b
+
b
g

z
b
. (9.83)
Proof: It is clear that (9.77) follows from (9.53), (9.61), and the fact that
b
is a
constant. Identity (9.78) can be deduced from (9.54), (9.56), (9.47), and (9.62). Sim-
ilarly, (9.79) is obtained from (9.55), (9.57), (9.48), and (9.63). Identity (9.80) can be
deduced from (9.56), (9.47), and (9.64). Similarly, (9.81) is obtained from (9.57), (9.48),
and (9.65). Identity (9.82) follows from (9.58), (9.49), and (9.66). Finally, (9.83) is
deduced from (9.59), (9.50-9.52), and (9.67).
9.3. CONSERVATION LAWS FOR THE WAVE-BODY PROBLEM 169
To arrive at conservation laws for the wave-body system, the dynamical interaction of
the uid and the body has to be stated mathematically. To be precise, an extra set
of equations is needed, specifying the mutual exchange of momentum and energy, and
describing the uid ow on the wetted body surface.
In the absence of external forces, the only forces acting on the body are the pressure
forces due to the hydrodynamic interaction with the uid. Under these circumstances
the equations of motion for the body read

x
b
=

S
p n
x
ds =

S
p dz , (9.84)

z
b
=

S
p n
z
ds
b
g =

S
p dx
b
g , (9.85)
^
b

b
=

S
p (r
x
n
z
r
z
n
x
) ds =

S
p (r
x
dx +r
z
dz) , (9.86)
where n = (n
x
, n
z
)
T
is the unit normal vector along the wetted body surface S; along S
we have
dx = n
z
ds , dz = n
x
ds . (9.87)
In (9.86) the vector r = (r
x
, r
z
)
T
denotes the position of a point (x, z) on S relative to
the centre of gravity ( x
b
, z
b
), that is
r
x
= x x
b
, r
z
= z z
b
. (9.88)
The equations of motion for the body, connecting the hydrodynamic pressure forces and
moment with the translational and rotational accelerations, are supplemented with a
contact condition on the wetted body surface:

n
= n = u
b
n on S , (9.89)
where u
b
= (u
b
, w
b
)
T
denotes the velocity of a point (x, z) on S:
u
b
=

x
b

b
(z z
b
) =

x
b

b
r
z
, (9.90)
w
b
=

z
b
+

b
(x x
b
) =

z
b
+

b
r
x
. (9.91)
The Neumann boundary condition (9.89) states that the normal uid velocity and the
normal body surface velocity coincide for all time, thus excluding the occurrence of cav-
ities on the body surface.
With the identities (9.77-9.83), the equations of motion (9.84-9.86), the boundary condi-
tion (9.89), and the vanishing of p on C ` S, seven conservation laws for the wave-body
system can be deduced; they are listed in the following theorem.
Theorem 8 : Conservation laws for the wave-body problem (2D)
The two-dimensional wave-body problem in the absence of external forces has the following
170 CHAPTER 9. SYMMETRIES AND CONSERVATION LAWS
seven constants of the motion:
c
1
=
s
, (9.92)
c
2
= 1
s
, (9.93)
c
3
= .
s
+c
1
gt = .
s
+
s
gt , (9.94)
c
4
=
s
x
s
c
2
t =
s
x
s
1
s
t , (9.95)
c
5
=
s
z
s
c
3
t +
1
2
c
1
gt
2
=
s

z
s

1
2
gt
2

.
s
t , (9.96)
c
6
= H
s
, (9.97)
c
7
= /
s
+g

c
4
t +
1
2
c
2
t
2

= /
s
+gt

s
x
s

1
2
1
s
t

. (9.98)
In the special case where the eect of gravity vanishes (i.e. g = 0), these quantities reduce
to
s
, 1
s
, .
s
, (
s
x
s
1
s
t), (
s
z
s
.
s
t), H
s
, and /
s
respectively.
Proof: Conservation law (9.92) follows directly from (9.77), expressing that
The total mass is a constant of the motion.
If the pressure p vanishes everywhere on C ` S, identity (9.80) yields
d1
s
dt
=

S
p dz +
b

x
b
. (9.99)
It is then obvious that (9.93) follows from (9.84), thus stating that
The total horizontal momentum is a constant of the motion.
Similarly, with the vanishing of p on C ` S, (9.81) gives
d.
s
dt
=

S
p dx
f
g +
b

z
b
, (9.100)
and hence, using (9.85)
d.
s
dt
= (
f
+
b
) g =
s
g . (9.101)
This leads to (9.94), expressing that in the absence of gravity
The total vertical momentum is a constant of the motion.
Next, (9.95) is obtained from (9.78) and (9.93), stating that
The horizontal velocity of the centre of mass
is a constant of the motion.
Similarly, (9.96) is derived from (9.79) and (9.94), expressing that in the absence of
gravity
The vertical velocity of the centre of mass
is a constant of the motion.
9.3. CONSERVATION LAWS FOR THE WAVE-BODY PROBLEM 171
In order to prove (9.97), note that with p vanishing on C ` S, (9.83) yields
dH
s
dt
=

S
p (
x
n
x
+
z
n
z
) ds
+
b
(

x
b

x
b
+

z
b

z
b
) +^
b

b
+
b
g

z
b
. (9.102)
With the Neumann condition (9.89), this can be rewritten to
dH
s
dt
=

x
b

x
b
+

S
p dz

+

z
b

z
b

S
p dx +
b
g

^
b

S
p (r
x
dx +r
z
dz)

. (9.103)
Substitution of the equations of motion (9.84-9.86) then yields (9.97), stating that
The total energy is a constant of the motion.
Finally, with p non-zero only on S, (9.82) can be written as
d/
s
dt
= ^
b

S
p (r
x
dx r
z
dz) M
f
g x
f
+ x
b

z
b

S
p dx

z
b

x
b
+

S
p dz

. (9.104)
Substitution of (9.84-9.86) then yields
d/
s
dt
= M
f
g x
f

b
g x
b
=
s
g x
s
, (9.105)
which completes the proof of (9.98), expressing that in the absence of gravity
The total angular momentum is a constant of the motion.
This completes the proof of theorem 8.
9.3.2 Invariants for the three-dimensional case
The generalization of the above results to the three-dimensional case is straightforward
and based on Gauss divergence theorem

V
f



U dV =

S
f


U ndS , (9.106)
where the vector eld

U is dened on the uid domain V
f
, bounded by a closed surface
S
f
.
172 CHAPTER 9. SYMMETRIES AND CONSERVATION LAWS
In this case we have twelve quantities for the uid, and each quantity is dened as a
volume integral of the corresponding density:

f
=

V
f

dV , (9.107)

f
x
f
=

V
f

xdV , (9.108)

f
y
f
=

V
f

y dV , (9.109)

f
z
f
=

V
f

z dV , (9.110)
1
x
f
=

V
f


x
dV =

S
f

n
x
dS , (9.111)
1
y
f
=

V
f


y
dV =

S
f

n
y
dS , (9.112)
1
z
f
=

V
f


z
dV =

S
f

n
z
dS , (9.113)
/
x
f
=

V
f


(z)
y
(y)
z

dV =

S
f

(zn
y
yn
z
) dS , (9.114)
/
y
f
=

V
f

[(x)
z
(z)
x
] dV =

S
f

(xn
z
zn
x
) dS , (9.115)
/
z
f
=

V
f


(y)
x
(x)
y

dV =

S
f

(yn
x
xn
y
) dS , (9.116)
B
f
=

V
f


(x)
x
+ (y)
y
+ (z)
z

dV
=

S
f

(xn
x
+yn
y
+zn
z
) dS , (9.117)
H
f
= /
f
+{
f
, (9.118)
with /
f
and {
f
dened as
/
f
=

V
f

1
2

2
x
+
2
y
+
2
z

dV
=

S
f

1
2
(
x
n
x
+
y
n
y
+
z
n
z
) dS , (9.119)
9.3. CONSERVATION LAWS FOR THE WAVE-BODY PROBLEM 173
{
f
=

V
f

gz dV =
f
g z
f
. (9.120)
Thus,
f
is familiar as the uid mass; x
f
, y
f
, and z
f
are the coordinates of the centre
of uid mass (x and y are the horizontal directions); 1
x
f
, 1
y
f
, and 1
z
f
are the three
components of the uid momentum; /
x
f
, /
y
f
, and /
z
f
are the three components of the
angular uid momentum; B
f
is the unfamiliar uid quantity; and H
f
is the uid energy,
being the sum of the kinetic uid energy /
f
and the potential uid energy {
f
.
The analogues for a rigid body characterized by its mass
b
, its moment of inertia

^
b
= (^
x
b
, ^
y
b
, ^
z
b
), its position x
b
= ( x
b
, y
b
, z
b
), and its orientation

b
=

x
b
,

y
b
,

z
b


read
1
x
b
=
b

x
b
, (9.121)
1
y
b
=
b

y
b
, (9.122)
1
z
b
=
b

z
b
, (9.123)
/
x
b
= ^
x
b

x
b
+
b
( z
b

y
b
y
b

z
b
) , (9.124)
/
y
b
= ^
y
b

y
b
+
b
( x
b

z
b
z
b

x
b
) , (9.125)
/
z
b
= ^
z
b

z
b
+
b
( y
b

x
b
x
b

y
b
) , (9.126)
H
b
= /
b
+{
b
, (9.127)
where /
b
and {
b
are dened as
/
b
=
1
2

x
b


x
b

+
1
2

^
b

b
, (9.128)
{
b
=
b
g z
b
. (9.129)
Thus, 1
x
b
, 1
y
b
, and 1
z
b
are the three components of the horizontal body momentum; /
x
b
,
/
y
b
, and /
z
b
are the three components of the angular body momentum; and H
b
is the
body energy, being the sum of the kinetic body energy /
b
and the potential body energy
{
b
.
Next, we put the system quantities as

s
=
f
+
b
, (9.130)

s
x
s
=
f
x
f
+
b
x
b
, (9.131)

s
y
s
=
f
y
f
+
b
y
b
, (9.132)

s
z
s
=
f
z
f
+
b
z
b
, (9.133)
1
x
s
= 1
x
f
+1
x
b
, 1
y
s
= 1
y
f
+1
y
b
, 1
z
s
= 1
z
f
+1
z
b
, (9.134)
/
x
s
= /
x
f
+/
x
b
, /
y
s
= /
y
f
+/
y
b
, /
z
s
= /
z
f
+/
z
b
, (9.135)
H
s
= H
f
+H
b
, /
s
= /
f
+/
b
, {
s
= {
f
+{
b
. (9.136)
Then, with the equations of motion for the unrestrained body

x
b
=

p n
x
dS , (9.137)

y
b
=

p n
y
dS , (9.138)
174 CHAPTER 9. SYMMETRIES AND CONSERVATION LAWS

z
b
=

p n
z
dS
b
g , (9.139)
^
x
b

x
b
=

p (r
y
n
z
r
z
n
y
) dS , (9.140)
^
y
b

y
b
=

p (r
z
n
x
r
x
n
z
) dS , (9.141)
^
z
b

z
b
=

p (r
x
n
y
r
y
n
x
) dS , (9.142)
with the contact condition on the wetted body surface
n = u
b
n on S with u
b
= (u
b
, v
b
, w
b
)
T
, (9.143)
and with the vanishing of the Bernoulli pressure
p =

t
+
1
2

2
x
+
2
y
+
2
z

+gz

(9.144)
on S
f
` S, we arrive at the conservation laws listed in the following theorem.
Theorem 9 : Conservation laws for the wave-body problem (3D)
The three-dimensional wave-body problem in the absence of external forces has the fol-
lowing eleven constants of the motion:
c
1
=
s
, (9.145)
c
2
= 1
x
s
, (9.146)
c
3
= 1
y
s
, (9.147)
c
4
= 1
z
s
+
s
gt , (9.148)
c
5
=
s
x
s
1
x
s
t , (9.149)
c
6
=
s
y
s
1
y
s
t , (9.150)
c
7
=
s

z
s

1
2
gt
2

1
z
s
t , (9.151)
c
8
= H
s
, (9.152)
c
9
= /
x
s
+gt

s
y
s

1
2
1
y
s
t

, (9.153)
c
10
= /
y
s
+gt

s
x
s

1
2
1
x
s
t

, (9.154)
c
11
= /
z
s
. (9.155)
In the special case where g = 0, these quantities reduce to
s
, 1
x
s
, 1
y
s
, 1
z
s
, (
s
x
s
1
x
s
t),
(
s
y
s
1
y
s
t), (
s
z
s
1
z
s
t), H
s
, /
x
s
, /
y
s
, and /
z
s
respectively.
9.4. THE VIRIAL AND CONSERVATION LAW NO.8 175
9.4 The virial and conservation law no.8
Attention is now focused on the unfamiliar uid quantity B
f
, dened in (9.43) for the
two-dimensional case and here repeated for convenience:
B
f
=

D
f

[(x)
x
+ (z)
z
] dxdz =

C
(z dx xdz) . (9.156)
For reasons that will be given later on, this quantity will be named virial hereafter. The
virial bears a strong resemblance to the angular uid momentum/
f
, see (9.42). However,
it will appear later on that the existence of the velocity potential is a prerequisite for
the denition of B
f
, while /
f
may as well be dened in terms of the velocity eld
(u
f
, w
f
) of the uid ow (see also Appendix C). This dierence appears to be the major
obstacle in the generalization of B
f
for the wave-body problem. An extensive discussion
from various points of view of the virial, its body and system analogues, and
the corresponding conservation law (which is lacking so far) is given in the following
subsections.
9.4.1 The virial connection
The fact that potential ow is a necessary premise for the present denition of B
f
becomes
more clear when (9.156) is rewritten to
B
f
=

D
f

(x
x
+z
z
) dxdz +

D
f

2dxdz (
f
+{
f
. (9.157)
The rst integral can also be written in terms of the velocity eld (u
f
, w
f
)
T
:
(
f
=

D
f


f
(xu
f
+zw
f
) dxdz , (9.158)
where the uid density
f
, which has been taken as unity before, has been reintroduced.
Thus, (
f
can be dened as the integral of the inner product of the position vector (x, z)
T
and the momentum vector (
f
u
f
,
f
w
f
)
T
, where the integration is carried out over the
uid domain D
f
. Dened in this way, (
f
is known to play an important role in the
so-called virial theorem.
The virial theorem is statistical in nature; for a large variety of systems it is concerned
with the time averages of various mechanical quantities. A brief discussion of the virial
theorem for a discrete system of mass points has been given by Goldstein (1980); equiv-
alent theorems for the continuous water-wave and wave-body problems will be deduced
hereafter.
17
Consider a system of free surface waves, where each innitesimal uid particle is char-
acterized by its transient position r
f
= (x, z)
T
and its momentum p
f
=
f

r
f
where
the latter vector quantity should not be confused with the Bernoulli pressure p. With a
17
For a discussion of the relation between uid mechanics and statistical physics, see Uhlenbecks
review article (1980).
176 CHAPTER 9. SYMMETRIES AND CONSERVATION LAWS
continuous force distribution

F
f
throughout the uid domain, the fundamental equations
of motion read

p
f
=

F
f
. (9.159)
Substitution of the inertial (pressure) forces and the gravitational forces for

F
f
yields the
well-known Euler equations for incompressible and inviscid uid ow.
With the above denitions, (9.158) is rewritten to
(
f
=

D
f

(r
f
p
f
) dxdz . (9.160)
The total time derivative of this quantity is
d(
f
dt
=

D
f

r
f
p
f

dxdz +

D
f

p
f
r
f

dxdz . (9.161)
The rst term can be transformed to

D
f

r
f
p
f

dxdz =

D
f

r
f


r
f

dxdz = 2 /
f
, (9.162)
while the second by (9.159) is

D
f

p
f
r
f

dxdz =

D
f

F
f
r
f

dxdz . (9.163)
Identity (9.161) therefore reduces to
d(
f
dt
= 2 /
f
+

D
f

F
f
r
f

dxdz . (9.164)
Averaging this result over a time interval [0, T] gives
1
T
T

0
d(
f
dt
dt =<
d(
f
dt
>=< 2 /
f
> + <

D
f

F
f
r
f

dxdz > , (9.165)


or
2 < /
f
> + <

D
f

F
f
r
f

dxdz >=
1
T
[(
f
(T) (
f
(0)] . (9.166)
If the motion is periodic, i.e. all particle positions repeat after a certain time, and if T
is chosen to be the period, then the right-hand side of (9.166) vanishes.
18
It then follows
that
< /
f
>=
1
2
<

D
f

F
f
r
f

dxdz > . (9.167)


18
A similar conclusion can be reached even if the motion is not periodic, provided that the coordinates
and velocities of all uid particles remain nite so that there is an upper bound to G
f
. Then, by choosing
T suciently long, the right-hand side of (9.166) can be made as small as desired.
9.4. THE VIRIAL AND CONSERVATION LAW NO.8 177
Identity (9.167) is known as the virial theorem
19
, and the right-hand side is called the
virial of Clausius
20
.
The above reasoning can also be applied to the second integral in (9.157):
{
f
=

D
f

2dxdz . (9.168)
Note that, since the potential eld is determinate up to an additive constant, we are
allowed to choose the base-level of at a certain time t
0
such that {
f
(t
0
) = 0. With
this special choice it follows that B
f
(t
0
) = (
f
(t
0
), which is the justication for labelling
B
f
as the virial quantity.
The total time derivative of {
f
is
d{
f
dt
=

D
f

2
D
Dt
dxdz = 2 /
f
2 {
f

D
f

2p dxdz , (9.169)
where we used the Bernoulli pressure denition (9.60).
Averaging over a time interval [0, T] gives
2 < /
f
> 2 < {
f
> <

D
f

2p dxdz >=
1
T
[{
f
(T) {
f
(0)] . (9.170)
Substitution of the inertial and gravitational forces

F
f
= (p
x
, p
z
g)
T
(9.171)
into (9.166) yields the identity
2 < /
f
> < {
f
> <

D
f

(xp
x
+zp
z
) dxdz >
=
1
T
[(
f
(T) (
f
(0)] . (9.172)
By simple addition of (9.170) and (9.172), we obtain with (9.157)
4 < /
f
> 3 < {
f
> <

D
f

[(xp)
x
+ (zp)
z
] dxdz >
=
1
T
[B
f
(T) B
f
(0)] . (9.173)
Now, if the potential ow is periodic, and if T is chosen to be the period of the uid
motion, then the right-hand side vanishes, and we arrive at
4 < /
f
> 3 < {
f
>=<

C
f
p (xn
x
+zn
z
) ds > , (9.174)
19
The discrete variant of this result is very useful in the kinetic theory of gases, in particular in the
proof of Boyles law PV = NkT.
20
Clausius, Rudolf Julius Emmanuel (1822-1888), German physicist and mathematician who made
important contributions to thermodynamics and the kinetic theory of gases.
178 CHAPTER 9. SYMMETRIES AND CONSERVATION LAWS
where Greens divergence theorem (9.36) has been applied.
Finally, with the vanishing of p on C
f
, we obtain the statistical result
< /
f
>
< {
f
>
=
3
4
. (9.175)
This special partition of the averaged kinetic and potential energies may be compared to
a similar result for discrete systems, wherein the forces are derivable from a potential; if
the potential behaves like r
n+1
, then the force law goes as r
n
. Application of the virial
theorem then leads to the well-known partition < K >: < P >= (n + 1) : 2.
The extension of the virial theorem to the more complex wave-body problem is rather
straightforward; the analogue of (
f
for a oating body is
(
b
=

D
b


b
(xu
b
+zw
b
) dxdz =
b
( x
b

x
b
+ z
b

z
b
) , (9.176)
where
b
is the body density and (9.90-9.91) have been used. Clearly, (
b
accounts for
the translational motions of the body only, irrespective of possible rotational motion.
However, the equivalent of {
f
for the body may be put as
{
b
= ^
b

b
. (9.177)
Clearly, this is not a direct generalization of {
f
, but merely a way to account for the
rolling motion of the body.
Next, with the generalized body position r
b
=

x
b
, z
b
,

T
and the generalized body
momentum p
b
=

x
b
,
b

z
b
, ^
b

T
, the following expression for the body virial is
proposed:
B
b
= r
b
p
b
= (
b
+{
b
. (9.178)
With the hydrodynamic pressure forces and moments represented by

F
b
, the fundamental
equations of motion for the body read

p
b
=

F
b
. (9.179)
The total time derivative of the body virial is
dB
b
dt
=

r
b
p
b
+

p
b
r
b
. (9.180)
Substitution of (9.179) and time averaging over [0, T] yields
2 < /
b
> + <

F
b
r
b
>=
1
T
[B
b
(T) B
b
(0)] . (9.181)
If the body performs a periodic motion with the roll angle

b
taken in [0, 2] and if
T is chosen to be the period, then the right-hand side of (9.181) vanishes; it then follows
that
< /
b
>=
1
2
<

F
b
r
b
> . (9.182)
9.4. THE VIRIAL AND CONSERVATION LAW NO.8 179
By simple addition of (9.167) and (9.182) one obtains the virial theorem for the wave-
body problem:
< /
f
+/
b
>=
1
2
<

D
f

F
f
r
f

dxdz +

F
b
r
b
> . (9.183)
Given explicit expressions for the periodic inertial and gravitational forces, the above
identity can be used to determine the averaged kinetic energy of the wave-body system.
9.4.2 The circulation alternative
When the uid motion is not irrotational, there no longer exists a velocity potential
. Nevertheless, Longuet-Higgins showed that the velocity-dependent integral quanti-
ties (9.40-9.42) and (9.45) still can be dened in terms of the uid velocity eld (u
f
, w
f
)
T
;
he also proved that the corresponding seven conservation laws for the water-wave prob-
lem remain valid. However, (quoting Longuet-Higgins) there appears to be no simple
analogue to B
f
for motion with vorticity.
For rotational ow, Longuet-Higgins introduced the uid circulation as a substitute for
B
f
:
(
f
=

D
f


f
dxdz , (9.184)
where the uid vorticity
f
is dened as

f
=
w
f
x

u
f
z
. (9.185)
Then, from the vorticity equation for two-dimensional ow
D
Dt
[
f
] = 0 , (9.186)
he obtained the well-known circulation theorem
d(
f
dt
= 0 , (9.187)
stating that for the two-dimensional water-wave problem the circulation is a constant of
the motion.
The straightforward generalization of the circulation concept to the wave-body problem
is doomed to failure, since the motion of a rigid body is generally rotational. However,
on the analogy of (
f
the body circulation may be put as
(
b
=

D
f


f
dxdz , (9.188)
where the body vorticity is dened as

b
=
w
b
x

u
b
z
= 2

b
. (9.189)
180 CHAPTER 9. SYMMETRIES AND CONSERVATION LAWS
To arrive at the last expression, (9.90-9.91) have been used. Thus, one obtains the
following expression for (
b
:
(
b
= 2
b

b
, (9.190)
Apparently, (
b
is conserved if and only if the rotational motion is uniform.
21
9.4.3 Fluid-lled deformable bodies
Consider the two-dimensional wave-body problem depicted in Figure 9.3, where an im-
permeable, exible bag S lled with an ideal uid the oating body is submerged
in an ideal, irrotational uid.
Figure 9.3: Free surface waves and submerged deformable body.
Assuming the uid ow inside the bag to be irrotational, a potential

can be introduced
to represent the interior velocity eld. For this particular conguration, the virials of the
body and the surrounding uid are dened as
B
b
=

D
b


b
[(x

)
x
+ (z

)
z
] dxdz =

(z dx xdz) , (9.191)
B
f
=

D
f


f
[(x)
x
+ (z)
z
] dxdz =

CS

f
(z dx xdz) . (9.192)
To obtain expressions for the total time derivatives of these virial quantities, note that
from (9.60)
D
Dt
[(x)
x
+ (z)
z
] =
D
Dt
[x
x
+z
z
+ 2]
=

2
x
+
2
z

xp
x
z (p
z
+g)
2

(p +gz)
1
2

2
x
+
2
z

= (xp)
x
(zp)
z
+ 2

2
x
+
2
z

3gz . (9.193)
21
Obviously, this problem is simply by-passed when
b
is replaced by
b
2

b
; it then follows that
C
b
= 0 for all times.
9.4. THE VIRIAL AND CONSERVATION LAW NO.8 181
A similar expression can be derived for the interior uid ow represented by

. Then,
from the same arguments that have been used in the proof of (9.61-9.67), the following
identities can be deduced:
dB
f
dt
=

CS
p (z dx xdz) + 4 /
f
3 {
f
, (9.194)
dB
b
dt
=

S
p

(z dx xdz) + 4 /
b
3 {
b
, (9.195)
where p

denotes the Bernoulli pressure inside the bag; /


b
and {
b
denote the kinetic
energy and the potential energy of the interior uid.
Adding (9.194) and (9.195), with the notion that n

= n along S, gives
d
dt
[B
f
+B
b
] = 4 (H
f
+H
b
) 7 ({
f
+{
b
)

S
(p p

) (z dx xdz) .
(9.196)
Permanent contact between the exterior uid and the deformable body is guaranteed by
the boundary condition
p = p

on S . (9.197)
Substitution into (9.196) yields
dB
s
dt
= 4 H
s
7 {
s
, (9.198)
where B
s
B
f
+B
b
, etc. Then, by integration in time, an extra constant of the motion
22
is obtained for this special conguration:
c
8
= B
s
(4 H
s
7 {
s
) t , (9.199)
which, in the absence of gravity, reduces to the statement that the rate of growth of the
virial equals the total energy multiplied by four.
9.4.4 The broken symmetry argument
Here an attempt is made towards a conclusive answer to the question whether there
exists an analogue to B
f
for a rigid body or not. Therefore, let us return to Benjamin
and Olver, and the summary of their analysis presented in section 9.2.
The vector elds v
i
from theorem 5, the symmetry groups G
i
from theorem 6, the
densities T
i
in theorem 7, and the integral quantities in theorem 8 are related according
to the table below. Obviously, all integral quantities have a one-to-one correspondence
with a single vector eld excepting the virial B
f
, which is related to two vector elds;
actually, Benjamin and Olver used a linear combination of v
7
and v
9
to obtain T
7
as a
conserved density.
In theorem 7 it was stated that when surface tension is operative i.e. > 0 in (9.1-
9.4) v
9
is no longer the innitesimal generator of a one-parameter symmetry subgroup;
22
With proper denitions, it can be shown that the seven conservation laws listed in theorem 8 also
apply to this problem.
182 CHAPTER 9. SYMMETRIES AND CONSERVATION LAWS
vector symmetry conserved integral physical
eld group density quantity interpretation
v
1
G
1
T
1
1
f
horizontal momentum
v
2
G
2
T
2
H
f
energy
v
3
G
3
T
3

f
mass
v
4
G
4
T
4
.
f
vertical momentum
v
5
G
5
T
5

f
x
f
horiz. pos. mass centroid
v
6
G
6
T
6

f
z
f
vert. pos. mass centroid
v
7
, v
9
G
7
, G
9
T
7
B
f
virial
v
8
G
8
T
8
/
f
angular momentum
Table 9.1: Interrelation between vector elds, symmetry groups, conserved densities, and
integrals quantities.
consequently, the density T
7
is no longer conserved. Hence, it can be expected that in the
presence of surface tension, any conservation law involving B
f
will be lost. This, however,
might be the answer to the question regarding the generalization of the virial to a wave-
body system; surface tension is usually viewed at as acting like a stretched membrane,
thus modifying the free surface pressure see the dynamic free surface condition (9.3).
The same eect can be expected from the presence of a rigid body, oating in or below
the free surface; the body exerts a (positive) pressure force on the surrounding uid.
Then, from the same argument, it follows that the scaling symmetry G
9
is broken, and
consequently there is no conservation of a virial-like quantity for the wave-body system.
As a nal remark, it is stressed that the above reasoning should not be interpreted
as a proof of the non-existence of a virial for wave-body systems. So far, it can only
be surmised that the formal extension of a uid quantity like B
f
and the corresponding
conservation law is impossible.
9.5 Symmetry-breaking boundaries
In the proof of conservation laws (9.92-9.98) it was assumed that the Bernoulli pressure
p vanished on C`S. However, in practical situations this is not a reasonable assumption;
in many physical applications impermeable xed boundaries for instance, a bottom
are present. In this section we discuss the impact of such boundaries on the previous
results.
Consider the problem depicted in Figure 9.4: the uid, with the body oating in or below
the free surface, is bounded by an impermeable xed bottom B (which is not necessarily
even) and lateral boundaries V and W (which are not necessarily xed). The free surface
and the wetted part of the body surface are denoted by F and S respectively.
It is obvious that, with V and W either impermeable and xed or moving in a Lagrangian
manner, the total mass is a constant of the motion, which is expressed by (9.77).
With regard to the horizontal momentum, note that from the vanishing of p on F
9.5. SYMMETRY-BREAKING BOUNDARIES 183
Figure 9.4: Wave-body system with impermeable xed boundaries.
and (9.64)
d1
f
dt
=

S
p dz +

B
p dz +

V W
p dz . (9.200)
Using (9.84) and (9.47), one arrives at
d1
s
dt
=

B
p dz +

V W
p dz . (9.201)
The integral over B equals zero if the bottom is horizontal. If the problem is periodic in
horizontal direction, the integrals over V and W are equal in magnitude but of opposite
sign. So, with these two conditions it is found that (d/dt) 1
s
= 0, stating that 1
s
is a
constant of the motion. The same result is obtained if the problem (i.e. the geometry,
the uid ow, and the body motion) is symmetric about a vertical plane x = x
0
.
Similarly, with regard to the vertical momentum, we deduce from p = 0 on F and (9.65)
d.
f
dt
=

S
p dx

B
p dx

V W
p dx
f
g . (9.202)
Then, using (9.85) and (9.48)
d.
s
dt
+
s
g =

B
p dx

V W
p dx . (9.203)
If V and W are impermeable xed vertical walls, then the corresponding integral con-
tributions vanish. If the problem is periodic in horizontal direction, then the integrals
over V and W are equal in magnitude but of opposite sign. However, symmetry about
a vertical plane x = x
0
does not imply the vanishing of these integrals. Likewise, the
integral over B is time dependent in general; therefore, .
s
+
s
gt is not a constant of
the motion.
184 CHAPTER 9. SYMMETRIES AND CONSERVATION LAWS
With the vanishing of p on F and (9.66) it follows that
d/
f
dt
=

S
p (xdx +z dz)

B
p (xdx +z dz)

V W
p (xdx +z dz)
f
g x
f
. (9.204)
Using (9.84-9.86) and (9.49), one obtains
d/
s
dt
+
s
g x
s
=

B
pxdx

V W
pxdx

B
pz dz

V W
pz dz .
(9.205)
Note that a horizontal bottom combined with horizontal periodicity is not a sucient
condition for the conservation of angular momentum, since the contribution of the rst
three integrals is not necessarily zero then. However, if the problem is symmetric about
the plane x = 0, then one has (d/dt) /
s
= 0, since the integrals over B vanish and the
integrals over V and W are equal in magnitude but of opposite sign; in that particular
case /
s
is a constant of the motion (use x
s
= 0).
From p = 0 on F and (9.67) it follows that
dH
f
dt
=

SBW
p (
z
dx
x
dz) =

SBW
p
n
ds . (9.206)
With the notion that
n
= 0 along B W and using (9.89), (9.84-9.86) and (9.50-9.52),
it is found that
dH
s
dt
= 0 . (9.207)
Finally, the eects of B and W on the conservation of the virial for the original problem
without a oating body are discussed. With p vanishing on F and (9.194) it follows that
dB
f
dt
=

BW
p (z dx xdz) + 4 H
f
7 {
f
. (9.208)
In case of a horizontal bottom and/or vertical walls some integral contributions vanish.
However, neither horizontal periodicity nor symmetry about a plane x = x
0
is sucient
to make the total contribution of the integrals vanish; therefore, B
f
(4 H
f
7 {
f
) t is
no longer a constant of the motion.
Summarizing, the presence of impermeable xed boundaries aects the horizontal, ver-
tical and angular momentum, and the virial of the system. In general, these quantities
are no longer constants of the motion; the bottom B breaks the corresponding symme-
tries. In the special case of a horizontal bottom and horizontal periodicity, the horizontal
momentum is a constant of the motion. The same result is obtained if the problem is
symmetric about a plane x = x
0
; if x
0
= 0 the angular momentum is conserved too. The
presence of an impermeable bottom breaks the vertical translation symmetry (even in
case of symmetry about a plane x = x
0
), and consequently the vertical momentum is no
longer a constant of the motion. The mass and energy conservation laws are not aected
by the introduction of impermeable xed boundaries.
9.6. NUMERICAL VALIDATION 185
9.6 Numerical validation
In this section the theory for conservation laws for the wave-body problem is conrmed
from numerical results for a simple test conguration.
Figure 9.5: Wave-body system: initial
conguration.
Figure 9.6: Wave-body system: equilib-
rium state.
The results presented apply to a simple test conguration; the initial position of the body
and the free surface are depicted in Figure 9.5, while the equilibrium state of this system
is shown in Figure 9.6. The tank length is 20 m, the still water level in the equilibrium
state is 5 m, and the initial upward displacement of the cylinder (with respect to the
initial water level) is 1 m; the cylinder diameter is 6 m. The ensuing free motion of the
body generates an outgoing wave eld; a direct exchange of energy (and other quantities)
from the uid to the body and vice versa is expected.
Figure 9.7 shows
f
,
b
and
s
; of course, each of these quantities is constant, which
is conrmed (fortunately!) by the computed results. The uid mass is computed by
using the contour integral expression in (9.37).
Since the problem is symmetric about the vertical plane through the body centre of
gravity, the horizontal momentum is a constant of the motion; this is conrmed by
Figure 9.8, showing 1
f
, 1
b
and 1
s
. Note that the order of magnitude is 10
5
, which makes
it plausible to attribute the variations to numerical discretization errors. A temporary
increase of the amplitude is observed in the time interval [8 s, 10 s]; this is due to the
reection of the radiated waves against the impermeable xed lateral boundaries.
With the presence of an impermeable xed bottom, it is expected that the vertical
momentum is not a constant of the motion. From Figure 9.9 it is clear that .
f
and .
b
are both approximately sinusoidal, and more or less in counter phase, but at dierent
amplitudes; their sum .
s
is nearly sinusoidal too, but at a doubled frequency. Thus, the
theoretical prediction that the vertical momentum is not preserved is conrmed by
the numerical computations.
Since the problem has been chosen such that it is symmetric about x = 0, the angular
momentum is a constant of the motion; this is conrmed by Figure 9.10, showing /
f
,
/
b
and /
s
(order of magnitude: 10
4
). Again, a temporary increase in amplitude is
observed in the time interval [8 s, 10 s].
186 CHAPTER 9. SYMMETRIES AND CONSERVATION LAWS
Figure 9.7: Conservation of mass.
Figure 9.11 shows the energies of the uid, the body and the wave-body system. It is
clear that H
f
and H
b
are both sinusoidal (with the same amplitude) and in exact counter
phase. Evidently, their sum H
s
is a constant of the motion.
Finally, from Figure 9.12 we note that, roughly speaking, B
f
oscillates sinusoidally round
a linearly decaying function; this is in fair agreement with Longuet-Higgins conclusion
that, for the original problem without a oating body (in the absence of gravity), B
f

4 H
f
t is a constant of the motion.
23
23
This result has also been obtained for a submerged uid-lled deformable body, see paragraph 9.4.3.
9.6. NUMERICAL VALIDATION 187
Figure 9.8: Conservation of horizontal momentum.
THIS PAGE INTENTIONALLY LEFT BLANK
188 CHAPTER 9. SYMMETRIES AND CONSERVATION LAWS
Figure 9.9: Exchange of vertical momentum no conservation.
Figure 9.10: Conservation of angular momentum.
9.6. NUMERICAL VALIDATION 189
Figure 9.11: Exchange and conservation of energy.
Figure 9.12: Virial no conservation.
190 CHAPTER 9. SYMMETRIES AND CONSERVATION LAWS
9.7 Bibliography
Few relevant publications which have not been cited in the text are also included here.
Arnold, V.I., and Khesin, B.A. 1992. Topological methods in hydrodynamics. An-
nual Review of Fluid Mechanics 24:145-166.
Benjamin, T.B. 1974. Lectures on nonlinear wave motion. Lectures in Applied Math-
ematics 15:3-47.
Benjamin, T.B. 1986a. On the Boussinesq model for two-dimensional wave motions in
heterogeneous uids. Journal of Fluid Mechanics 165:445-474.
Benjamin, T.B. 1986b. Note on added mass and drift. Journal of Fluid Mechanics
169:251-256.
Benjamin, T.B., and Mahony, J.J. 1971. On an invariant property of water waves.
Journal of Fluid Mechanics 49:385-389.
Benjamin, T.B., and Olver, P.J. 1982. Hamiltonian structure, symmetries and
conservation laws for water waves. Journal of Fluid Mechanics 125:137-185.
Broer, L.J.F, van Groesen, E.W.C., and Timmers, J.M.W. 1976. Stable model
equations for long water waves. Applied Scientic Research 32:619-636.
Crapper, G.D. 1979. Energy and momentum integrals for progressive capillary-gravity
waves. Journal of Fluid Mechanics 94(1):13-24.
van Daalen, E.F.G., and van Groesen, E.W.C. 1993. Conservation laws for water
waves and oating bodies. Submitted to the Journal of Fluid Mechanics.
Goldstein, H. 1980. Classical Mechanics. Addison-Wesley.
Lax, P.D. 1968. Integrals of nonlinear equations of evolution and solitary waves. Com-
munications on Pure and Applied Mathematics 21:467-490.
Longuet-Higgins, M.S. 1974. On the mass, momentum, energy and circulation of a
solitary wave. Proceedings of the Royal Society of London, Series A, 337:1-13.
Longuet-Higgins, M.S., and Fenton, J.D. 1974. On the mass, momentum, en-
ergy and circulation of a solitary wave II. Proceedings of the Royal Society of London,
Series A, 340:471-493.
Longuet-Higgins, M.S. 1975. Integral properties of periodic gravity waves of nite
amplitude. Proceedings of the Royal Society of London, Series A, 342:157-174.
Longuet-Higgins, M.S. 1980. Spin and angular momentum in gravity waves. Journal
of Fluid Mechanics 97:1-25.
Longuet-Higgins, M.S. 1983. On integrals and invariants for inviscid, irrotational
ow under gravity. Journal of Fluid Mechanics 134:155-159.
Longuet-Higgins, M.S. 1984. New integral relations for gravity waves of nite ampli-
tude. Journal of Fluid Mechanics 149:205-215.
Miles, J.W., and Salmon, R. 1985. Weakly dispersive nonlinear gravity waves.
Journal of Fluid Mechanics 157:519-531.
Miura, R.M., and Gardner, C.S. 1968. Korteweg-de Vries equation and general-
izations II. Existence of conservation laws and constants of motion. Journal of Mathe-
9.7. BIBLIOGRAPHY 191
matical Physics 9:1204-1209.
Noether, E. 1918. Invariante Invariationsprobleme. Nachrichten der Gesellschaft der
Wissenschaften Gottingen 2:235-237.
Olver, P.J. 1980a. On the Hamiltonian structure of evolution equations. Mathematical
Proceedings of the Cambridge Philosophical Society 88:71-88.
Olver, P.J. 1980b. Applications of Lie Groups to Dierential Equations. Lecture Notes,
Mathematical Institute, University of Oxford.
Olver, P.J. 1983. Conservation laws of free boundary problems and the classication of
conservation laws for water waves. Transactions of the American Mathematical Society
277(1):353-380.
Olver, P.J. 1986. Applications of Lie Groups to Dierential Equations.
Springer-Verlag.
Ripa, P. 1981. Symmetries and conservation laws for internal gravity waves. Proceed-
ings of the American Institute of Physics 7b:281-386.
Starr, V.P. 1949a. A momentum integral for surface waves in deep water. Journal of
Maritime Research 6:126-135.
Starr, V.P. 1949b. Momentum and energy integrals for gravity waves of nite height.
Journal of Maritime Research 6:175-193.
Uhlenbeck, G.E. 1980. Some notes on the relation between uid mechanics and sta-
tistical physics. Annual Review of Fluid Mechanics 12:1-9.
192 CHAPTER 9. SYMMETRIES AND CONSERVATION LAWS
Now had they left behind the gulf named after the Ambracians, now with
sails wide spread the land of the Curetes, and in next order the narrow islands
with the Echinades, and the land of Pelops was just descried; even then a
baleful blast of the north wind seized them in mid-course and swept them
towards the Libyan sea nine nights and as many days, until they came far
within Syrtis
24
, wherefrom is no return for ships, when they are once forced
into that gulf. For on every hand are shoals, on every hand masses of seaweed
from the depths; and over them the light foam of the wave washes without
noise; and there is a stretch of sand to the dim horizon; and there moveth
nothing that creeps or ies. Here accordingly to the ood-tide for this tide
often retreats from the land and bursts back again over the beach coming on
with a rush and roar thrust them suddenly on to the innermost shore, and
but little of the keel was left in the water. And they leapt forth from the ship,
and sorrow seized them when they gazed on the mist and the levels of vast
land stretching far like a mist and continuous into the distance; no spot for
water, no path, no steading of herdsmen did they descry afar o, but all the
scene was possessed by a dead calm. And thus did one hero, vexed in spirit,
ask another:
What land is this? Whither has the tempest hurled us? Would that,
reckless of deadly fear, we had dared to rush on by that same path between the
clashing rocks! Better were it to have overleapt the will of Zeus and perished
in venturing some mighty deed. But now what should we do, held back by the
winds to stay here, if ever so short a time? How desolate looms before us the
edge of the limitless land!
Thus one spake; and among them Ancaeus the helmsman, in despair at
their evil case, spoke with grieving heart: Verily we are undone by a terrible
doom; there is no escape from ruin; we must suer the cruellest woes, having
fallen on this desolation, even though breezes should blow from the land; for,
as I gaze far around, on every side do I behold a sea of shoals, and masses
of water, fretted line upon line, run over the hoary sand. And miserably long
ago would our sacred ship have been shattered far from the shore; but the tide
itself bore her high on to the land from the deep sea. But now the tide rushes
back to the sea, and only the foam, whereon no ship can sail, rolls round us,
just covering the land. Wherefore I deem that all hope of our voyage and of
our return is cut o. Let someone else show his skill; let him sit at the helm
the man that is eager for our deliverance. But Zeus has no will to full
our day of return after all our toils.
Argonautica, Book IV, Verses 1228-1276.
24
quicksands in Libya
Chapter 10
Radiation Boundary
Conditions
. . . This is the tale the Muses told; and I sing obedient to the Pierides, and
this report have I heard most truly; that ye, O mightiest far of the sons of
kings, by your might and your valour over the desert sands of Libya raised
high aloft on your shoulders the ship and all that ye brought therein, and bare
her twelve days and nights alike. Yet who could tell the pain and grief which
they endured in that toil? Surely they were of the blood of the immortals, such
a task did they take on them, constrained by necessity. How forward and how
far they bore her gladly to the waters of the Tritonian lake! How they strode
in and set her down from their stalwart shoulders!
Argonautica, Book IV, Verses 1381-1392.
10.1 Introduction
In this chapter it is shown how variational principles and conservation laws can be used
to derive radiation (or absorbing) boundary conditions for partial dierential equations
which describe wave phenomena. Such boundary conditions are desirable in order to
limit the size of the computational domain and thus computation time and data storage
as much as possible, while minimizing the eects of reected waves on the solution in the
area of interest.
Many methods for developing radiation boundary conditions have been used.
1
For
wave equations, the boundary conditions proposed by Bayliss and Turkel (1980, 1982),
Engquist and Halpern (1988), Engquist and Majda (1977, 1979), and Higdon (1986, 1987,
1990) are well-known. These boundary conditions have in common that they are based
on (properties of) solutions to the partial dierential equations under consideration.
Here, radiation boundary conditions are derived without the assumption that solu-
tions are available beforehand, and therefore they are applicable to nonlinear and disper-
sive systems too. Moreover, these boundary conditions render the problem well-posed
1
A recent overview of non-reecting boundary conditions in the numerical solution of wave problems
is due to Givoli (1991).
193
194 CHAPTER 10. RADIATION BOUNDARY CONDITIONS
in the sense that the integral of some appropriate density for instance, the energy
density over the computational domain does not increase.
The present chapter is based on a paper written by van Daalen, Broeze, and van Groe-
sen (1992), with an accompanying paper by Broeze and van Daalen (1992). The outline
of this chapter is as follows: in section 10.2 we develop a theory of radiation bound-
ary conditions for continuous systems that are governed by a variational (Lagrangian)
principle. This theory is illustrated in section 10.3 with explicit boundary conditions
for a nonlinear version of the one-dimensional Klein-Gordon equation; numerical test
results are also presented. In section 10.4 we apply the theory to the classical (nonlinear)
water-wave problem; some of the ndings in the previous chapter will be re-established.
10.2 Theory for continuous systems
Consider a continuous system described by the variable u, which is a function of the
spatial coordinates x = (x
1
, x
2
, . . . , x
n
)
T
and the time t.
It is assumed that the evolution of the system can be derived from a variational
principle. With a Lagrangian density L, an expression in x, t, u, and partial derivatives
of u up to some nite order, the action integral reads
J (u) =

(t)
Lddt (10.1)
where T is some time interval, and the spatial domain may depend on time.
Restrictions on the motion of will be imposed further on, but rst the Euler-
Lagrange equation for L and natural boundary conditions are derived under the assump-
tion that the motion of is given.
For the system under consideration, Hamiltons principle can be summarized by say-
ing that the evolution of the system is such that the action integral J (u) has a stationary
value, see (for instance) Goldstein (1980). The rst variation of J (u) with respect to a
variation u is dened as
J (u; u)

d
d
J (u +u)

=0
=

(t)
DL(u) uddt , (10.2)
where
DL(u) u

d
d
L(u +u)

=0
(10.3)
denotes the Frechet
2
derivative of L, dened for the variation u.
To arrive at the Euler-Lagrange equation for L, partial integrations with respect
to x and t are required, leaving expressions on the boundary when u and its partial
derivatives do not vanish there. In general, the formula for partial integration reads
DL(u) u = L(u) u +
t
B
0
(u; u) + div

B
1
(u; u) (10.4)
2
Frechet, Maurice (1878-1975), French mathematician. He made notable contributions to the theory
of functionals.
10.2. THEORY FOR CONTINUOUS SYSTEMS 195
where L(u) is the variational derivative of L, and B
0
and

B
1
are expressions linear
in u.
Substitution of (10.4) into (10.2) gives
J (u; u) =

(t)
L(u) uddt
+

(t)

t
B
0
(u; u) + div

B
1
(u; u)

ddt . (10.5)
Taking into account the prescribed motion of the spatial domain and applying Gauss
divergence theorem, this equation can be written as
J (u; u) =

(t)
L(u) ud +

(t)

B
1
(u; u) ndS

dt
+

d
dt

(t)
B
0
(u; u) d

(t)
B
0
(u; u) v
n
dS

dt , (10.6)
where n is the outward pointing normal to , and v
n
is the velocity of the boundary in
normal direction.
Allowing u and its partial derivatives to be arbitrary at the boundary , but
vanishing at the end points of the time interval T, the second term in the right-hand side
vanishes (since B
0
is linear in u), and the rst variation of J (u) with respect to u is
then given by
J (u; u) =

(t)
L(u) uddt
+

(t)

B
1
(u; u) n B
0
(u; u) v
n

dS dt . (10.7)
From this expression the evolution of the system follows; the Euler-Lagrange equation
reads
L(u) = 0 in (10.8)
and the natural boundary condition on the moving boundary is given by
B
0
(u; u) v
n


B
1
(u; u) n = 0 on (10.9)
for arbitrary u.
Example 1: If the Lagrangian density L depends on x, t, u, and partial derivatives up
to rst order only, the Frechet derivative of L with respect to a variation u reads
DL(u) u =
L
u
u +
L
u
t
u
t
+
L
u
x
i
u
x
i
, (10.10)
196 CHAPTER 10. RADIATION BOUNDARY CONDITIONS
where we used Einsteins summation convention. This can be written as
DL(u) u =

L
u

t
L
u
t

x
i
L
u
x
i

u
+
t

L
u
t
u

+
x
i

L
u
x
i
u

. (10.11)
The rst variation of J is given by (10.5), with the variational derivative of L given by
L(u) =
L
u

t
L
u
t

x
i
L
u
x
i
, (10.12)
and with the following boundary expressions:
B
0
(u; u) =
L
u
t
u , (10.13)

B
1
(u; u) =

L
u
x
1
,
L
u
x
2
, . . . ,
L
u
x
n

T
u . (10.14)
The vanishing of J (u; u) provides the Euler-Lagrange equation see (10.8)
L
u

t
L
u
t

x
i
L
u
x
i
= 0 on (10.15)
and the natural boundary condition see (10.9)
L
u
t
v
n

L
u
x
i
n
i
= 0 on (10.16)
Aiming to derive boundary conditions on a xed domain of integration that do not
inuence the solution in the interior of the domain, several options are available. One
important a priori requirement in the development of radiation boundary conditions is
that they should not be based on (properties of) solutions to the partial dierential
equations considered, since in general (for nonlinear equations) such solutions can not
easily be found. Compare this, for instance, with Higdons (1987) derivation of absorbing
boundary conditions, that is based on an exact plane monochromatic wave.
The possibility that will be examined in this chapter is to nd boundary conditions
that simulate a moving domain of integration such that the energy is conserved as well
as possible. The naive idea is that if the correct amount of energy is uxed through the
xed boundary, any reection caused by the presence of the boundary will have little
energy and therefore will be small in this sense. This reasoning may be valuable, since
energy is a denite quantity. However, other densities for example, the momentum
density may do as well, as will be shown further on.
In the derivation above, the motion of the domain was assumed to be given. Then the
evolution of the system is completely determined by the Euler-Lagrange equation (10.8)
and the natural boundary condition (10.9). From now on, the motion of the domain is
not prescribed, but instead it is demanded that moves in such a way that the integral
over of some density is conserved.
10.2. THEORY FOR CONTINUOUS SYSTEMS 197
In order to show the dependence of the results on the choice of density later on,
we shall consider quite generally a density for which a local conservation law can be
obtained with Noethers theorem. For a comprehensive description of this well-known
theorem the reader is referred to Olver (1986) and the historic work of Emmy Noether
herself (1918). The main principle of Noethers theorem can be formulated as follows
3
:
if the Lagrangian density is invariant under a given transformation of the variables, then
there is a corresponding functional that is conserved.
So, let the Lagrangian density L be invariant under a variation (u), then it can be
proven that for some scalar density
0
(u; (u)) and some vector density
1
(u; (u)) it
holds that for all u
DL(u) (u) =
t

0
(u; (u)) + div
1
(u; (u)) (10.17)
A combination of this equation and the integration-by-part formula (10.4) yields, taking
u = (u)
L(u) (u) =
t

0
(u; (u)) B
0
(u; (u))

+ div


1
(u; (u))

B
1
(u; (u))

, (10.18)
leading to a local conservation law of the form

t
e (u) + div

f
e
(u) = 0 (10.19)
for solutions of the Euler-Lagrange equation (10.8). In this equation the density e and
the corresponding ux density

f
e
are given by
e (u)
0
(u; (u)) B
0
(u; (u)) (10.20)

f
e
(u)
1
(u; (u))

B
1
(u; (u)) (10.21)
As motivated above, let the motion of be restricted in such a way that the quantity
E (u)

(t)
e (u) d (10.22)
is conserved.
Dierentiation of (10.22) with respect to the time gives
dE
dt
=

(t)

t
e (u) d +

(t)
e (u) v
n
dS . (10.23)
Substitution of (10.19) and application of Gauss divergence theorem yields
dE
dt
=

(t)
div

e (u) v

f
e
(u)

d , (10.24)
where v represents a velocity eld on , such that
v
n
= v n on . (10.25)
3
See also, for instance, Goldstein (1980).
198 CHAPTER 10. RADIATION BOUNDARY CONDITIONS
If the quantity E (u) is conserved, then v must satisfy

(t)
div

e (u) v

f
e
(u)

d = 0 . (10.26)
Since the initial domain can be taken arbitrarily, v necessarily equals the local ux
velocity v
f
, dened by

f
e
(u) = v
f
(u) e (u) . (10.27)
So, in particular, the normal velocity v
n
on the boundary equals the local ux velocity
in normal direction:
v
n
= v
f
(u) n =

f
e
(u) n
e (u)
on (10.28)
at points where e (u) does not vanish. The latter condition can also be obtained directly
from (10.23); substitution of (10.19) and application of Gauss divergence theorem gives
dE
dt
=

(t)

e (u) v
n


f
e
(u) n

dS . (10.29)
If the quantity E is conserved, then v
n
is determined by
e (u) v
n


f
e
(u) n = 0 on (10.30)
in accordance with (10.28).
Remark 1: Substitution of (10.27) into (10.19) gives

t
e (u) +v
f
(u) e (u) +e (u) divv
f
(u) = 0 , (10.31)
which in case of a divergence-free ux velocity eld v
f
(u) reduces to
De (u)
Dt
=
t
e (u) +v
f
(u) e (u) = 0 , (10.32)
expressing that e (u) is constant along the streamlines of the corresponding ux velocity
eld v
f
(u).
Remark 2: If e (u) equals the energy density, then the local ux velocity v
f
(u) dened
by (10.27) is well-known, in particular for linear systems for which it equals the group
velocity for monochromatic waves; see, for instance, Broer (1951), Biot (1957), and
Lighthill (1965). It is the pointwise version of the centro-velocity i.e. the velocity of
the centre of gravity of the density e (u) which is given by
v
C
(u) =

F
e
(u)
E (u)
, (10.33)
where

F
e
(u)

f
e
(u) d , E (u)

e (u) d , (10.34)
10.2. THEORY FOR CONTINUOUS SYSTEMS 199
see also Wehausen and Laitone (1960), van Groesen (1980), and van Groesen and Mainardi
(1990).
Remark 3: If e equals the energy density, then the ideas above are related to some parts
of the earlier work of Whitham (1974). In his treatment of energy propagation for the
Klein-Gordon equation a slowly varying wave train is considered. It is found that the
averaged energy density

E and the averaged energy ux density

F (i.e. averaged over
one period) satisfy

F = C (k)

E , (10.35)
where C (k) denotes the group velocity depending on the wave number k. Compare this
averaged result with the denition of the local ux velocity in (10.27).
Next, an averaged energy equation is proposed:

t

E +
x

C (k)

E

= 0 , (10.36)
which is the dierential form of the statement that
The total energy between any two group lines remains constant.
Evidently, (10.36) in combination with (10.35) can be regarded as a global conservation
law. Compare this with our local conservation law (10.19), and note that the present
derivation of radiation boundary conditions started from a similar, yet more general,
principle, namely the conservation of an appropriate density on a moving domain of
integration.
With the explicit expressions (10.20) and (10.21) for e and

f
e
respectively, condition (10.30)
reads

0
(u; (u)) B
0
(u; (u))

v
n


1
(u; (u))

B
1
(u; (u))

n = 0 .
(10.37)
However, the natural boundary condition, given in (10.9), already implies, taking u =
(u)
B
0
(u; (u)) v
n


B
1
(u; (u)) n = 0 on , (10.38)
and so the additional boundary condition for conservation of the quantity E (u) is given
by

0
(u; (u)) v
n

1
(u; (u)) n = 0 on (10.39)
where
0
(u; (u)) and
1
(u; (u)) are obtained from (10.17).
Summarizing, the proposed boundary conditions for a domain , chosen in such a way
that the total amount of a density e (u) is preserved, are given by the natural boundary
condition (10.9) and the additional boundary condition (10.39).
200 CHAPTER 10. RADIATION BOUNDARY CONDITIONS
Example 2: If the rst order Lagrangian density L does not depend explicitly on x
and t, then the energy density can be found with Noethers theorem from a variation
corresponding to the invariance of L with respect to time translations:
(u) = u
t
. (10.40)
Similarly, the momentum density in a given direction is obtained with a variation
corresponding to the invariance of L with respect to spatial translations:
(u) = u with =

0) . (10.41)
Substitution of u = (u) into (10.10) yields
DL(u) u
t
=
L
u
u
t
+
L
u
t
u
tt
+
L
u
x
i
u
tx
i
=
t
L(u) , (10.42)
and therefore (10.17) holds with

0
(u; (u)) = L(u) , (10.43)

1
(u; (u)) =

0 . (10.44)
Then, indeed, substitution of (10.13) and (10.43) into (10.20) yields the energy density
e (u) = L(u)
L
u
t
u
t
, (10.45)
with the energy ux density see (10.14), (10.21), and (10.44)

f
e
(u) =

L
u
x
1
,
L
u
x
2
, . . . ,
L
u
x
n

T
u
t
. (10.46)
Similarly, substitution of u = (u) into (10.10) gives
DL(u) (u ) =
L
u
(u ) +
L
u
t
(u )
t
+
L
u
x
i
(u )
x
i
= L(u)
= div (L(u)) , (10.47)
and (10.17) holds with

0
(u; (u)) = 0 , (10.48)

1
(u; (u)) = L(u) . (10.49)
The momentum density m(u) in the direction reads
m(u) =
L
u
t
(u ) , (10.50)
with momentum ux density

f
m
(u) = L(u)

L
u
x
1
,
L
u
x
2
, . . . ,
L
u
x
n

T
(u ) . (10.51)
10.2. THEORY FOR CONTINUOUS SYSTEMS 201
Substitution of (10.43) and (10.44) into (10.39) yields the additional boundary condition
for energy conservation
L(u) v
n
= 0 on . (10.52)
Similarly, the additional boundary condition for momentum conservation reads
L(u) ( n) = 0 on . (10.53)
Consequently, taking
L(u) = 0 on , (10.54)
it follows that both energy and momentum (in any direction) are conserved in this special
case.
Remark 4: Whitham (1970) has considered dispersive wave problems in which the
Euler-Lagrange equation has approximate solutions of the form
u = U (, a) , = kx t , (10.55)
where, in the nonlinear case, the amplitude a, the wave number k and the frequency
will generally vary in time; k and are generalized by dening them as
k (x, t) =

x
, (x, t) =

t
. (10.56)
It is assumed that , k and a are slowly varying functions of x and t, corresponding to
the slow modulation of the wave train considered. The averaged Lagrangian

L is then
dened as

L(, k, a) =
1
2
2

0
L(u) d , (10.57)
and is calculated by substitution of the uniform periodic solution u = U (, a) in L.
The equations for , k, and a are then obtained from the averaged variational principle



L(, k, a) dt dx = 0 . (10.58)
In the linear case we have

L(, k, a) = G(, k) a
2
, (10.59)
and a variation of

L with respect to a yields the linear dispersion relation
4
G(, k) = 0 . (10.60)
It is then observed that the stationary value of

L is zero. Compare this global condition
under the assumptions of slowly varying wave trains and periodic solutions to the
pointwise boundary condition (10.54).
In the following sections the theory will be applied to a nonlinear one-dimensional wave
equation, and to the nonlinear three-dimensional water-wave problem.
4
Note that this relation does not involve the amplitude a, and compare this result with the discussion
in paragraph 2.4.3, in particular the dispersion relation (2.39).
202 CHAPTER 10. RADIATION BOUNDARY CONDITIONS
10.3 Application to one-dimensional wave equations
In this section the theory is applied to a nonlinear one-dimensional wave equation. The
boundary conditions obtained will be used to simulate open boundaries, i.e. boundaries
which give low reections for radiating waves.
Consider a nonlinear version of the Klein-Gordon equation:
u
tt
= c
2
u
xx
V

(u) , (10.61)
where V

(u) is some nonlinear function of u the prime denoting dierentiation with


respect to u and c is the wave velocity. This hyperbolic equation arises in various
physical problems; for example, if V

(u) = sin u, then (10.61) is known as the sine-


Gordon equation.
5
The action integral for this problem is given by
J (u) =

T
x
1

x
0
L(u) dxdt , (10.62)
where the Lagrangian density L is a function of u and its rst order derivatives note
that L does not depend explicitly on t:
L(u) =
1
2

u
2
t
c
2
u
2
x

+V (u) . (10.63)
Here the potential function V (u) is given by
V (u) =
1
2
u
2
+
1
4
u
4
, (10.64)
which is an appropriate expansion for even functions V . In our test cases and are
restricted to positive values to guarantee the positive deniteness of the energy density,
which for this problem reads see (10.45) and (10.63-10.64)
e (u) =
1
2

u
2
t
+c
2
u
2
x

+V (u) . (10.65)
The natural boundary condition for this system is obtained from (10.16):
c
2
u
n
+v
n
u
t
= 0 at x = x
0,1
, (10.66)
and (10.52) yields the additional boundary condition:

1
2

u
2
t
c
2
u
2
x

V (u)

v
n
= 0 at x = x
0,1
. (10.67)
5
Clearly, the sine-Gordon equation is nonlinear, with a nonlinear amplitude-dependent dispersion
relation. It can have solutions with properties shared by only a few other nonlinear equations. These
solutions are travelling wave disturbances that can pass through each other and emerge with unchanged
shape, apart perhaps from a phase shift. Such solutions are also found, for example, for the nonlinear
Korteweg-de Vries equation u
t
+ uu
x
+ u
xxx
= 0, where and are constants. These waves that
preserve their shape even through interactions have been termed solitons or solitary waves and are
nding an expanding area of application throughout physics, from elementary particles to solid-state
physics.
10.3. APPLICATION TO ONE-DIMENSIONAL WAVE EQUATIONS 203
From (10.66) and (10.67) it can be deduced that either v
n
= u
n
= 0 must hold or
v
n
= c

cu
n
u
t

at x = x
0,1
, (10.68)
where u
t
and u
n
are related by L(u) = 0, that is
u
2
t
c
2
u
2
x
= 2V (u) at x = x
0,1
. (10.69)
The sign of v
n
has to be chosen such as to make sure that no energy ows into the spatial
domain, since energy inow might spoil the stability of the problem; if an increasing
spatial domain is simulated, and conservation of energy is required on this domain, no
energy inow will occur on the xed domain . It is for this reason that v
n
is chosen
positive. From (10.66) it can be seen that this is enforced by requiring
u
n
u
t
0 . (10.70)
If = 0, then (10.61) and (10.64) reduce to the linear Klein-Gordon equation
u
tt
= c
2
u
xx
u . (10.71)
Solutions to (10.71) can be written as
u = ae
i
, (x, t) = kx t , (10.72)
where k is the wave number and is the frequency.
Substitution of (10.72) into (10.71) yields the (linear) dispersion relation
6

2
= c
2
k
2
+ =

c
2
k
2
+

1/2
, (10.73)
from which the phase velocity c
f
and the group velocity c
g
can be calculated:
c
f


k
= c

1 +

c
2
k
2

1/2
, c
g

d
dk
= c

1 +

c
2
k
2

1/2
. (10.74)
Substitution of (10.72) into (10.68) gives (n is taken in positive x-direction)
v
n
=
c
2
k

= c
g
, (10.75)
thus arriving at a result that might have been expected for the linear case; conservation of
energy on a moving domain leads to a boundary velocity which equals the group velocity.
Finally, note that if both and equal zero, condition (10.69) reduces to Sommer-
felds radiation condition see Orlanski (1976)
u
t
= cu
n
, (10.76)
corresponding to a boundary velocity which equals the local phase velocity c
f
if and only
if = 0:
v
Sommerfeld
=

c
2
k
= c
f
. (10.77)
6
See also remark 4.
204 CHAPTER 10. RADIATION BOUNDARY CONDITIONS
Next, we discuss the results from these numerical tests that were done for this one-
dimensional wave problem; the eld equation is given by the nonlinear Klein-Gordon
equation
u
tt
= u
xx
u u
3
with 0 , 0 . (10.78)
Three-point central discretizations have been used for the second order derivatives in (10.78).
An explicit Euler scheme has been used here, which is stable if the time step is chosen
suciently small.
The spatial domain and the time domain are given by
= [, ] , (10.79)
T = [0, 4] . (10.80)
The initial conditions have been chosen as
u(x, 0) =

cos
2
x for [x[ /2
0 for [x[ > /2
, (10.81)
u
t
(x, 0) =

(1 +)
1/2
sin 2x for [x[ /2
0 for [x[ > /2
. (10.82)
These conditions represent a wave travelling in positive direction. It is obvious that (10.81-
10.82) does not represent a steady wave for (10.78) if (, ) = (0, 0). In the small
amplitude limit, these initial conditions may be regarded as a superposition of many
monochromatic solutions, each of them travelling at a dierent phase velocity. There-
fore, it can be expected that from the very beginning some energy will propagate to the
left. This means that the radiation boundary conditions have to be implemented in both
end points x = .
The new energy-transmitting radiation boundary condition is given by
u
t
= u
n

1 +
u
2
+ (/2) u
4
u
2
n

1/2
, (10.83)
and Sommerfelds radiation condition reads
u
t
= u
n
. (10.84)
Condition (10.83) can be written as
u
t
= c (u) u
n
, (10.85)
which can be regarded as a u-dependent Sommerfeld radiation condition see (10.76)
where the correction factor (or velocity) c (u) is given by
c (u) =

1 +
u
2
+ (/2) u
4
u
2
n

1/2
. (10.86)
If u
n
approaches zero, which corresponds to a local extremum of the solution u near
the boundary, then (10.83) would give a signicant value for u
t
, involving an undesired
change in u at the boundary. Therefore, an upper limit is set to the correction velocity:
c

(u) = min

c (u) , (1 +)
1/2

. (10.87)
10.3. APPLICATION TO ONE-DIMENSIONAL WAVE EQUATIONS 205
The new radiation boundary condition then reads
u
t
= c

(u) u
n
. (10.88)
Since an explicit discretization of the eld equation has been chosen, conditions (10.84)
and (10.88) are implemented explicitly too.
For want of analytical solutions to (10.78) with initial conditions (10.81-10.82), the
numerical solutions obtained with (10.84) and (10.88) will be compared with the solution
obtained on an extended spatial domain

. It is important that

is chosen large
enough, so that reections from the boundaries of

do not reach the smaller domain


during the time interval T. Test results have indicated that this requirement is met by
the choice

= [3, 3] . (10.89)
On this extended domain the initial conditions (10.81-10.82) are imposed. Since the
numerical errors on the interior of and those on the interior of

are equal, a good


impression of the errors due to the radiation boundary conditions is obtained in this way.
Three congurations have been used for the calculation of the numerical solution
of (10.78) with initial conditions (10.81-10.82):
1. The numerical solution is determined on the spatial domain with the new energy-
transmitting boundary condition (10.88) implemented in the end points x = ,
where the correction velocity c

(u) is determined by (10.86-10.87).


2. The numerical solution is determined on the spatial domain with Sommerfelds
radiation condition (10.84) implemented in the end points x = .
3. The numerical solution is determined on the extended spatial domain

with
Sommerfelds radiation condition (10.84) implemented in the end points x = 3.
In this way we obtain an undisturbed i.e. not aected by articial boundaries
numerical solution on the smaller domain , which can be used to calculate the
errors due to the radiation boundary conditions used in congurations 1 and 2.
In order to get an impression of the error = u u, where u denotes the numerical
approximation to u, two norms are used. The rst norm is the L
2
-norm of , which is
dened as
||
2

2
dx

1/2
. (10.90)
The second norm is an energy norm of :
||
E

1
2

2
t
+
2
x

dx . (10.91)
We also compute the transient energy of the solutions on the restricted domain , dened
by
E (u)

e (u) dx , (10.92)
206 CHAPTER 10. RADIATION BOUNDARY CONDITIONS
with
e (u) =
1
2

u
2
t
+u
2
x

+
1
2
u
2
+
1
4
u
4
, (10.93)
and the transient momentum of the solutions, dened by
M (u)

m(u) dx , (10.94)
with
m(u) = u
x
u
t
. (10.95)
Five tests have been done for dierent combinations of and . In the rst three tests
equals zero, yielding a linear eld equation. However, note that boundary condi-
tion (10.88) (conguration 1) is nonlinear if either or is nonzero. In the last two
tests the values of both and are nonzero, yielding a nonlinear eld equation.
Test 1: Take = 0.1, = 0. In this case the linear disturbance term in (10.78) is
rather small. Figure 10.2 shows the L
2
-norm of the errors in the solutions obtained with
congurations 1 and 2. In Figure 10.3 the energy norms of the errors are presented. The
errors obtained with the new radiation boundary condition (10.88) are about the same
as the errors obtained with Sommerfelds radiation condition (10.84). An explanation is
deduced from (10.86); if both and are small, then the correction factor c

(u) will be
but slightly larger than unity.
It is convenient to combine Figures 10.4-10.5 and 10.1. The rst two gures show the
energy and momentum of the numerical solution obtained with congurations 1 and 2
during the rst two periods
7
i.e. from t = 0 to t = 4. In these graphs the parts
of rapid decrease in energy correspond to high gradients in the solution passing through
the right boundary x = , see Figure 10.1.
8
Similarly, the inexion points at t = in
Figures 10.4-10.5 correspond to the crest of the wave arriving at the right boundary.
Test 2: Take = 1, = 0. Due to the larger value of which implies a stronger
linear disturbance the errors are expected to be larger than in test 1. This is conrmed
by the graphs in Figures 10.6-10.7, which show the L
2
-norm and the energy norm of the
errors in the solutions obtained with congurations 1 and 2.
For 0 t 3 the errors obtained with the rst conguration are somewhat larger
than the errors obtained with conguration 2. For t > 3 there is a clear advantage of
the rst method over the second. Apparently, is suciently large now to make the
correction velocity c

(u) in (10.88) eective. After two periods the energy norms of the
errors are 0.01 for conguration 1 and 0.03 for conguration 2. Figures 10.8-10.9 show
the energy and the momentum of the computed solutions in all three congurations.
Note that there are two inexion points in these graphs now, indicating two successive
wave crests travelling through the open boundary.
Test 3: Take = 5, = 0. Figures 10.10- 10.11 show a great advantage of the
new radiation boundary condition over Sommerfelds radiation condition. After two
7
Note from Figures 10.4-10.5 that E(u) M (u); it can easily be shown that with the present
initial conditions in the case where = = 0 the equality E (u) = M (u) holds.
8
A close look at Figure 10.1 reveals a small wave disturbance travelling to the left.
10.3. APPLICATION TO ONE-DIMENSIONAL WAVE EQUATIONS 207
periods the L
2
-norms of the errors obtained with congurations 1 and 2 are 0.13 and 0.32
respectively. The energy norms of the errors are 0.10 for conguration 1 and 0.45 for
conguration 2.
It is clear that the new radiation boundary condition provides better results for larger
values of . As mentioned before, this is due to the correction factor c

(u) in (10.88).
Figures 10.12-10.13 show the energy and momentum of the solutions; note that the graphs
representing congurations 1 and 3 nearly coincide.
From the above three tests it is clear that the new radiation condition (10.88) pro-
vides much better results than Sommerfelds condition (10.84), especially for large val-
ues of . As mentioned earlier, this is due to the eectiveness of the correction fac-
tor c

(u) in (10.86). In the last two tests we shall investigate the inuence of a nonlinear
disturbance in the eld equation on the performance of boundary conditions (10.88)
and (10.84).
Test 4: Take = 1, = 0.5. Since is nonzero, the eld equation (10.78) is nonlinear.
The results obtained with conguration 1 are better than those obtained with cong-
uration 2, as can be seen in Figures 10.14 and 10.15. The L
2
-norms after two periods
are 0.21 for conguration 1 and 0.30 for conguration 2. The maximum values of the en-
ergy norms are 0.02 for conguration 1 and 0.03 for conguration 2. Figures 10.16-10.17
present the energy and momentum of the computed solutions.
Note that for both congurations the errors are but slightly larger than in test 2
( = 1.0 , = 0), confer Figures 10.6-10.7. This indicates a dominant linear term u
in the disturbance in the eld equation for this combination of and . Therefore, we
shall increase the impact of the nonlinear term u
3
by raising the value of in the next
test.
Test 5: Take = 1, = 2. The expected growth in the errors (see test 4) can only
be found in Figure 10.19, which shows the energy norm of the errors. The L
2
-norm
of the error is (for both congurations) of the same order as in test 2 ( = 0). This
indicates a stronger variation in the derivatives of the errors due to the stronger nonlinear
disturbance, see (10.91).
For this combination of and conguration 1 provides better results than cong-
uration 2 (see Figures 10.18 and 10.19). The L
2
-norms of the errors after two periods
are 0.17 for conguration 1 and 0.19 for conguration 2. The average values however
show better results for conguration 1 than for conguration 2. The energy norms after
two periods are 0.04 for conguration 1 and 0.09 for conguration 2. Figures 10.20-10.21
show the energy and momentum of the solutions.
208 CHAPTER 10. RADIATION BOUNDARY CONDITIONS
Figure 10.1: Test 1: = 0.1, = 0. Solution u in conguration 1.
10.3. APPLICATION TO ONE-DIMENSIONAL WAVE EQUATIONS 209
Figure 10.2: Test 1: = 0.1, = 0. Euclidean norm of error in u.
dashed line = Sommerfeld condition (10.84)
dotted line = energy-transmitting condition (10.88)
210 CHAPTER 10. RADIATION BOUNDARY CONDITIONS
Figure 10.3: Test 1: = 0.1, = 0. Energy norm of error in u.
solid line = extended domain solution
dashed line = Sommerfeld condition (10.84)
dotted line = energy-transmitting condition (10.88)
10.3. APPLICATION TO ONE-DIMENSIONAL WAVE EQUATIONS 211
Figure 10.4: Test 1: = 0.1, = 0. Energy of u.
dashed line = Sommerfeld condition (10.84)
dotted line = energy-transmitting condition (10.88)
212 CHAPTER 10. RADIATION BOUNDARY CONDITIONS
Figure 10.5: Test 1: = 0.1, = 0. Momentum of u.
solid line = extended domain solution
dashed line = Sommerfeld condition (10.84)
dotted line = energy-transmitting condition (10.88)
10.3. APPLICATION TO ONE-DIMENSIONAL WAVE EQUATIONS 213
Figure 10.6: Test 2: = 1, = 0. Euclidean norm of error in u.
dashed line = Sommerfeld condition (10.84)
dotted line = energy-transmitting condition (10.88)
214 CHAPTER 10. RADIATION BOUNDARY CONDITIONS
Figure 10.7: Test 2: = 1, = 0. Energy norm of error in u.
solid line = extended domain solution
dashed line = Sommerfeld condition (10.84)
dotted line = energy-transmitting condition (10.88)
10.3. APPLICATION TO ONE-DIMENSIONAL WAVE EQUATIONS 215
Figure 10.8: Test 2: = 1, = 0. Energy of u.
dashed line = Sommerfeld condition (10.84)
dotted line = energy-transmitting condition (10.88)
216 CHAPTER 10. RADIATION BOUNDARY CONDITIONS
Figure 10.9: Test 2: = 1, = 0. Momentum of u.
solid line = extended domain solution
dashed line = Sommerfeld condition (10.84)
dotted line = energy-transmitting condition (10.88)
10.3. APPLICATION TO ONE-DIMENSIONAL WAVE EQUATIONS 217
Figure 10.10: Test 3: = 5, = 0. Euclidean norm of error in u.
dashed line = Sommerfeld condition (10.84)
dotted line = energy-transmitting condition (10.88)
218 CHAPTER 10. RADIATION BOUNDARY CONDITIONS
Figure 10.11: Test 3: = 5, = 0. Energy norm of error in u.
solid line = extended domain solution
dashed line = Sommerfeld condition (10.84)
dotted line = energy-transmitting condition (10.88)
10.4 Application to the water-wave problem
In this section we apply the above theory for radiation boundary conditions to the three-
dimensional problem of nonlinear water waves travelling over a bottom.
In the numerical simulation of free surface waves under gravity, the quest for eective
radiation boundary conditions has not been completed to full satisfaction so far. In order
to obtain a good approximation of the moving free surface, relatively dense grids have
to be used in general; consequently, the number of data to be stored is extremely high.
Then, given the limited amount of available computer memory, it is clear that the uid
domain must be cut o by an articial boundary. With an appropriate implementation
of suitable radiation conditions on these boundaries, any outgoing wave will be fully
transmitted; there will be no reection at all, and the solution in the area of interest is
not spoiled.
The majority of the radiation conditions proposed hitherto are based on approxima-
tive models, such as linear outgoing wave elds. In practical calculations however, the
outgoing waves are generated by diraction round a oating structure or by radiation
from an oscillating body. As a consequence, these waves generally consist of an innite
number of components interacting in a nonlinear fashion.
Here radiation boundary conditions are derived that simulate a moving uid domain
for which the total energy, or the total horizontal momentum, is conserved. This means
that the greater part of the energy of the outgoing waves will be transmitted, and hence
the reected wave will be negligible in the sense that its energy is small. In order to
10.4. APPLICATION TO THE WATER-WAVE PROBLEM 219
Figure 10.12: Test 3: = 5, = 0. Energy of u.
demonstrate the signicance of a proper variational principle for the problem under
consideration, two dierent approaches will be followed to arrive at such conditions. The
rst approach is based on Lukes (1967) variation principle and the related Hamiltonian
formulation for the water wave problem described in sections 8.2 and 8.3. The second
approach starts from the Eulerian description for an incompressible, inviscid uid and an
irrotational ow, without emphasis on the presence of a free surface. Both routes lead to
radiation boundary conditions which are directly accessible to physical interpretations,
but it will appear that only the rst approach yields signicant boundary conditions for
free surface ow.
10.4.1 The Lagrangian-Hamiltonian approach
In chapter 8 it has been demonstrated that a system of water waves travelling over
an uneven bottom can be described by surface variables only. With the free surface
elevation in the role of canonical coordinate, and the restriction = []
z=
of the
velocity potential to the free surface as the canonical conjugate momentum, it has been
shown that the nonlinear free surface conditions describe a so-called Hamiltonian system;
it has been proven that the canonical equations, involving the energy as Hamiltonian
density, are equivalent with the nonlinear free surface conditions.
Thus we have the composed variable (, ) characterizing the system of water waves
under gravity; in the three-dimensional case, and are both functions of the horizontal
coordinate vector x = (x, y)
T
and the time t. The action integral is given by
J (, ) =

(t)

L(, ) ddt , (10.96)


where the two-dimensional spatial domain (t) consists of all admissible horizontal po-
220 CHAPTER 10. RADIATION BOUNDARY CONDITIONS
Figure 10.13: Test 3: = 5, = 0. Momentum of u.
sition vectors x.
In chapter 8 it was also shown that the evolution equations for water waves can be
derived from a variational principle, with the Lagrangian density L chosen as
L(, ) =
(x;t)

h(x)
p dz =
(x;t)

h(x)

t
+
1
2
( ) +gz

dz , (10.97)
where p denotes the Bernoulli pressure; the water density is taken as unity.
Further, it is assumed that satises the following boundary value problem:

2
= 0 for h(x) < z < (x; t) , (10.98)

n
= 0 at z = h(x) , (10.99)
(x, z; t) = (x; t) at z = (x; t) . (10.100)
Formally, the Frechet derivative of L with respect to a variation is dened as
D

L(, ) =

d
d
L( + , )

=0
. (10.101)
So, let vary such that its variation corresponds to a change in the free surface
potential . Then (10.101) can be worked out to
D

L(, ) =
(x;t)

h(x)
(
t
+ ) dz
10.4. APPLICATION TO THE WATER-WAVE PROBLEM 221
Figure 10.14: Test 4: = 1, = 0.5. Euclidean norm of error in u.
=
t
(x;t)

h(x)
dz
x
(x;t)

h(x)

x
dz
+
(x;t)

h(x)

2
x
+
2
z

dz
+ [(
t
+
x

x

z
) ]
z=
+ [(
x
h
x
+
z
) ]
z=h
, (10.102)
where
x
= (
x
,
y
)
T
is the two-dimensional restriction of the gradient operator =
(
x
,
y
,
z
)
T
.
Using (10.98-10.99), the following expression for the Frechet derivative of L with
respect to a variation in is obtained:
D

L(, ) =

L(, )
+
t
B
0

(, ; ) + div

B
1

(, ; ) , (10.103)
where the variational derivative of L with respect to is given by

L(, ) = [
t
+
x

x

z
]
z=
, (10.104)
and the corresponding boundary expressions read
B
0

(, ; ) =
(x;t)

h(x)
dz , (10.105)
222 CHAPTER 10. RADIATION BOUNDARY CONDITIONS
Figure 10.15: Test 4: = 1, = 0.5. Energy norm of error in u.

B
1

(, ; ) =
(x;t)

h(x)

x
dz . (10.106)
The formal denition of the Frechet derivative of L with respect to a variation in reads
D

L(, ) =

d
d
L(, + )

=0
. (10.107)
However, a variation leads to a change in the free surface potential . In rst order,
this change is given by see also (8.45-8.46)

= [
z
]
z=
. (10.108)
Taking into account the eect of on , the correct expression for the Frechet-derivative
of L with respect to a variation in is
D

L(, ) =

L(, )
+
t
B
0

(, ; ) + div

B
1

(, ; ) , (10.109)
where the variational derivative of L with respect to is given by

L(, ) =

t
+
1
2
( ) +gz

z=

L(, ) [
z
]
z=
, (10.110)
and the corresponding boundary expressions read
B
0

(, ; ) = 0 , (10.111)

B
1

(, ; ) =

0 . (10.112)
10.4. APPLICATION TO THE WATER-WAVE PROBLEM 223
Figure 10.16: Test 4: = 1, = 0.5. Energy of u.
With these results, it is straightforward to show that the Euler-Lagrange equation (10.8)
yields

t
+
x

x

z
= 0 at z = (x; t) , (10.113)

t
+
1
2
( ) +gz = 0 at z = (x; t) . (10.114)
The so-called natural boundary condition (10.9) gives
(x;t)

h(x)
(
x
n v
n
) dz = 0 , (10.115)
where n is the unit normal vector on (t) and v
n
= vn is the normal velocity of (t).
Note that n is independent of the vertical position, whereas the boundary velocity v may
depend on z.
Since (10.115) must hold for arbitrary , and thus for arbitrary , the natural
boundary condition reads
(
x
v) n = 0 for h(x) < z < (x; t) . (10.116)
To obtain an additional, density conserving, boundary condition, it is worthwhile to have
a closer look at the Lagrangian density, dened in (10.97). It is then observed that in
case of a horizontal bottom z = h, L does not depend explicitly on the horizontal
position x and the time t, and that L depends on derivatives of u = (, ) up to rst
order only. As a consequence, the results of example 2 apply, and hence it is concluded
that a necessary and sucient condition for conservation of energy and for conservation
224 CHAPTER 10. RADIATION BOUNDARY CONDITIONS
Figure 10.17: Test 4: = 1, = 0.5. Momentum of u.
of momentum in any horizontal direction is given by
L(, ) =
(x;t)

h(x)

t
+
1
2
( ) +gz

dz = 0 on , (10.117)
stating that for all x on the boundary of (t), the integral of the Bernoulli pressure from
the bottom to the free surface must equal zero.
Remark 5: Since the Lagrangian density L(, ) does not depend explicitly on t, the
energy density can be obtained through Noethers theorem from the variations see
example 2
=
t
, =
t
. (10.118)
The corresponding densities are given by see (10.43-10.44)

0
(, ) = L(, ) =
(x;t)

h(x)

t
+
1
2
( ) +gz

dz , (10.119)

1
(, ) =

0 . (10.120)
With the boundary expressions see (10.105-10.106) and (10.111-10.112)
B
0
(, ) =
(x;t)

h(x)

t
dz , (10.121)
10.4. APPLICATION TO THE WATER-WAVE PROBLEM 225
Figure 10.18: Test 5: = 1, = 2. Euclidean norm of error in u.

B
1
(, ) =
(x;t)

h(x)

x
dz , (10.122)
the density e =
0
B
0
and the corresponding ux density

f
e
=
1


B
1
see (10.20-
10.21) emerge as
e (, ) =
(x;t)

h(x)
1
2
( ) dz
1
2
g

2
h
2

, (10.123)

f
e
(, ) =
(x;t)

h(x)

x
dz . (10.124)
Note that apart from the minus sign e (, ) equals the energy density indeed.
Then, for solutions (, ) to the Euler-Lagrange equations (10.113-10.114) the follow-
ing local energy conservation law holds:

t
e +
x

f
e
= 0 . (10.125)
The proof directly follows from substitution of (10.123-10.124).
Remark 6: The above natural and additional conditions for conservation of energy and
horizontal momentum are related to some results obtained in chapter 9, in particular
identities (9.64) and (9.67), repeated here for convenience:
d1
f
dt
=

C
p dz , (10.126)
226 CHAPTER 10. RADIATION BOUNDARY CONDITIONS
Figure 10.19: Test 5: = 1, = 2. Energy norm of error in u.
dH
f
dt
=

C
p (
x
dz
z
dx) . (10.127)
Here 1
f
and H
f
denote the horizontal uid momentum and the uid energy respectively.
Integration is over the contour C bounding the uid domain; with p naturally vanishing
on the free surface, and with dz = 0 on a horizontal bottom, the necessary and sucient
condition for conservation of horizontal momentum is then found to be

h
p dz = 0 , (10.128)
at the vertical boundaries.
Similarly, with
x
= U and dx = 0 on the vertical boundaries, it is found that a
necessary and sucient condition for conservation of energy is
U

h
p dz = 0 , (10.129)
i.e. either U = 0, in which case we have xed vertical boundaries, or the integrated
pressure equals zero. Hence, both horizontal momentum and energy are conserved if the
additional boundary condition (10.117) is satised.
10.4.2 The Eulerian approach
The previous expressions for the Euler-Lagrange equations, the natural boundary condi-
tions, and the additional boundary conditions for conservation of energy (or horizontal
10.4. APPLICATION TO THE WATER-WAVE PROBLEM 227
Figure 10.20: Test 5: = 1, = 2. Energy of u.
momentum) were obtained at the outset of a velocity potential satisfying the bound-
ary value problem (10.98-10.100). This approach is along the lines of the Hamiltonian
formulation in chapter 8, where it is shown that the system of water waves is governed
by the free surface elevation and the free surface potential .
Perhaps a more obvious and simple approach would be to consider the classical water-
wave problem as a potential ow problem, without emphasizing the presence of a free
surface. From this point of view, the system is characterized by the velocity potential ,
dened throughout the transient three-dimensional uid domain (t). The bounding
surface (t) consists of the free surface, the bottom, and lateral boundaries which are
either physical or articial. The proposed variational principle is
J () =

(t)

L() ddt = 0 , (10.130)


where the Lagrangian density equals the Bernoulli pressure:
L() = p =

t
+
1
2
( ) +gz

. (10.131)
The Frechet derivative of L with respect to a variation is
DL() =
t

=
t
() +
2

= L() +
t
B
0
(; ) + div

B
1
(; ) . (10.132)
Hence, the Euler-Lagrange equation becomes Laplaces equation:
L() =
2
= 0 , (10.133)
228 CHAPTER 10. RADIATION BOUNDARY CONDITIONS
Figure 10.21: Test 5: = 1, = 2. Momentum of u.
and the natural boundary condition states that the normal velocity of each boundary
particle coincides with the local boundary velocity in normal direction:

n
= v
n
. (10.134)
For an impermeable bottom z = h(x, y) we have v
n
= 0, and (10.134) reduces to
the zero-ux condition (10.99). For the free surface z = (x, y; t), condition (10.134)
becomes (10.113), so that in this formulation the kinematic free surface condition
plays the role of a natural boundary condition, while in the Lagrangian-Hamiltonian
formulation it was part of the Euler-Lagrange equations.
Note that the second (dynamic) free surface condition (10.114) has not been obtained
from (10.130-10.131) so far. In this respect, the present variation principle does not
provide the complete set of governing equations for nonlinear free surface ow under the
action of gravity. It does provide, however, the eld equation for bounded or unbounded
potential ow for instance, the ow between two parallel horizontal plates.
Since the Lagrangian density L does not depend explicitly on the time t, conservation of
energy is ensured by
L() v
n
= p v
n
= 0 . (10.135)
At the bottom we have v
n
= 0, so the additional condition is automatically satised
there. At a moving free surface, we have v
n
= 0; hence, the additional condition yields
p = 0, which is exactly the dynamic condition (10.114).
Since the Lagrangian density does not depend explicitly on the horizontal coordinates
x and y, conservation of momentum in some horizontal direction = (
x
,
y
, 0)
T
is
provided by
L() ( n) = p ( n) = 0 . (10.136)
10.4. APPLICATION TO THE WATER-WAVE PROBLEM 229
Note that this condition is automatically satised at an even (horizontal) bottom, while
in the presence of an uneven bottom the horizontal momentum will not be preserved. At
the free surface the additional condition reduces to the dynamic condition (10.114).
With respect to the lateral boundaries, we arrive at the vanishing of the Bernoulli pres-
sure as a sucient condition for conservation of energy and horizontal momentum; this is
the point-wise version of condition (10.117) obtained with the Lagrangian-Hamiltonian
approach. This local condition, however, is insignicant from a physical point of view,
since any lateral outow boundary connects the free surface, where p = 0, with the
bottom, where p > 0.
This example shows the necessity of an appropriate variation principle for the problem
under consideration, in the application of our theory for radiation boundary conditions.
230 CHAPTER 10. RADIATION BOUNDARY CONDITIONS
10.5 Bibliography
Bayliss, A., and Turkel, E. 1980. Radiation boundary conditions for wave-like
equations. Communications on Pure and Applied Mathematics 33:707-725.
Bayliss, A., and Turkel, E. 1982. Far eld boundary conditions for compressible
ows. Journal of Computational Physics 48:182-199.
Biot, M.A. 1957. General theorem on the equivalence of group velocity and energy
transport. Physical Reviews 105:1129-1137.
Broer, L.J.F. 1951. On the propagation of energy in linear conservative waves. Ap-
plied Scientic Research, Series A, 329-344.
Broeze, J., and van Daalen, E.F.G. 1992. Radiation boundary conditions for the
two-dimensional wave equation from a variational principle. Mathematics of Computa-
tion 58(197):73-82.
van Daalen, E.F.G., Broeze, J., and van Groesen, E.W.C. 1992. Variational
methods and conservation laws in the derivation of radiation boundary conditions for
wave equations. Mathematics of Computation 58(197):55-71.
Engquist, B., and Halpern, L. 1988. Far eld boundary conditions for computation
over long time. Applied Numerical Mathematics 4:21-45.
Engquist, B., and Majda, A. 1977. Absorbing boundary conditions for the numerical
simulation of waves. Mathematics of Computation 31(139):629-651.
Engquist, B., and Majda, A. 1979. Radiation boundary conditions for acoustic and
elastic wave calculations. Communications on Pure and Applied Mathematics 32:313-
357.
Givoli, D. 1991. Non-reecting boundary conditions. Journal of Computational Physics
94(1):1-29.
Goldstein, H. 1980. Classical Mechanics. Addison-Wesley.
van Groesen, E.W.C. 1980. Unidirectional wave propagation in one dimensional, rst
order Hamiltonian systems. Journal of Mathematical Physics 21:1646-1655.
van Groesen, E.W.C., and Mainardi, F. 1990. Balance laws and centro velocity
in dissipative systems. Journal of Mathematical Physics 31(9):2136-2140.
Higdon, R.L. 1986. Absorbing boundary conditions for dierence approximations to the
multi-dimensional wave equation. Mathematics of Computation 47(176):437-459.
Higdon, R.L. 1987. Numerical absorbing boundary conditions for the wave equation.
Mathematics of Computation 49(179):65-90.
Higdon, R.L. 1990. Radiation boundary conditions for elastic wave propagation. SIAM
Journal on Numerical Analysis 27(4):831-870.
Lighthill, M.J. 1965. Group velocity. J. Inst. Math. Appl. 1:1-28.
Noether, E. 1918. Invariante Variationsprobleme. Nachrichten der Gesellschaft der
Wissenschaften Gottingen 2:235-257.
Olver, P.J. 1986. Application of Lie Groups to Dierential Equations. Springer-Verlag.
Orlanski, I. 1976. A simple boundary condition for unbounded hyperbolic ows. Jour-
nal of Computational Physics 21:251-269.
10.5. BIBLIOGRAPHY 231
Wehausen, J.V., and Laitone, E.V. 1960. Surface Waves. Volume IX of Encyclo-
pedia of Physics. Springer-Verlag.
Whitham, G.B. 1970. Two-timing, variational principles and waves. Journal of Fluid
Mechanics 44(2):373-395.
Whitham, G.B. 1974. Linear and Nonlinear Waves. Wiley-Interscience.
232 CHAPTER 10. RADIATION BOUNDARY CONDITIONS
. . . And the god
9
rose up from the depths in form such as he really was. And
as when a man trains a swift steed for the broad race-course, and runs along,
grasping the bushy mane, while the steed follows obeying his master, and rears
his neck aloft in his pride, and the gleaming bit rings loud as he champs it
in his jaws from side to side; so the god, seizing hollow Argos keel, guided
her onward to the sea. And his body, from the crown of his head, round his
back and waist as far as the belly, was wondrously like that of the blessed ones
in form; but below his sides the tail of a sea monster lengthened far, forking
to this side and that; and he smote the surface of the waves with the spines,
which below parted into curving ns, like the horns of the new moon. And
he guided Argo on until he sped her into the sea on her course; and quickly
he plunged into the vast abyss; and the heroes shouted when they gazed with
their eyes on that dread portent. There is the harbour of Argo and there are
the signs of her stay, and altars to Poseidon and Triton; for during that day
they tarried . . .
Be gracious, race of blessed chieftains! And may these songs year after
year be sweeter to sing among men. For now have I come to the glorious
end of your toils; for no adventure befell you as ye came home from Aegina,
and no tempest of winds opposed you; but quietly did ye skirt the Cecropian
land and Aulis inside of Euboea and the Opuntian cities of the Locrians, and
gladly did ye step forth upon the beach of Pagasae.
Argonautica, Book IV, Verses 1602-1622 and 1773-1781.
9
i.e. Triton
Chapter 11
Conclusions and
Recommendations
In parts I and II of this thesis we have described the development of a numerical algorithm
for the time domain simulation of nonlinear wave-body interactions. This method is based
on Romates higher order panel method
1
for nonlinear free surface wave simulations.
The theory of nonlinear waves interacting with rigid bodies at zero forward speed
that is, both the mathematical statement of the problem and the formulation in
terms of boundary integral equations was treated in chapters 2-3. An outline of the
numerical algorithm was presented in chapter 4. The most important feature of this
panel method is that the dynamic equilibrium of the waves and the body is preserved in
the time marching, by solving Laplaces equation both for the velocity potential and for
its partial time derivative.
Numerical test results for a wide range of applications were presented in chapters 5-7.
The main conclusion is that the present panel method is very accurate in view of
the excellent agreement of the numerical results with the analytical predictions for an
impulsively started vertical wavemaker, see chapter 5 and very robust in view of
the results on cylinders in free heaving and rolling motion, see chapter 7. Fair agreement
with linear theory was observed for the hydrodynamic mass and damping coecients of
circular cylinders in forced sinusoidal heaving and swaying motion.
With respect to the further development of this panel method and the future
extension to three dimensions we recommend an extensive validation of the two-
dimensional method. In particular, the response of cylinders on incoming wave elds
including the computation of (mean) drift forces and an elaborated study of added
mass and damping for various cross sections in all three modes of motion deserve
much more attention than we have been able to give so far. Furthermore, this nonlinear
model enables the investigation of the behaviour of hydrodynamic mass and damping for
large amplitude motion, and of nonlinear eects for cross sections of special shape.
Part III of this thesis is concerned with the use of variation principles and Hamiltonian
formulations in problems involving water waves and oating bodies.
In chapter 8 we have presented a variation principle and a Hamiltonian formulation for
1
Romate, J.E. 1989. The Numerical Simulation of Nonlinear Gravity Waves in Three Dimensions
using a Higher Order Panel Method. Ph.D. thesis, University of Twente. Enschede, The Netherlands.
233
234 CHAPTER 11. CONCLUSIONS AND RECOMMENDATIONS
the nonlinear wave-body problem, using the Legendre transformation as a tool to transit
from the rst description to the latter. The main advantage of these formulations is that
they clearly describe the physical mechanisms that are active in wave-body systems, such
as the transfer of momentum and energy.
Based on earlier results for water waves only, a full account of symmetries and the
corresponding conservation laws for the wave-body problem was given in chapter 9. It
appears that all constants of the motion but one are preserved when a surface piercing
body is introduced. The remaining water-wave invariant and the possibility of its gen-
eralization to the wave-body problem was discussed in detail. The validity of the theory
was supported with numerical results obtained with our panel method for a simple
wave-body conguration.
Finally, in chapter 10 we demonstrated the optimal use of variation principles in
the quest for radiation boundary conditions for general wave-like problems. Based on
the conservation of a chosen density for instance, the energy density transmitting
boundary conditions are obtained that simulate a moving domain for which the integrated
density is conserved. An application to a nonlinear one-dimensional wave equation was
given, and the corresponding numerical results indicate that these density-conserving
boundary conditions are superior to standard type boundary conditions, such as Som-
merfeld/Orlanski conditions. The theory was also applied to the classical water wave
problem, and some of the ndings in the related chapters 8 and 9 were re-established.
For the near future we recommend the stable implementation of these radiation boundary
conditions in our panel method for nonlinear free surface waves and ship motions.
Part IV
Appendices
235
Appendix A
An Expression for
tn
on the
Body
In this appendix we shall derive an expression for

tn
=

2

tn
=
t
n on S (A.1)
in terms of the body velocities and accelerations, tangential derivatives of up to second
order, and rst order tangential derivatives of
n
. The strategy is to derive an expression
for the normal component of the acceleration of a point on the wetted body surface S,
since this variable is directly related to
tn
.
Figure A.1: Denition of three-dimensional Cartesian and orthogonal curvilinear coor-
dinate systems.
237
238 APPENDIX A. AN EXPRESSION FOR
TN
ON THE BODY
The description is in terms of a Cartesian coordinate system x
1
, x
2
, x
3
, where x
1
and
x
2
denote the horizontal directions and x
3
denotes the vertical direction, and in terms
of an orthogonal curvilinear coordinate system s
1
, s
2
, s
3
which has its origin on S; we
choose s
1
and s
2
to be the tangential coordinates (i.e. the coordinates on S) and s
3
= n
as the normal coordinate, see Figure A.1. If x is the position of a point on S, its local
acceleration reads

x =
d
dt

=
d
dt

k=1

x s
k

s
k

=
3

k=1

d
dt

x s
k

s
k
+

x s
k

d
dt
[s
k
]

. (A.2)
Using the orthogonality of s
1
, s
2
, s
3
, the following expressions for the material deriva-
tives of s
1
, s
2
and s
3
can be derived:
d
dt
[s
1
] =

G
s
1
=

G
(s
2
s
3
) =

G
s
3

s
2

G
s
2

s
3
, (A.3)
d
dt
[s
2
] =

G
s
2
=

G
(s
3
s
1
) =

G
s
1

s
3

G
s
3

s
1
, (A.4)
d
dt
[s
3
] =

G
s
3
=

G
(s
1
s
2
) =

G
s
2

s
1

G
s
1

s
2
, (A.5)
where

G
is the body rotation vector with respect to the principal axes, corresponding
to roll, pitch and yaw motions respectively.
Substitution of (A.3-A.5) into (A.2) yields

x =

d
dt

x s
1

G
s
2

x s
3

G
s
3

x s
2

s
1
+

d
dt

x s
2

G
s
3

x s
1

G
s
1

x s
3

s
2
+

d
dt

x s
3

G
s
1

x s
2

G
s
2

x s
1

s
3
. (A.6)
Taking the inner product with the unit normal vector n = s
3
gives the normal component
of the local acceleration:

x s
3
=
d
dt

x s
3

G
s
1

x s
2

G
s
2

x s
1

. (A.7)
Substituting in terms of curvilinear coordinates for

x, the rst term on the right-hand
side of (A.7) reads:
d
dt

x s
3

=
d
dt
[ s
3
] =
d
dt

k=1
1
h
k

s
k
s
k

s
3

, (A.8)
where the scale factors h
k
arise from the transformation from Cartesian to curvilinear
coordinates.
Next, equation (A.8) is elaborated to
d
dt

x s
3

=
d
dt

k=1
1
h
k

s
k
s
k

s
3
+

k=1
1
h
k

s
k
s
k

d
dt
[s
3
] . (A.9)
239
Then, using (A.5), it is found that
d
dt

x s
3

=
d
dt

k=1
1
h
k

s
k
s
k

s
3
+
1
h
1

s
1

G
s
2

1
h
2

s
2

G
s
1

. (A.10)
Substitution of (A.10) into (A.7) gives the rst interim expression for the local accelera-
tion in normal direction:

x s
3
=
d
dt

1
h
1

s
1
s
1

s
3
+
d
dt

1
h
2

s
2
s
2

s
3
+
d
dt

1
h
3

s
3
s
3

s
3
+

G
s
1

x s
2

1
h
2

s
2

G
s
2

x s
1

1
h
1

s
1

. (A.11)
Attention is now focused on the rst three terms on the right-hand side of (A.11); applying
the denition of the material derivative of a scalar function f in x, that is
d
dt
[f] =

t
[f] +[f]

x =

t
[f] +
3

k=1
1
h
k

x s
k


s
k
[f] , (A.12)
and using the equations for the derivatives of the unit base vectors with respect to the
curvilinear coordinates see, for instance, Malvern (1969)
s
m
s
m
=
1
h
n
h
m
s
n
s
n

1
h
r
h
m
s
r
s
r
for m, n, r all dierent , (A.13)
s
m
s
n
= +
1
h
m
h
n
s
m
s
n
for m = n , (A.14)
the rst vector in the rst term on the right-hand side of (A.11) is rewritten to
d
dt

1
h
1

s
1
s
1

1
h
1

ts
1
+
3

k=1
1
h
k

x s
k


s
k

1
h
1

s
1

s
1
+

1
h
2
1
h
2

s
1

x s
2

h
2
s
1

x s
1

h
1
s
2

s
2
+

1
h
2
1
h
3

s
1

x s
3

h
3
s
1

x s
1

h
1
s
3

s
3
. (A.15)
It then follows that the rst term on the right-hand side of (A.11) gives
d
dt

1
h
1

s
1
s
1

s
3
=
1
h
2
1
h
3

s
1

x s
3

h
3
s
1

x s
1

h
1
s
3

. (A.16)
Similarly, it is shown that the second term yields
d
dt

1
h
2

s
2
s
2

s
3
=
1
h
2
2
h
3

s
2

x s
3

h
3
s
2

x s
2

h
2
s
3

, (A.17)
240 APPENDIX A. AN EXPRESSION FOR
TN
ON THE BODY
and the third term
d
dt

1
h
3

s
3
s
3

s
3
=
1
h
3

ts
3
+
3

k=1
1
h
k

x s
k


s
k

1
h
3

s
3

. (A.18)
Substitution of (A.16-A.18) into (A.11) yields the second interim expression for the local
acceleration in normal direction:

x s
3
=
1
h
2
1
h
3

s
1

x s
3

h
3
s
1

x s
1

h
1
s
3

+
1
h
2
2
h
3

s
2

x s
3

h
3
s
2

x s
2

h
2
s
3

+
1
h
3

ts
3
+
3

k=1
1
h
k

x s
k


s
k

1
h
3

s
3

G
s
1

x s
2

1
h
2

s
2

G
s
2

x s
1

1
h
1

s
1

. (A.19)
In general, curvilinear coordinates s
k
are dened in terms of rectangular Cartesian coor-
dinates x
m
by functional equations of the form
s
k
= s
k
(x
1
, x
2
, x
3
) (A.20)
Reversely, the Cartesian coordinates can formally be written as functions of the curvi-
linear coordinates:
x
m
= x
m
(s
1
, s
2
, s
3
) (A.21)
The scale factors h
k
are then dened by
h
2
k
=
3

m=1

x
m
s
k

2
, k = 1, 2, 3 (A.22)
With x restricted to S, it follows that
x = x(s
1
, s
2
) , (A.23)
since s
1
and s
2
are the tangential coordinates on S.
From (A.22) and (A.23) it follows that h
3
is constant along S, and therefore
h
3
s
1
=
h
3
s
2
= 0 . (A.24)
Contrary to h
3
, the scale factors h
1
and h
2
are not constant, since S is curved in general.
The radii of curvature are dened as
1
R
1
=
1
h
1
h
1
s
3
,
1
R
2
=
1
h
2
h
2
s
3
(A.25)
241
With these simplications and denitions the third and nal expression for the
local acceleration in normal direction is obtained:

x s
3
=

2

ts
3
+

x s
1

s
1
s
3
+

x s
2

s
2
s
3
+

x s
3

s
2
3
+

G
s
1

x s
2


s
2

G
s
2

x s
1


s
1

+
1
R
1

s
1

x s
1

+
1
R
2

s
2

x s
2

. (A.26)
Using Laplaces equation in terms of the curvilinear coordinates, that is

2
= h
1
h
2
h
3


s
1

h
1
h
2
h
3

s
1

+

s
2

h
2
h
1
h
3

s
2

+

s
3

h
3
h
1
h
2

s
3

,
(A.27)
the term
2
/s
2
3
can be eliminated from (A.26). Rewriting this equation then yields
the following expression for
tn
on S:

tn
=

x n
+

G
s
2

x s
1


s
1

G
s
1

x s
2


s
2

1
R
1

s
1
+

2

s
1
n

x s
1

1
R
2

s
2
+

2

s
2
n

x s
2

s
2
1
+

2

s
2
2

1
R
1
+
1
R
2

x n

(A.28)
where s
3
has been replaced by n.
Bibliography
Malvern, L.E. 1969. Introduction to the Mechanics of a Continuous Medium. Prentice-
Hall.
242 APPENDIX A. AN EXPRESSION FOR
TN
ON THE BODY
Appendix B
A Body Surface Integral
Condition
In this appendix a boundary integral condition
(
t
,
tn
, x) = 0 on S (B.1)
is derived, using the expression for
tn
derived in the previous appendix, and the pressure
integral expressions for the hydrodynamic forces and moments acting on the body. It
will appear that these equalities enable us to transform the hydrodynamic equations
of motion into a boundary integral equation over the wetted part of the body surface
S. Discretization of this boundary integral equation gives a system of linear equations
connecting
t
and
tn
in the N
S
collocation points on S.
Expression (A.28), applied in a point x on S, reads in condensed notation

tn
(x) =

x n(x) + (x) (B.2)
where the rest term (x) is dened as
(x) =

G
s
2
(x)

x s
1
(x)


s
1
(x)

G
s
1
(x)

x s
2
(x)


s
2
(x)

1
R
1
(x)

s
1
(x) +

2

s
1
n
(x)

x s
1
(x)

1
R
2
(x)

s
2
(x) +

2

s
2
n
(x)

x s
2
(x)

s
2
1
(x) +

2

s
2
2
(x)

1
R
1
(x)
+
1
R
2
(x)

n
(x)

x n(x)

.
(B.3)
The acceleration of a point x on S can be expressed in terms of the acceleration of the
centre of mass G and the angular acceleration:

x =

x
G
+

G
r (x) +

G
r (x)

, (B.4)
243
244 APPENDIX B. A BODY SURFACE INTEGRAL CONDITION
where r (x) x x
G
denotes the position of x relative to G.
Substitution of (B.4) into (B.2) gives

tn
(x) =

x
G
n(x) +

G
r (x)

n(x)
+

G
r (x)

r (x)

n(x) + (x) , (B.5)


where we used the identity
a

b c

= (a c)

c . (B.6)
The accelerations

x
G
and

G
are given by the equations of motion:
M

x
G
=

F Mge
3
,

I

G
=

L , (B.7)
and the hydrodynamic (inertial) forces and moments (exerted by the uid) are obtained
from pressure integrations over S:

F =

dS

,

L =

dS

. (B.8)
Substitution of (B.7) and (B.8) into (B.5) yields after a little elementary algebra

tn
(x) =

1
M
p

n(x) n

dS

(r (x) n(x))

dS

G
r (x)

G
n(x)

(r (x) n(x))
+ (x) gn
3
(x) , (B.9)
where
a .

a
1
b
1
,
a
2
b
2
,
a
3
b
3

T
and a

b (a
1
b
1
, a
2
b
2
, a
3
b
3
)
T
(B.10)
dene a component-wise vector division and product respectively. In the derivation
of (B.9) we used the identity
a

b c

c . (B.11)
The pressure along S is obtained from Bernoullis equation
p

+
1
2

+g
3

. (B.12)
Substitution of (B.12) into (B.9) yields a boundary integral equation, connecting
tn
with
t
along S:

tn
(x) +

x,

dS

= (x) (B.13)
245
where the kernel function K

x,

is regular and symmetric, and depends on the geom-


etry of the body only:
K

x,

1
M

n(x) n

+ (r (x) n(x))

r (y) n

(B.14)
The discretized version of (B.13), applied in collocation point i on S, can be written as

i
tn
+
N
S

j=1
C
ij
k

j
t
=
i
(B.15)
where summation is over all N
S
collocation points on S.
The coecients C
ij
k
read
C
ij
k
= S
j

1
M

n
i
n
j

r
i
n
i

r
j
n
j

(B.16)
where S
j
is the area of panel j.
Finally, the right-hand side terms
i
are dened as

i
=

G
r
i

G
n
i

r
i
n
i

+
i
gn
i
3
+
N
S

j=1
C
ij
k

1
2

j

j

+gx
j
3

, (B.17)
with the rest terms
i
given by

i
=

G
s
i
2

x
i
s
i
1

i
s
1

G
s
i
1

x
i
s
i
2

i
s
2

1
R
i
1

i
s
1
+
i
s
1
n

x
i
s
i
1

1
R
i
2

i
s
2
+
i
s
2
n

x
i
s
i
2

i
s
1
s
1
+
i
s
2
s
2

1
R
i
1
+
1
R
i
2

i
n

x
i
n
i

. (B.18)
246 APPENDIX B. A BODY SURFACE INTEGRAL CONDITION
Appendix C
The Angular Body Momentum
In this appendix we derive an expression for the angular body momentum in terms of
the mass
b
, the moment of inertia ^
b
, the position of the centre of gravity ( x
b
, z
b
),
and the angle of roll

b
.
Figure C.1: Two-dimensional wave-body system.
The angular uid momentum is dened as
/
f
=

D
f


f
(x
z
z
x
) dxdz =

D
f


f
(xw
f
zu
f
) dxdz , (C.1)
where
f
is the uid density and (u
f
, w
f
) is the local uid velocity in (x, z).
247
248 APPENDIX C. THE ANGULAR BODY MOMENTUM
The analogue of /
f
for the body is put as
/
b
=

D
b


b
(xw
b
zu
b
) dxdz (C.2)
Here
b
is the body density, and (u
b
, w
b
) denotes the velocity of a point (x, z) in the rigid
body domain D
b
:
u
b
=

x
b

b
r
z
, w
b
=

z
b
+

b
r
x
, (C.3)
where the coordinates relative to G are see Figure C.1
r
x
= x x
b
, r
z
= z z
b
, (C.4)
Substitution of (C.3-C.4) into (C.2) yields
/
b
=

z
b
+

b
x
b

D
b


b
r
x
dxdz

x
b

b
z
b

D
b


b
r
z
dxdz
+

D
b

r
2
x
+r
2
z

dxdz + ( x
b

z
b
z
b

x
b
)

D
b


b
dxdz . (C.5)
From the denition of the centre of gravity ( x
b
, z
b
) it follows that

D
b


b
r
x
dxdz =

D
b


b
r
z
dxdz = 0 . (C.6)
The body mass and moment of inertia are dened as

b
=

D
b


b
dxdz , ^
b
=

D
b

r
2
x
+r
2
z

dxdz (C.7)
Substitution of (C.6-C.7) into (C.5) yields the following expression for the angular body
momentum:
/
b
= ^
b

b
+
b
( x
b

z
b
z
b

x
b
) (C.8)
List of Figures
2.1 Fluid domain and bounding surfaces. . . . . . . . . . . . . . . . . . . . 19
2.2 Free surface with a oating body. . . . . . . . . . . . . . . . . . . . . . . . 23
3.1 Denition of , , and n

. . . . . . . . . . . . . . . . . . . . . . . . . . . 38
3.2 Boundary consisting of two smooth subsurfaces S
1
and S
2
. . . . . . . 44
4.1 Structure of matrix equation. . . . . . . . . . . . . . . . . . . . . . . . . . 54
4.2 Body motion and geometry denitions. . . . . . . . . . . . . . . . . . . . . 55
4.3 Structure of extra matrix equation. . . . . . . . . . . . . . . . . . . . . . . 57
5.1 Impulsive wavemaker problem: geometry denition. . . . . . . . . . . . . 70
5.2 Element distributions for impulsive wavemaker problem. . . . . . . . . . . 76
5.3 Initial potential distribution on the wavemaker. . . . . . . . . . . . . . . . 83
5.4 Initial vertical velocity along the wavemaker. . . . . . . . . . . . . . . . . 84
5.5 Initial potential distribution close to the intersection point. . . . . . . . . 85
5.6 Initial vertical velocity close to the intersection point. . . . . . . . . . . . 86
5.7 Initial pressure distribution on the wavemaker. . . . . . . . . . . . . . . . 87
5.8 Initial potential distribution on the bottom. . . . . . . . . . . . . . . . . . 88
5.9 Initial horizontal velocity along the bottom. . . . . . . . . . . . . . . . . . 89
5.10 Initial vertical velocity of the free surface. . . . . . . . . . . . . . . . . . . 90
5.11 Initial vertical free surface velocity close to the intersection point. . . . . . 91
5.12 Free surface elevation at times t = 0.0025 (0.0025) 0.0200 s. . . . . . . 92
5.13 Free surface elevation at times t = 0.025 (0.025) 0.200 s. . . . . . . . . 93
6.1 Two-dimensional cylinder oating on a free surface. . . . . . . . . . . . . . 98
6.2 Added mass coecients for circular cylinder in heaving motion. . . . . . . 107
6.3 Damping coecients for circular cylinder in heaving motion. . . . . . . . . 108
6.4 Improved added mass coecients for circular cylinder in heaving motion. 109
6.5 Improved damping coecients for circular cylinder in heaving motion. . . 110
6.6 Added mass coecients for circular cylinder in swaying motion. . . . . . . 111
6.7 Damping coecients for circular cylinder in swaying motion. . . . . . . . 112
7.1 Transient heaving motion of a circular cylinder due to moderate initial
displacements. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
7.2 Transient heaving motion of a circular cylinder due to large initial dis-
placements. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122
7.3 Transient heaving motion of a circular cylinder due to initial displacement.
Numerical versus analytical results. . . . . . . . . . . . . . . . . . . . . . . 123
249
7.4 Transient heaving motion of a circular cylinder due to initial unit displace-
ment and initial unit velocity. . . . . . . . . . . . . . . . . . . . . . . . . . 124
7.5 Transient heaving motion of a rectangular cylinder due to initial displace-
ments. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
7.6 Transient heaving motion of a rectangular cylinder due to initial displace-
ment and initial velocity. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126
7.7 Transient rolling motion of a rectangular cylinder due to initial displace-
ments. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
8.1 Free surface potential ow. . . . . . . . . . . . . . . . . . . . . . . . . . . 133
8.2 Wave-body system. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142
9.1 Two-dimensional water waves: the innite depth case. . . . . . . . . . . . 160
9.2 Two-dimensional wave-body system. . . . . . . . . . . . . . . . . . . . . . 164
9.3 Free surface waves and submerged deformable body. . . . . . . . . . . . . 180
9.4 Wave-body system with impermeable xed boundaries. . . . . . . . . . . . 183
9.5 Wave-body system: initial conguration. . . . . . . . . . . . . . . . . . . . 185
9.6 Wave-body system: equilibrium state. . . . . . . . . . . . . . . . . . . . . 185
9.7 Conservation of mass. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 186
9.8 Conservation of horizontal momentum. . . . . . . . . . . . . . . . . . . . . 187
9.9 Exchange of vertical momentum no conservation. . . . . . . . . . . . . 188
9.10 Conservation of angular momentum. . . . . . . . . . . . . . . . . . . . . . 188
9.11 Exchange and conservation of energy. . . . . . . . . . . . . . . . . . . . . . 189
9.12 Virial no conservation. . . . . . . . . . . . . . . . . . . . . . . . . . . . 189
10.1 Test 1: = 0.1, = 0. Solution u in conguration 1. . . . . . . . . . . . . 208
10.2 Test 1: = 0.1, = 0. Euclidean norm of error in u. . . . . . . . . . . . . 209
10.3 Test 1: = 0.1, = 0. Energy norm of error in u. . . . . . . . . . . . . . 210
10.4 Test 1: = 0.1, = 0. Energy of u. . . . . . . . . . . . . . . . . . . . . . 211
10.5 Test 1: = 0.1, = 0. Momentum of u. . . . . . . . . . . . . . . . . . . . 212
10.6 Test 2: = 1, = 0. Euclidean norm of error in u. . . . . . . . . . . . . . 213
10.7 Test 2: = 1, = 0. Energy norm of error in u. . . . . . . . . . . . . . . 214
10.8 Test 2: = 1, = 0. Energy of u. . . . . . . . . . . . . . . . . . . . . . . 215
10.9 Test 2: = 1, = 0. Momentum of u. . . . . . . . . . . . . . . . . . . . . 216
10.10Test 3: = 5, = 0. Euclidean norm of error in u. . . . . . . . . . . . . . 217
10.11Test 3: = 5, = 0. Energy norm of error in u. . . . . . . . . . . . . . . 218
10.12Test 3: = 5, = 0. Energy of u. . . . . . . . . . . . . . . . . . . . . . . 219
10.13Test 3: = 5, = 0. Momentum of u. . . . . . . . . . . . . . . . . . . . . 220
10.14Test 4: = 1, = 0.5. Euclidean norm of error in u. . . . . . . . . . . . . 221
10.15Test 4: = 1, = 0.5. Energy norm of error in u. . . . . . . . . . . . . . 222
10.16Test 4: = 1, = 0.5. Energy of u. . . . . . . . . . . . . . . . . . . . . . 223
10.17Test 4: = 1, = 0.5. Momentum of u. . . . . . . . . . . . . . . . . . . . 224
10.18Test 5: = 1, = 2. Euclidean norm of error in u. . . . . . . . . . . . . . 225
10.19Test 5: = 1, = 2. Energy norm of error in u. . . . . . . . . . . . . . . 226
10.20Test 5: = 1, = 2. Energy of u. . . . . . . . . . . . . . . . . . . . . . . 227
10.21Test 5: = 1, = 2. Momentum of u. . . . . . . . . . . . . . . . . . . . . 228
A.1 Denition of three-dimensional Cartesian and orthogonal curvilinear co-
ordinate systems. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 237
250
C.1 Two-dimensional wave-body system. . . . . . . . . . . . . . . . . . . . . . 247
251
252
List of Tables
2.1 Order estimation of force magnitudes. . . . . . . . . . . . . . . . . . . . . 14
3.1 Uniqueness for solutions of (3.7-3.8) with various boundary conditions. . . 43
3.2 Uniqueness for various surface distributions. . . . . . . . . . . . . . . . . . 43
4.1 Numerical methods for nonlinear ship motion simulations. . . . . . . . . . 52
6.1 Added mass and damping coecients for heaving circular cylinder. . . . . 101
6.2 Added mass and damping coecients for heaving circular cylinder: eects
due to longer wave tank for = 0.452. . . . . . . . . . . . . . . . . . . . . 102
6.3 Added mass and damping coecients for heaving circular cylinder: eects
due to smaller cylinders. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
6.4 Added mass and damping coecients for swaying circular cylinder. . . . . 103
9.1 Interrelation between vector elds, symmetry groups, conserved densities,
and integrals quantities. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182
253
254
Abstract
In ocean engineering many problems involve the analysis of free surface waves interact-
ing with large xed or oating bodies either partially or totally submerged where
surface tension, viscosity, and compressibility eects are of minor importance and may
be neglected. Under these assumptions the uid ow is governed by a velocity potential
which satises the time-independent Laplace equation throughout the uid domain. The
wave evolution is described by the dynamic and kinematic free surface conditions, which
are time-dependent partial dierential equations. The motion of a oating body is de-
scribed by the hydrodynamic equations of motion, involving pressure integrations over
the wetted body surface.
The rst and second parts of this thesis are devoted to the development of a numerical
model for the simulation of these water-wave and wave-body systems. The algorithm
is based on a boundary integral equation method for the spatial discretization, and a
fourth order Runge-Kutta method for the discrete time marching.
First, the formulations of both problems are considered, as well as the transition to
the boundary integral equation formulation through Greens theorem. A brief literature
survey of numerical solution procedures for the nonlinear wave-body problem is presented.
The most important feature of the approach followed here is that the dynamic equilibrium
of the uid and the body is preserved for all times, through the solution of integral
equations for both the velocity potential and its partial time derivative.
Numerical results for various water-wave and wave-body problems show fair to excel-
lent agreement with analytical predictions and experimental measurements. The present
method appears to be quite able to reveal nonlinear eects.
In the third part we discuss the description of water-wave and wave-body problems in
terms of variational principles and Hamiltonian equations of motion. Based on earlier
results for water waves only, it is demonstrated that the equations of motion for free
surface waves interacting with freely oating bodies constitute an innite-dimensional
Hamiltonian system. Explicit expressions for the canonical variables and equations are
presented.
The complete set of constants of the motion for the wave-body problem is given for
both the two-dimensional and the three-dimensional problem; explicit proofs are given,
and this theory is supported with numerical results obtained with the panel method
described above.
Finally, the use of variational principles and conservation laws is demonstrated in
the development of absorbing boundary conditions for general wave-like problems, and
applications to a nonlinear one-dimensional Klein-Gordon equation are discussed. This
theory is also applied to the nonlinear water-wave problem.
255

You might also like