You are on page 1of 486

SEAKEEPING:

Ship Behaviour in Rough Weather

~ rl

ELLIS HORWOOD SERIES IN MARINE TECHNOLOGY


Series Editor: Professor JAMES PAFFETT
Lloyd, A.R.J.M.
Walker, G. & Reader, G.T.

Seakeeping: Ship Behaviour in Rough Weather


Underwater Power Plant

ELLIS HORWOOD SERIES IN MARINE SCIENCE


Series Editor: T. D. ALLAN, Institute of Oceanograpic Sciences, Wormley, Surrey
Allan, T.D.
Bowden, K.F.
Cracknell, A. P.
Robinson, I.S.
Searle, R.C.
Mercer, J.R.
Sleigh, M.A.

Satellite Microwave Remote Sensing


Physical Oceanography of Coastal Waters
Remote Sensing in Meteorology, Oceanography and Hydrology
Satellite Oceanography
New Perspectives in Marine Geology
Marine Corrosion in Offshore Structures
Microbes in the Sea

~
l~

ARGONAUTICS

MARINE ENGINEERING

naval architects
.

- marine surveyors .
heavy lift transportation consultants .

3030 Bridgeway, Suite 114

Sausalito, California 94965

SEAKEEPING:
Ship Behaviour in
Rough Weather
A. R. J. M. LLOYD,

B.Sc.,Ph.D.

Senior Principal Scientific Officer


Admiralty Research Establishment
Haslar, Gosport, Hampshire

ELLIS HORWOOD LIMITED


Publishers Chichester
o

Halsted Press: a division of


JOHN WILEY & SONS
New,York Chichester Brisbane Toronto
o

First published in 1989 by


ELLIS HORWOOD LIMITED
Market Cross House, Cooper Street,
Chichester, West Sussex, P0191EB, England
The publisher's colophon is reproduced from James Gillison's drawing of the ancient Market Cross,
Chichester.

Distributors:
Australia and New Zealand:
JACARANDA WILEY LIMITED
GPO Box 859, Brisbane, Queensland 4001, Australia
Canada:
JOHN WILEY & SONS CANADA LIMITED
22 Worcester Road, Rexdale, Ontario, Canada
Europe and Africa:
JOHN WILEY & SONS LIMITED
Baffins Lane, Chichester, West Sussex, England
North and South America and the rest of the world:
Halsted Press: a division of
JOHN WILEY & SONS
605 Third Avenue, New York, NY 10158, USA
South-East Asia
JOHN WILEY & SONS (SEA) PTE LIMITED
37 J alan Pemimpin # 05-04
Block B, Union Industrial Building, Singapore 2057
Indian Subcontinent
WILEY EASTERN LIMITED
4835/24 Ansari Road
Daryaganj, New Delhi 110002, India

1989 A.R.J .M. Lloyd/Ellis Horwood Limited


British Library Cataloguing in Publication Data
Lloyd, A.R.J.M., 1941Seakeeping.
1. Ships. Hydrodynamics
I. Title
623.8'171
Library of Congress Card No. 88--8402

ISBN. 0--7458-0230--3 (Ellis Horwood Limited)


ISBN 0-470--21232-2 (Halsted Press)
Typeset in Times by Ellis Horwood Limited
Printed in Great Britain by The Camelot Press, Southampton

COPYRIGHT NOTICE
All Ri!lhts Reserved. No part of this publication may be reproduced, stored in a retrieval system, or
transmitted, in any form or by any means, electronic, mechamcal, photocopying, recording or otherwise,
without the permission ofElhs Horwood Limited, Market Cross House, Cooper Street, Chichester, West
Sussex, England.

Table of contents

Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
Notation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
A note on units . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
1 Seakeeping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
___/

Fluid dynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28

2.1
2.2
2.3
2.4
2.5
2.6
2. 7
2.8
2.9
2.10
2.11
3

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Euler's equations of motion for an inviscid fluid . 1. . . . . . . . . . . . .
Equation of continuity . . . . . . . . . . . . . . . . .
The velocity potential . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Integration of Euler's equations of motion: Bernoulli's equation ...
Laplace's equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
The stream function. . . . . . . -. . . . . . . . . . . . . . . . . . . . . . . . .
Some simple flows . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Conformal transformations . . . . . . . . . . . . . . . . . . . . . . . . . . .
Viscosity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Lifting surface characteristics . . . . . . . . . . . . . . . . . . . . . . . . . .

28
29
33
35
38
40
41
43
52
58
59

Regular waves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64

3.1
3.2
3.3
3.4
3.5
3.6
3.7
3.8

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
The potential function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Pressure contours and the surface profile . . . . . . . . . . . . . . . . . .
Wave slope . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Regular wave characteristics . . . . . . . . . . . . . . . . . . . . . . . . . .
Particle orbits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Pressure fluctuations under a wave . . . . . . . . . . . . . . . . . . . . . .
Energy of a regular wave ......... ,, . . . . . . . . . . . . . . . . . .

64
65
66
72
73
74
86
86

Table of contents
3.9

Energy transmission and group velocity . . . . . . . . . . . . . . . . . . . 89

4 Ocean waves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
4.1
Wave generation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
4.2 Statistical analysis of time histories of irregular waves . . . . . . . . . . 94
4.3
Fourier analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
4.4 The wave energy spectrum . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
4.5 Spectral moments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
4.6 Idealised wave energy spectra . . . . . . . . . . . . . . . . . . . . . . . . 106
4.7 Wave slope spectra . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
4.8 Wave spreading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
5 Ocean wave statistics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.1
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.2 Visual observations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.3 Wave atlases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

121
121
121
124

6 The spring-mass system . . . . .


6.1
Introduction . . . . . . .
6.2 Harmonic response . . .
6.3 Free decay. . . . . . . . .
6.4 System with no stiffness

132
132
133
139
141

............................
.. .. .. ... .. .. ... .. .. . .. . .. . .
.. .. .. .. ... .. ... . .. .. .. . .. . .
. ... .. .. ... .. .. .. .. .. . .. . .. .
............................

7 Heading and encounter frequency . . . . . . . . . . . . . . . . . . . . . . . . . . . 144


7.1
Heading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144
7.2 Encounter frequency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145
8 Basic equations for ship motions in regular waves . . . . . . . . . . . . . . . . .
8.1
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
8.2 Axes and ship motion definitions . . . . . . . . . . . . . . . . . . . . . .
8.3
General equations for ship motions in regular waves. . . . . . . . . .
8.4 Coefficients in the equations of motion . . . . . . . . . . . . . . . . . .

151
151
151
154
158

9 Strip theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
9.1
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
9.2 Strip motions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
9.3 Hydrodynamic coefficients . . . . . . . . . . . . . . . . . . . . . . . . . .
9.4 Excitations in regular waves . . . . . . . . . . . . . . . . . . . . . . . . .

170
170
171
173
181

10 Hydrostatic coefficients . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
10.1
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
10.2 Vertical plane . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
10.3 Lateral plane . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

191
191
191
192

11

196
196
196
206
211

Local hydrodynamic properties . . . . . . . . . . . . . . . . . . . . . . . . . . . .


11.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
11.2 Hydrodynamic properties of a heaving circular cylinder . . . . . . . .
11.3 Lewis forms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
11.4 Hydrodynamic properties of Lewis forms . . . . . . . . . . . . . . . . .

Table of contents
11.5
12

Measurements of local hydrodynamic properties

7
218

Roll damping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
12.1
Sources of roll damping . . . . . . . . . . . . . . . . . . . . . . . . . . . .
12.2
Non-linear roll damping: equivalent linearisation . . . . . . . . . . . .
12.3
Eddy roll damping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
12.4
Skin friction roll damping . . . . . . . . . . . . . . . . . . . . . . . . . . .
12.5
Appendage roll damping . . . . . . . . . . . . . . . . . . . . . . . . . . .
12.6
Total roll damping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

223
223
223
225
228

13

Ship
13.1
13.2
13.3
13.4
13.5
13.6
13.7
13.8
13.9
13.10
13.11
13.12

motions in regular waves. . . . . . . . . . . . . . . . . . . . . . . . . . . . .


Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Transfer functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Vertical plane motions in regular head waves . . . . . . . . . . . . . .
Vertical plane motions in regular following waves . . . . . . . . . . .
Vertical plane motions in regular oblique waves . . . . . . . . . . . . .
Lateral plane motions in regular beam waves . . . . . . . . . . . . . .
Lateral plane motions in regular oblique waves . . . . . . . . . . . . .
Absolute motions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Relative motions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Velocities and acceleratirms . . . . . . . . . . . . . . . . . . . . . . . . . .
Lateral force estimator. . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Non-linearities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

234
234
234
235
240
243.
247
252
253
257
257
259
262

14

Ship motions in irregular waves . . . . . . . . . . . . . . . . . . . . . . . . . . . .


14.1
The electronic filter analogy . . . . . . . . . . . . . . . . . . . . . . . . .
14.2
The encountered wave spectrum . . . . . . . . . . . . . . . . . . . . . . .
14.3
The motion energy spectrum . . . . . . . . . . . . . I ~ . . . . . . . . . . .
14.4
Alternative method of calculating motion statistics . . . . . . . . . . .
14.5
Effect of matching the wave spectrum and the transfer function ...
14.6
Motions in short crested waves . . . . . . . . . . . . . . . . . . . . . . . .
14.7
Spectral calculations for non-linear motion responses . . . . . . . . .

263
263
264
266
269
271
272
275

15

Seakeeping trials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 277


15.1
Full-scale trials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 277
15.2
Wave measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 279
15.3
Ship motion measurements . . . . . . . . . . . . . . . . . . . . . . . . . . 279
15.4
Measurements of other seakeeping responses . . . . . . . . . . . . . . 281
15.5
Run lengths and ship courses . . . . . . . . . . . . . . . . . . . . . . . . . 282

16

Model testing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
16.1
Reasons for model seakeepi))g experiments . . . . . . . . . . . . . . .
16.2
Model experiment scaling . . . . . . . . . . . . . . . . . . . . . . . . . . .
Open water model experiments . . . . . . . . . . . . . . . . . . . . . . .
16.3
16.4
Laboratory test facilities . . . . . . . . . . . . . . . . . . . . . . . . . . . .
16.5
Wave makers and.beaches . . . . . . . . . . . . . . . . . . . . . . . . . . .
16.6
Instrumentation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
16.7
Model materials . . . . . . . . . . . . . . . -. . . . . . . . . . . . . . . . . .

2~

233

286
286
286
297
297
300
303
307

Table of contents

8
16.8
16.9
16.10
16.11

Trimming and ballasting . . . . . . . . . . . . . . . . . . . . . . . . . . . .


Testing in regular waves . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Testing in irregular waves . . . . . . . . . . . . . . . . . . . . . . . . . . .
Tank wall interference . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

308
312
318
323

17 Probability formulae . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
17.1
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
17.2
Probability analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
17.3
HistQgrams . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
17.4 The probability density function . . . . . . . . . . . . . . . . . . . . . . .
17.5
The Gaussian probability density function . . . . . . . . . . . . . . . .
17.6 The Rayleigh probability density function . . . . . . . . . . . . . . . . .
17.7 Significant wave height and related statistics . . . . . . . . . . . . . . .
17.8 Joint probabilities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

327
327
327
327
331
333
337
339
342

18 Roll stabilisation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
18.1
Motion reduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
18.2
Bilge keels . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
18.3
Active roll stabiliser fins . . . . . . . . . . . . . . . . . . . . . . . . . . . .
18.4
Passive tanks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

343
343
344
349
377

19 Added resistance and involuntary speed loss in waves . . . . . . . . . . . . . .


19.1
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
19.2
Simple theory for added resistance in regular head waves . . . . . . .
19.3
Added resistance in irregular head waves . . . . . . . . . . . . . . . . .
19.4
Increase of resistance due to wind . . . . . . . . . . . . . . . . . . . . . .
19.5
Propeller characteristics . . . . . . . . . . . . . . . . . . . . . . . . . . . .
19.6 Speed loss . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

398
398
398
400
401
403
406

20

Slamming, deck wetness and propeller emergence . . . . . . . . . . . . . . . .


20.1
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
20.2
Probability of occurrence . . . . . . . . . . . . . . . . . . . . . . . . . . .
20.3
Slamming . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
20.4
Deck wetness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
20.5
Freeboard exceedance . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
20.6
Effect of bow shape . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

409
409
410
413
421
422
424

21

Effects of ship motions on passengers and crew . . . . . . . . . . . . . . . . . .


21.1
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
21.2
Motion sickness incidence . . . . . . . . . . . . . . . . . . . . . . . . . . .
21.3
Subjective motion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
21.4
Lateral force estimator and motion induced interruptions . . . . . .

425
425
426
429
433

22

Seakeeping criteria and voluntary speed loss in rough weather . . . . . . . .


22.1
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
22.2
Equipment criteria . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
22.3
Questionnaires . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
22.4 Speed loss in rough weather . . . . . . . . . . . . . . . . . . . . . . . . .

437
437
439
440
449

Table of contents
22.5

Criteria for speed loss . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 451

23

Operational effectiveness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
23.1
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
23.2
Sea area and season . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
23.3
Ship speed and course . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
23.4
Calculation of operational effectiveness . . . . . . . . . . . . . . . . . .

455
455
455
456
457

24

The effect of hull size form on seakeeping . . . . . . . . . . . . . . . . . . . . .


24.1
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
24.2
Parent hull form . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
24.3
Effect of hull size: changing the size of the hull while keeping
the shape constant . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
24.4
Effect of hull shape . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
24.5
Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

460
460
461
461
468
473

Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 475
Glossary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 479
Numerical values . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 482
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 483

For
Sonya, Abigail and Tobin
who never complained in four and a half years

Acknowledgements

I have been engaged in research on seakeeping since 1968. During that time I have
been helped by many colleagues both within the Admiralty Research Establishment
and outside. In particular I would like to pay tribute to the outstanding contributions
of Mrs P.R. Loader, Dr R.N. Andrew and Mr W. B. Marshfield. Without their
constant support, advice and inspiration over the years this book would probably
never have been written.
Gosport
January 1988

A. R. J. M. Lloyd

j
j
j
j
j
j
j
j
j
j
j
j
j
j
j
j
j
j
j
j
j
j
j
j
j
j
j
j
j
j
j
j
j
j
j
j
j
j
j
j
j
j
j
j
j
j
j
j
j
j
j
j
j
j
j
j
j
j

Notation
(a)

ROMAN SYMBOLS

Symbol
A

Aij (i,j= 1,6)

Meaning
area;
parameter in Bretschneider wave
energy spectrum formula;
Fourier coefficient;
parameter in SM formula
generalised virtual mass or inertia
coefficient: ith force or moment
due to jth unit acceleration
cross-section area of superstructure and hull above waterline

Units
metres 2
metres 2/second 4
metres
kN/(metre/second 2 )

or
kN metres/(metre/second 2 )
metres 2 or
kN/(radian/second 2 )

or
kN metres/(radian/second 2 )

aspect ratio;
inertia;
radius of cylinder;
resistance augment
mapping coefficients
stabiliser fin servo coefficients

a;(i=l,6)

aij (i, j

= 1,6)

ith force or moment due to unit


wave depression acceleration
generalised added mass or inertia
coefficient: ith force or moment
due to jth unit acceleration

tonnes
metres
r;

- , S{!conds,
seconds 2
kN/(metre/second 2 )

or

I,

'

kN metres/(metre/second 2 )
kN /(metre/second 2 ) or
kN metres/(metre/second 2 )

or
kN/(radian/second 2 )

or
a;,(i=1,6)

ith force or mome,nt due to. unit


passive tank angle acceleration
passive tank moment due to ith
unit acce-leration

kN metres ( radian/second 2 )
kN/(radian/second 2 )

or
kN metres/(radian/second 2 )
kN metres/(metre/second 2 )

or
kN metres/(radian/second 2 )

Notation

14

b; (i = 1,6)

b;j

(i,j = 1,6)

passive tank added mass coefficient


parameter in Bretschneider wave
energy spectrum formula;
beam;
breadth;
Fourier coefficient
appendage outreach;
damping coefficient
ith force or moment due to unit
wave depression velocity
generalised damping coefficient:
ith force or moment due to jth
unit velocity

kN metres/(radian/second 2 )
seconds- 4
metres
metres
metres
metres
kN seconds/metre
kN/(metre/second)

or
kN metres/(metre/second)
kN/(metre/second)

or
kN metres/(metre/second)
or

kN/(radian/second)

or

b;~

b~;

(i = 1,6)
(i = 1,6)

fixed stabiliser fin controller coefficients


ith force or moment due to passive tank angle velocity
passive tank moment due to ith
velocity
passive tank damping coefficient
parameter in JONSW AP wave
energy spectrum formula
confidence level
parameter defining propeller
operating point
added resistance coefficient in
waves
block coefficient
drag coefficient
drag coefficient for eddy-making
roll damping
drag coefficient for skin friction
roll damping
lift coefficient
slamming pressure coefficient
beach reflection coefficient
swell-up coefficient
forward waterplane area coefficient

kn metres/(radian/second)
-,seconds,
seconds2
kN/(radian/second)

or
kN metres/(radian/second)
kN metres/(metre/second)

or
kN metres/(radian/second)
kN metres/(radian/second)

kN/metre 2

Notation

C;

C;j

(i = 1,6)

(i, j

= 1,6)

wave celerity;
stiffness coefficient;
fin chord
ith force or moment due to unit
wave depression
generalised stiffness coefficient;
ith force or moment due to jth
unit displacement

15
metres/second
kN/metre
metres
kN/metre

or
kN metres/metre
kN/metre

or
kN metres/metre

or
kN/radian

or

E
F

F; (i = 1,6)

roll moment applied by passive


tank due to unit roll displacement
passive
tank
displacement
coefficient:
tank moment due to unit roll displacement
passive tank stiffness coefficient:
tank moment due to unit tank
angle
drag force;
parameter in wave spreading
formula;
draught
depth of water;
depth of experiment tank;
propeller diameter
energy on one wave length;
effectiveness
force;
test function;
freeboard
ith force or moment required to
sustain general oscillation

kN metres/radian
kN metres/radian
kN metres/radian

kN metres/radian

kN

metres
metres
metres
metres
joules/metre width
kN
metres
kN

or
kN metre

Froude number
ith force or moment due to waves
on restrained ship

probability density function or


frequency distribution ordinate
forward path complex gain
fluid metacentric height
solid metacentric height

kN

or
kN metres
metres - 1 , radians -

metres
metres

etc.

Notation

16

I
I;;

(i = 4,6)

Iij (i, j = 4,6)


(i=l=j)

gain margin
roll-righting lever
acceleration due to gravity
wave height;
beam/draught ratio;
feedback path complex gain
characteristic wave height
significant wave height
distance from pivot point to
centre of gravity;
height
mass moment of inertia
ith mass moment of inertia
cross-products of inertia
second moment of area of waterplane about transverse axis
second moment of area of waterplane about longitudinal axis

y( -1)

propeller advance coefficient


overall gain setting
propeller torque coefficient
speed-dependent gain setting
propeller thrust coefficient
roll controller sensitivities
distance from keel to centre of
gravity
wave number;
ith radius of gyration
lift;
length
lateral force estimator
doublet strength
ith local momentum per unit
length of strip in calm water

KG
Ko
Ku

KT
Kl, K2, K3

KG
k
k; (i = 4,6)
L

LFE
M
MI (i = 2,4)

metres
metres/second 2
metres

metres
metres
metres
metres
tonne metres 2
tonne metres 2
tonne metres 2
metres 4
metres 4

-,seconds, seconds 2
metres
metres- 1
metres

kN
metres
metres/second 2
metres 3/second
tonne metres/second
per metre
or

M~;

(i= 2,4)

first moment of area of waterplane about transverse axis


ith local momentum per unit
length for stationary strip in
waves

tonne metres 2 radians/


second per metre
metres 3
tonne metres/second per
metre
or

tonne metres 2 radians/


second per metre

Notation

Mil
MSI

motion-induced interruptions per


minute
motion
sickness
incidence:
percentage of passengers or crew
who are seasick
source strength;
wave-spreading index;
mass
variance of displacement

17
minutes- 1
per cent

metres 2/second
tonnes
metres 2
or

variance of velocity

radians 2
metres 2/second 2
or

variance of acceleration

radians 2/second 2
metres 2/second 4
or

nth moment of area of energy


spectrum

radians 2/second4
metres 2/second"
or

radians2 /second"
N

n
p

number of observations;
number of motion cycles;
number of observations per hour;
propeller revolutions per second
number of ways in which r positive answers can be achieved from
N questionnaires
dimension normal to passive tank
axis
pressure;
probability;
proportion of time;
power
proportion of questionnaires
returning a positive response to a
particular question
coefficients to weight contribution of 2mth multi pole
kx 81 cos 11-;
propeller torque
passive tank parameter
total velocity;
passive tank coefficient of
resistance;
minimum number of votes in a
questionnaire to establish a
majority opinion with a 95% confidence level

hours- 1
seconds- 1

metres
kN/metre 2

kW

kN metres
tonne metres
metres/second

18

S(ro)

Notation
scale ratio Ls/Lm;
resistance
Reynolds number
radial coordinate;
number of positive questionnaire
answers
nth ship response
relative local vertical motion
between ship and sea surface
energy spectral ordinate

kN
metres

various
metres
metres 2/(radian/second)
or

SM
SA;(i=1,3)

s8 ; (i= 1,3)
S;

(i = 1,3)

u,v

v
v

standard error of the differences


half-separation of source and sink
in a doublet;
girth coordinate;
Laplace transform operator
subjective motion
apparent accelerations in Earthfixed axes experienced by an
object on the deck
accelerations experienced in
body axes by an object on the
deck
ith local absolute motion displacement relative to 0
period;
thrust
duration of time history
time;
Student's test function
time of jth peak or trough
freestream velocity;
ship speed
fluid velocities in x andy
directions
radial and tangential velocity
components
volume of submerged and
emerged wedges;
relative wind speed
voltage;
disturbance
weighting function;
work done by whole ship in one
motion cycle

radians2/(radian/second)
various
metres
metres

metres/second 2

metres/second 2

metres
seconds
kN
seconds
seconds
seconds
metres/second
metres/second
metres/second
or knots
metres/second
metres 3
metres/second
volts

kN metres

Notation

w
w(z)
WT

X,Y

x,y
xi (i = 1,6)
XoJ

Xoo

Z 1, Z2

(b)

width
complex potential function
<Pz + i'l'z in z plane
Taylor wake fraction
force per unit mass in x and y
directions
cartesian coordinate system
ith displacement of centre of gravity relative to moving origin 0
amplitude of Jth peak or trough
initial value of x in decaying
oscillation
eddy-making roll-damping
functions
complex variable x + iy;
displacement of fluid surface
from stabiliser tank datum level

19
metres
metres 2/second

metres/second 2
metres
metres or radians
metres
metres

metres
metres

GREEK SYMBOLS

Symbol
(X

r"
y

Yi (i = 1,6)

v
Lix,~y

Meaning
angle of incidence;
wave slope;
slope of pressure contour
slope of beach;
fin depression angle
deadrise angle at keel
counting functional for nth ship
response
parameter in JONSWAP wave
energy spectrum formula;
inclination of hull section at
waterline
ith peak wave force or moment
leads maximum wave depression
at 0 by Yi radians
volume of displacement
deviations of particle from datum
position beneath a wave
boundarylayer thickness
ith peak (positive) motion displ(!cement leads maximum wave
depression at 0 by oi radians

Units
radians
radians
radians
radians
radians
radians

radians
radians

metres:>
metres
metres
radians
radians

20

J.tw
v
p
(J

CJo

Notation
bandwith parameter;
phase
ith peak (positive) force or
moment leads maximum positive
motion by Ei radians
phase margin
passive tank motion phase: tank
angle leads roll motion by 11
radians
passive tank moment phase: tank
moment leads roll motion by 12
radians
complex variable x 82 + ix 83 ;
depression of water surface below
mean level
significant wave amplitude
wave amplitude exceeded with
probability 1/n
propeller efficiency;
decay coefficient
angular coordinate
tuning factor co/coc
wave length
primary wave direction
fractional loss in metacentric
height due to passive stabiliser
tank
viscosity
secondary wave direction
mass density
section area coefficient
rms displacement

radians
radians

degrees
radians

radians

metres
metres
metres
metres

radians
metres
radians

tonnes/(metre second)
radians
tonnes/metre 3
metres
or

rms velocity

radians
metres/second
or

rms acceieration

radians/second
metres/second 2
or

<j), <P'
X

time constant;
stabiliser tank angle;
shear stress
velocity potential
angle of inclination of flow to y
axis
stream function

radians/second 2
seconds
radians
kN/metre 2
metres 2/second
radians
metres 2/second

Notation
force potential
frequency;
wave frequency

Q
(J)

(c)

21
metres 2/second 2
radians/second
radians/second

SUFFIXES, SUPERSCRIPTS, ETC.

Unless otherwise shown in the table of notation the following suffixes are used:

Symbol
A
a
area
B

BK
BL
b
c
crit
D
D
d
ds
E
e

FP
FS
f

GH
H
h
I

in
J
k
L

LA
M
m
max
n
obs
out
p
r

Meaning
appendage; apparent; air
aft; irregular amplitude or height; added; augmented
sea area
axes fixed in hull; Bretschneider
bilge keel
boundary layer
bilge
calm water; fin controller
critical value or criterion
demanded angle; drag
mean draught
duct; decay
deck submergence
Earth-fixed axes; eddy
encounter; effective
fin; skin friction; frame
measured from forward perpendicular
fin servo
i
forward
group; forward path
open loop
time history or run time; feedback path
histogram bin
interference
f
suffix used to indicate degree of freedom or direction of motion
input
JONSWAP
keel; kinetic
lift; experiment tank length; beach
lateral acceleration
measured from midships
measured value; model value
maximum
nth observation or frequency
visually observed
output
propeller; wave probe; parent; peaks; port; pressure P; potential
half-separation of wires, struts or potentiometer strings; reservoir; root
I'

22
r3
s

s3
season
slam
stag
stall

TH
t

u
v
w

z
z

s
s
(X

X
0
113

Notation

relative motion
ship value; starboard
absolute vertical motion
season
slamming
stagnation
stall
experiment tank
wave period and height
passive tank; tip
ship speed
voltage
waves; wavemaking; wind; waterplane
zero crossing
value in complex z plane
value in complex plane
wave slope
wave amplitude
wave direction
regular amplitude; modal value
significant value
undamped natural frequency or period
average; complex conjugate
local value; non-dimensional value
oscillating part

A note on units

Systeme Internationale (SI) units are used throughout this book. Tonnes are
adopted as the unit of mass with kilonewtons (kN) as the corresponding unit of force
which will accelerate a unit mass at one metre/second 2 . Metres and seconds are used
as the units of length and time. Following common practice, ship speeds are given in
knots or nautical miles per hour.
Appropriate units are quoted for all equations since this serves to remind the
reader of the physical reality expressed by the equations. The reader may, however,
substitute his own units providing these are based on a rational system in which a unit
force is defined as one which imparts unit acceleration to a unit mass. Thus, for
example, a pressure equation quoted here in kilonewtons per square metre is equally
valid in pounds (force) per square foot (when the unit of mass is the slug) or poundals
per square foot (if the mass unit is the pound).
i

I'

>

1
Seakeeping

There are three things which are too wonderful for me, yea, four which I
know not:
the way of an eagle in the air;
the way of a serpent on a rock;
the way of a ship in the midst of the sea;
and the way of a man with a maid.
Proverbs Chapter 30 verses 18-19
In the days of sail ships were very much more dependent on the weather than they
are today. Square-rigged sailing vessels could not sail directly into the wind and were
strictly limited in their ability to go where the master wanted tin severe conditions it
was necessary to shorten sail and even to ride out a storm under bare poles. Many a
ship was lost because she was driven ashore under such circumstances.
Economic pressures often demanded that the ship's ma~ter spread as much
canvas as he dared in order to make the best speed. This is nowhere more graphically
illustrated than in the stories of the clipper 'races' from China to Europe in the
nineteenth century. The first ship home with the newly harvested tea crop co~ld
demand a premium price for her cargo. Speed was of the essen,ce and these ships
sprouted all sorts of additional sails to make the most of every breath of wind
available.
A heavily laden over-canvassed ship must have been an unpleasant home for the
sailors and passengers in rough weather. With the lee gunwhale submerged, the
decks continually awash, deckhouses damp and cold, life must have been miserable.
Yet even in these circumstances the crew would be expected to continue to navigate
and steer the ship and to go aloft in order to shorten sail or spread additional canvas
as the master demanded.
However, the real problems of seakeeping only came to be recognised with the
demise of sail and the advenLof steam as the prime motive power. Now, for the first
time, ships could steam directly into the wind and s~a with a consequent increase of

26

Seakeeping

[Ch. 1

pitch and heave motions. The damaging effects of shipping heavy seas over the bow
began to be experienced. The punishing effects of high speed in rough weather were
not fully understood and at least one ship (HMS Cobra in 1901) is believed to have
been lost after her hull broke in two after slamming in rough weather.
At the same time the steadying effects of tall masts and a good spread of canvas
were lost and the new steam ships were found to roll heavily. It is ironic that this
beneficial effect of sails has only recently been rediscovered with the emerging
technology of wind-assisted propulsion for low-powered merchant ships.
At about this time William Froude, an eminent Victorian engineer, proposed to
build what would become the world's first model towing tank at Torquay in Great
Britain. He had recently developed scaling laws for predicting the resistance of ships
from tests on models and he intended to use the tank for the required scale model
experiments. The British Admiralty accepted Froude's proposal on the condition
that he also used the tank to inyestigate ways of reducing the rolling motion of ships.
In due course towing tanks were built in many different countries. These were
often fitted with wave makers which allowed the behaviour of model ships in waves to
be studied at leisure and provided, for the first time, a technique for refining a fullscale design to ensure adequate performance in rough weather. These model
experiments were usually confined to tests in regular head or following waves with
occasional tests at zero speed in beam waves. Tests at other headings or in more
realistic irregular waves were impossible because of the long narrow shape of the
towing tanks and the simplicity of the wave makers.
These early model experiments allowed some limited developments in the study
of seakeeping but they could not be used to predict the actual performance of ships at
sea because no technique for relating the behaviour of the model in the regular waves
of the laboratory to the behaviour of the ship in the chaotic environment of the real
ocean was available. This situation prevailed for sixty years or more and the study of
seakeeping remained in effective limbo until the publication of a landmark paper by
StDenis and Pierson in 1953. This showed, for the first time, how this problem could
be solved using the techniques of spectral analysis borrowed from the field of
electromagnetic communications.
At about the same time theoretical methods of predicting the behaviour of ships
in regular waves were being developed. The breakthrough came with Ursell's
(1949a,b) theory for predicting the characteristics of the flow around a circular
cylinder oscillating in a free surface. Classical transformation techniques allowed
these results to be applied to a wide range of shapes of ship-like cross-section and the
fundamentals of modern ship motion theory were born.
These developments some forty years ago provided the basic tools requiied to
develop routine techniques for the prediction of ship motions in something
approaching the real irregular wave environment of the ocean. It was now possible
for the first time to predict the rough weather performance of a ship at the design
stage and to allow seakeeping to take its rightful place in the design process.
Since that time seakeeping has remained an active field of research, but
developments have been in the nature of progressive refinements rather than
spectacular advances. Techniques for designing roll stabilisers, criteria determination, prediction of long-term motion statistics and operational effectiveness have

Ch. 1]

27

Seakeeping

all been added to the naval architect's armoury of weapons: seakeeping performance
prediction should now be a routine in any ship design office.
Unfortunately these developments have not been accompanied by much readily
obtainable literature on the subject outside the specialist papers and publications of
the learned societies and research institutes. The time therefore seems ripe for the
publication of a standard text on seakeeping covering all essential aspects in some
detail.
Although the underlying physical principles of seakeeping theory are not generally difficult to understand, the intimate details are mathematically complicated. It
follows that calculations of ship motions and related phenomena require access to
suitable computer programs and computers. No real progress can be made without
them. Fortunately many suitable programs are available in educational, research and
design establishments as well as computer bureaux throughout the world. The PAT86 suite of seakeeping computer programs (available at the Admiralty Research
Establishment at Haslar in the United Kingdom) was used for the examples of ship
motion calculations presented in this book.
No book can hope to cover such a complex subject completely. Indeed such an
undertaking would be inappropriate for all but the most specialised readers. This
book is therefore intended for the practising (and practical) naval architect and the
student. It is hoped that others on the fringes of the profession will also find the book
useful.

I'

2
Fluid dynamics

2.1

INTRODUCTION

Many aspects of the behaviour of ships in rough weather depend on the general laws
of fluid flow. The study of the waves on the sea surface, the resulting responses to
them and detailed considerations of the flow around appendages such as roll
stabiliser fins are all based on the classical equations of fluid dynamics which were
first expounded by the great mathematicians of the eighteenth century. An extensive
knowledge of fluid dynamics is not required for an understanding of seakeeping
theory: nevertheless, a basic knowledge of certain aspects of the subject is needed
and this chapter is intended to give the necessary grounding. Readers requiring a
more detailed treatment of the subject are referred to O'Neill and Charlton (1986).
At the time when the basic axioms were first established the results had little
practical application, and remained only of academic interest until the advent of
flight and powered ships in the late nineteenth century. Even then the practical
application of this rigorous early work was often hampered by the impossibility of
solving the resulting equations for any but the simplest of cases. It was usually
necessary to pretend that the fluid was 'ideal'; that is, it had no viscosity or surface
tension and was incompressible (i.e. the density remains constant at all times). These
rather limiting assumptions resulted in predictions of fluid flow which were sometimes at variance with the observations of experiments.
Fortunately for the student of seakeeping, the negiect of viscosity, surface
tension and compressibility in these equations allows good theoretical predictions of
many kinds of fluid flow which are important in the determination of the behaviour of
ships in rough weather. This is not to say that water does not have viscosity, surface
tension or compressibility; rather that these qualities apparently have little effect on
the fluid flows concerned. So relatively simple solutions to many relevant problems
are possible using the classical equations of fluid dynamics described in the following
sections.
Fluid flows are, in general, three-dimensional and those associated with the

Sec. 2.2]

Euler's equations of motion for an inviscid fluid

29

behaviour of ships in rough weather are no exception. However, we shall see that it is
often possible to simplify matters and treat certain limited aspects of these flows as
two-dimensional. In this case the flow is always in a plane parallel to the x-y plane as
shown in Fig. 2.1. There are no variations of velocity, pressure, density or any other

Fig. 2.1 -Frame of reference for two-dimension:ojl flows.

r:

property of fluid in the direction normal to the x-y plane. For convenience the fluid is
assuJ11ed to have a constant depth d metrest.
I'

2.2

EULER'S EQUATIONS OF MOTION FOR AN INVISCID FLUID

Euler developed the basic equations of motion for fluid particles by considering the
forces on a small rectangular block of fluid. For present purposes his derivation may
be simplified by considering only two-dimensional flow: in this case the block has
sides of length ox, oy and d metres and has its centre at a point (x, y) in the x-y plane
as shown in Fig. 2.2. ox and oy are at first supposed to be small but finite.
In. general the properties of the fluid which are of interest (pressure, density,
velocity, acceleration etc) will vary throughout the x-y plane but will have specific
values at the point (x, y). Suppose that the pressure at the point (x, y) is P kN/metre 2 .
A possible variation of pressure in the x direction is shown in Fig. 2.3. Obviously this
t The use of the word 'depth' does not necessarily imply that

t~e

x-y plane is horizontal.

[Ch.2

Fluid dynamics

30

0~----------------------------~

r-1

-----~
.........-(?+

P-aP ,)x)ov--~

ax

(x,y)

ap
ax

i'lx)bv
2

'

Fig. 2.2- Euler's equations of motion: forces acting on a particle in the x direction.

<)X

<)X
I

Fig. 2.3- Variation of pressure in the x direction.

Sec. 2.2]

31

Euler's equations of motion for an inviscid fluid

will result in different pressures on the end faces of the block and give rise to a
'pressure force' in the x direction. Other possible forces might be externally applied
(such as gravity) or due to friction as adjacent fluid overtakes or is overtaken by the
block under consideration. If the fluid is assumed inviscid (that is, it has no viscosity)
there can be no frictional forces and the only forces which can exist are therefore due
to pressure differences or are applied externally.
If the dimension ox is small the variation of pressure in the x direction may be
approximated by the straight line of slope

oP
ox

kN/metre 2/metre

as shown in Fig. 2.3. Then the pressures at the centres of each end face of the block
shown in Fig. 2.2 are approximately

P- oP ox
ox 2

kN metre 2

on the side nearest to the origin

P+ oP ox
ax 2

kN/metre 2

on the side furthest from the origin

These expressions are approximate because the pressure vanatwn in the x


direction will not in general be linear. We introduce a further approximation by
assuming that the average pressure over each end face is the same as the pressure at
the centre of each end face. These approximations will improve as the block
dimension ox is reduced and eventually become exact if the block size becomes
1
infinitesimal.
Since the end faces have an area ox d metres 2 the net pressure force in the x
direction is approximately

ox) oy d( p-ap
ox 2

(p + oPox ox) oy
2

= -

oP ox oy d
ox

kN

I'

The volume of the block is ox oy d metres 3 . If the mass density of the fluid is p
tonnes/metre 3 the mass of fluid in the block is p ox oy d tonnes. So if X is the
externally applied force per unit mass the resulting force in the x direction is

Xp

ox oy d

kN

and the total force (pressure plus external) is


"
oP
- ax ox oy d+ Xp ox oy d

kN

Fluid dynamics

32

[Ch. 2

These forces result in an acceleration it of the fluid block in the x direction. From
Newton's second law
force

mass x acceleration

so

aP

--ox oy d+Xp ox oy d
ax

pox oy d it kN

(2.1)

Hence dividing by the volume of the block, ox oy d,

1 aP
X --p ax

it metres/second 2

(2.2)

The fluid block will have some velocity which will have components in the x andy
directions. We shall designate these as u and v respectively. These component
velocities will vary with time and with the position (x, y) ofthe particle. For example

u = u(x, y, t)

metres/second

so that the total differential of u is

du

au
au
au
dx +- dy +- dt metres/second
ax
ay
at

(2.3)

The component it of the acceleration in the x direction is

au dx au dy au
du
u = - = - - +- - +- metres/second 2
dt
ax dt ay dt at

(2.4)

but

u-

dx

dt'

dy
dt

metres/second

(2.5)

so that

au
au au
+ v - +- metres/second 2
ax
ay at

u-

(2.6)

Equation of continuity

Sec. 2.3]

33

Substituting equation (2.6) in equation (2.2) we obtain Euler's equation of motion


for the x direction:

I oP
x--P ox

ou + v -ou +ou
ox oy ot

u-

metres/second 2

(2.7a)

A similar analysis for they direction gives

Y-~ oP

P ay

ov
ax

au ov
oy ar

u - + v - +- metres/second 2

(2.7b)

These equations, as we have seen, are approximate for finite dimensions of the
fluid block but become exact if the block dimensions are made infinitesimal. They are
valid for inviscid flow but the fluid may be compress;ble. In the general form stated
they are insoluble.

2.3

EQUATION OF CONTINUITY

A second fundamental equation, the equation of continuity, states that fluid is


neither created nor destroyed. Consider the small rectangular volume shown in
Fig. 2.4. Again the centre is at (x, y), the sides are of'Iength ox, oy metres and the
fluid is d metres deep. Note, however, that in contrast to the moving block of fluid
used to derive Euler's equations, this volume remains stationary relative to the axes.
In physical terms it may be imagined as a wire framework and."Ye are concerned with
the rate of fluid flow through its open faces.

If the velocity components of the fluid at the centre of the framework are again u
and v metres/second and the mass density at that point is p tonnes/ metre 3 , the mass
of fluici flowing in the x direction through the centre of the frame is approximately
pu oy d tonnes/second. Again this result is only generally exact if the volume is
infinitesimal so that the value of pu at its centre is the same as the average value ofpz:t
over the area oy d. The product pu will in general vary over the entire flow regime and
have different values at the two end faces of the framework. Following the approach
used for pressure variations in the derivation of Euler's equations, the mass flow in
the x direction through the face nearest the origin is approximately

ox) us:y d

tonnes/second

into the framework

a(pu) ox) uy
s: d
( pu+--ax 2

tonnes/second

out of the framework


through the opposite face

(pu-

a(pu)

ox

Hence the net mass flow into the framework through these two faces is

34

Fluid dynamics

[Ch.2

0~------------------------------------~x~

Fig. 2.4- Equation of continuity: mass flow in the x direction.

- CJ(pu) bx by d
ax

tonnes/second

The same approach is used to find the net mass flow into the framework through the
other pair of faces:
- CJ(pu) bx by d
CJy

tonnes/second

and the total mass flow into the framework is

- (

a~xu) + a~v))

bx by d

tonnes/second

The mass of fluid within the framework is approximately p bx by d tonnes. Since


the dimensions of the framework are fixed, this mass can only change if the density

The velocity potential

Sec. 2.4]

35

changes. In this case the rate of increase of the mass of fluid inside the framework is
(op/cU) ox oy d tonnes/second and this must balance the net mass flow through the
faces into the framework. Hence

op

ot

ox oy d = -

(o(pu)

ox

+ o(pv))

oy

ox oy d

tonnes/second

and the general equation of continuity is

op

ot

= -

(o(pu) + o(pv))/ tonnes/(metre 3 second)

ox

oy

(2.8)

which is valid for both real and ideal fluids. If the density is constant (i.e. the fluid is
incompressible) this reduces to

ou ov
ox uy

- + ;- =

seconds- 1

(2.9)

which is the equation of continuity for an ideal fluid.

2.4

THE VELOCITY POTENTIAL

The flow of an ideal fluid can be described in terms of the velocity potential. This is a
function <l>(x, y, t) which has some value everywhere in a fluid flow and varies such
that the velocity components are given by its partial derivatives:

u =
v -

o<l>

ox
o<l>

oy

metres/second

(2.10a)

metres/second

(2.10b)

I'

As an illustration Fig. 2.5 shows velocity potential contours for a particular


two-dimensional fluid flow in the x-y plane. The potential may be visualised as a 'hill'
with the fluid velocities given by the slopes in the appropriate directions. So the
velocity u in the x direction at point Q is obtained by taking a section at the plane AA
and finding the slope of the velocity poteptial at Q. The velocity v may be found by a
similar method.
The velocity potential has no physical significance in itself. Although it has a
value everywhere in an ideal fluid flow it cannot be measured directly and is simply a
mathematical artifact which is .useful in deriving the characteristics of a wide range of
ideal fluid flows.
It is often more convenient to work in terms of polar coordinates as defined in Fig.

[Ch. 2

Fluid dynamics

36

_ _ _ _ _ _o

~-e-

til
til

Is

Section at AA

0~----------------~x
Fig. 2.5- Velocity potential in a two-dimensional fluid flow.

2.6. If the velocity potential at A (r, 8) is <P and the velocity potential at
B (r +or, 8 + o8) is <J> + o<j>, it follows that

<P + o<J>

o<J>
o<J>
<P+ox+- oy

ax

oy

o<J>
o<J>
<P +-or+oe

ae

or

metres 2/second

or

o<J> ox + o<J> oy = o<J> or+ o<J> oe metres 2/second

ax

Oy

Or

ae

Now

ox
oy
so

r oe cos e+ or sin

e metres
or cos e- r oe sin e metres

Sec. 2.4]

37

The velocity potential

br Cos II

bx

Fig. 2.6- Conversion to polar coordjnates.

(i

Dr(~~ sine+~; cos e) +r De(~~ cos e- ~;sine)


=

o<j> Dr+ o<j> De


or
oe

metres 2/second
I'

and, from equations (2.10)

Dr(u sin e + v cos e)+ r De (u cos e- v sin e)


o<J>

o<J>

or Dr+ oe De

metres 2/second

Separating radial and tangential components we obtain for the radial velocity

u,

= u sine+ v cos e
o<J>

or

z d
metres /secon

positive away from the origin

(2.1la)

38

Fluid dynamics

[Ch.2

and for the tangential velocity


Ue

2.5

= U COS 8 - V sin 8
1 aq,
= -;. ae metres2/second

positive anticlockwise

(2.11b)

INTEGRATION OF EULER'S EQUATIONS OF MOTION: BERNOULLI'S


EQUATION

An analogous force potential n may also be defined for the externally applied forces
so that

X - ClQ

ax

metres/second 2

= an

metres/second 2

ay

(2.12b)

Partial differentiation of equations (2.10) gives

au
ay

av
ax

a
axay

2
<j>
= -- =-

second- 1

(2.13)

Substituting the 'x' equations (2.10a), (2.12a) and (2.13) into Euler's equation of
motion for the x direction (equation (2. 7a)) we obtain

an 1 aP
au av a2 <j>
=
u
ax + v ax+ axat
ax pax

metres/second

(2.14)

Now

(v2

v av = ~
ax ax

metres/second 2

(2.15a)

metres/second 2

(2.15b)

2
)

and equation (2.14) may be written

~
(u + v + aq,- n + ~)
=
ax 2 2 at
p
2

metres/second 2

(2.16)

Sec. 2.5] Integration of Euler's equations of motion: Bernoulli's equation

39

if pis assumed to be constant (i.e. the fluid is incompressible). Equation (2.16) may
now be integrated to give

qz + o<J>- ~~
n + _P -- Fx (y, t)
me t res z;secon dz
2
at
p

(2.17a)

where q 2 = u 2 + v2 (q is the total velocity) and Fx is an arbitrary function which does


not vary with x (in other words Fx is a 'constant' of partial integration with respect to
x). A similar procedure for they component yields

o<J>

qz

-+--Q+2
at
p

(2.17b)

where Fy is not a function of y. Since the left-hand sides of equations (2.17a) and
(2.17b) are the same, it is clear that the functions Fx and Fv must be identical and have
the same values at all positions and that they are therefore functions only of time.
Hence the integrated form of Euler's equations of motion for an in viscid incompressible fluid is

qz

o<J>

at

-+--Q+-

F(t)

metres 2/second 2

(2.18)

Now equation (2.10) shows that the velocities are functions only of the derivatives of the velocity potential and not of the potential function itself. It follows that
any arbitrary constant may be subtracted from the potential function without
affecting the velocities. The quantity

J~ F(t) dt

metres 2/second

I'

has the same value everywhere in the fluid at some arbitrary timet and may therefore
be regarded as a constant. We may define a new velocity potential

<P'

<P- J~ F(t)

dt

metres 2/~econd

without affecting the velocities in any way. Then

o<J>' = o<J>- F(t) dt metres 2/second

at

at

40

Fluid dynamics

[Ch. 2

and equation (2.18) becomes

qz aq,'
P
-+--Q+-

at

0 metres 2/second 2

This alternative form of Bernoulli's equation is sometimes convenient. The prime


may of course be omitted so that

qz aq,
P
-+--Q+-

at

0 metres 2/second 2

(2.19)

If the flow is steady (i.e. no variations with time) equation (2.18) reduces to
q2

-- Q

+- = F
p

metres 2/second 2

(2.20)

where F is now an arbitrary constant. If there is no external applied force the force
potential Q is constant and the equation reduces to the well known form

P+

pi

= Pstag

kN/metre 2

(2.21)

where Pstag is another arbitrary constant. If the fluid is brought to rest at some point
so that q = 0, the pressure will be

P
and
2.6

Pstag

= PStag kN/metre 2

is known as the stagnation pressure.

LAPLACE'S EQUATION

Substituting equations (2.10) for the velocity potential in the continuity equation for
an incompressible fluid (equation (2.9)) leads to the Laplace equation
0 seconds- 1

(2.22)

which must be satisfied at every point in an ideal fluid flow. A potential function
which satisfies Laplace's equation will therefore describe some ideal flow of an
inviscid incompressible fluid.

The stream function

Sec. 2.7]
2. 7

41

THE STREAM FUNCTION

Consider a small rectangular element of sides 8x and 8y and depth d metres in the
two-dimensional flow field shown in Fig. 2. 7. The volume flow through the element is

'I'=

(u dy- v dx)

Fig. 2.7- Definition of the stream function.

u 8y d

metres 3/second
I(!

in the positive x direction and

v 8x d metres 3/second
in the positive y direction. So the total volume flow across the line OS 1R is

d (u dy- v dx)

I'

metres 3/second

The related quantity

\v

J(u dy- v dx)

metres 2/second

(2.23)

is the volume flow across the line OS 1 R per unit depth of fluid.
Consider now a second line OS 2 R joining the origin to the point R. Provided that

42

Fluid dynamics

[Ch. 2

no fluid is created or destroyed in the space between the two lines (i.e. the equation
of continuity (2. 9) is satisfied) the value of \jl will be unchanged. \jl is therefore a
function only of the position of Rand is independent of the path of integration. \jl is
called the stream function.
Differentiating equation (2.23) gives
d\jl

= u dy- v dx

d\11

metres 2/second

but

't'

a\jl dx + a\jl dy

ax

ay

metres 2/second

so the veiocities are

u =

~;

metres/second

a\jl

v -

metres/second

ax

(2.24a)
(2.24b)

Referring to equations (2.10) we see that the stream function is related to the velocity
potential by
a\jl

a<j>

ay = ax

metres/second

a\jl

a<j>

ax

ay

metres/second

(2.25a)
(2.25b)

Lines on which \jl is constant are called streamlines, and the fluid flows along these
lines when the flow is steady. Streamlines intersect lines of constant potential at right
angles.
In the polar coordinates defined in Fig. 2.6 the stream function is related to the
tangential and radial velocities by

u0

a\jl
a,

ur

1 a\jl
- ~ ae

metres/second
metres/second

(2.26a)
(2.26b)

Some simple Bows

Sec. 2.8]

2.8

43

SOME SIMPLE FLOWS

2.8.1 Uniform stream


Consider a uniform stream with velocity U metres/second inclined at an angle xto the
y axis as shown in Fig. 2.8. The velocities everywhere are given by

USin X

~-

,..

'

UCosx

Fig. 2.8- Uniform stream in the x-y plane.

u
v

= U sin x
= U cos x

metres/second
metres/second

, ii

and the stream function at any point R is, from equation (2:23)
\jl

Uy sin

x- Ux cos x

metres 2/second

(2.27)
I'

From equation (2.10)

u - U sin

x = ~!

U cos X

acp

ay

metres/second
metres/second

and the potential function is therefore

cp = Ux sin x + Uy .cos x +an arbitrary constant metres 2/second (2.28)


We note in passing that the potential function satisfies Laplace's equation, confirm-

44

Fluid dynamics

[Ch.2

ing that it represents a valid flow of an ideal fluid. Fig. 2.9 shows streamlines and
equipotential contours for a uniform stream derived from equations (2.27) and
(2.28).
2.8.2 Sources and sinks
Fig. 2.10 illustrates a point source at the origin in the x-y plane. The source may be
visualised as a narrow tube with porous walls extending over the depth of the fluid.
Fluid is created within the tube at a rate md metres 3/second and the source strength is
said to be m metres 2/second. At some radial distance r metres the fluid recently
created by the source is flowing outwards at a rate md metres 3 /second across a
cylindrical boundary of circumference 2rrr metres and depth d metres. The radial
velocity is therefore

ur

md

2rrrd

2rrr

metres/second

where

The component velocities are

= u, sine =

mx
2rrr 2
my

a<jl

~;

ax

metres/second

a'I'
ax

a<P
ay

metres/second

Integrating, we obtain the stream function for a point source at the origin:

'I'

(~)

m tan- 1
2rr
x

m (rr/2 - e)
2rr

metres 2/second

(2.29)

and the potential function is

m
2

rr loge r

metres 2/second

(2.30)

Streamlines and equipotential contours for a point source are shown in Fig. 2.11.
As expected, the streamlines radiate outwards from the source and the equipotential
contours are concentric circles centred on the source.
A sink is simply a source with a negative strength. Fluid is therefore drawn into
the source at a rate md metres 3/second and the stream and potential functions are

Sec. 2.8]

45

Some simple flows

Fig. 2.9- Streamlines and equipotential contours for a uniform stream.

y
I'

Fig. 2.10- Point source at the origin in the x-y plane.

m(rr/2- e)

2rr

metres 2/second

(2.31)
(2.32)

Note that the equation of continuity is violated at the source (or sink) but is valid
elsewhere in the flow.

46

Fluid dynamics

[Ch. 2

Fig. 2.11 -Streamlines and equipotential contours for a point source at the origin in the x-y
plane.

2.8.3 Doublet or dipole


Stream and potential functions for sources, sinks and uniform streams may be added
to build more complicated flow patterns. One of the simplest combinations is a
source of strength m and a sink of strength - m on the x axis as shown in Fig. 2.12.

Fig. 2.12- Combination of a source and a sink on the x axis.

The stream function is then

21T

m
_ 1 ( -Y-)
tan- 1 -Y-) --tan

(x +S

21T

X- S

metres 2/second

47

Some simple flows

Sec. 2.8]
Now

tan- 1

tan - 1 a - tan- 1 b

a-b)
( 1 +ab

and the stream function is


- 2ys
m
_1 (
)
\jl =-tan
21T
xz + yz- sz

metres 2/second

if s 2 is small compared with x 2 + y 2 , this is given with sufficient accuracy by

If we now choose to move the source and sink towards the origin and at the same time
to increase their strengths in such a way that the product M = 2ms metre 3 /second
remains constant, we may write for the stream function

\jl

My

M cos

metres 2/second

(2.33)

and the potential function is


It!

M sin 9

'

metres 2/second

(2.34)

The combination of a source and a sink at the origin is called a doublet or a dipole.
Fig. 2.13 shows the streamlines and equipotential contours associated with a dipqle
aligned with the x axis. These are all circles centred on they and x axes respectively.
The centres of the circles are at (0, M/27T\jl) and (M/21T(j>, 0).
In a similar way it can be shown that the stream and potential functions for a
dipole aligned with the y axis are given by

\jl

<1>

Mx
27T y(xz + yz)
My
27T y(xz + yz)

=
"

M sin 9
metres 2/second
27Tr
M cos
21Tr

metres 2/second

(2.35)
(2.36)

So changing the orientation of the dipole exchanges streamlines for equipotential


contours and vice versa.

48

Fluid dynamics

.....

[Ch. 2

'

'
X

,/

Fig. 2.13 -

Flow patterns associated with dipole aligned with x axis: - , streamlines;


- - - , equipotential.

2.8.4 Multipoles
Sources, sinks and dipoles are often termed singularities because their stream and
potential functions adopt infinite values at the origin. The potential functions satisfy
Laplace's equation everywhere except at the singularity itself and these functions
therefore represent valid flows of an ideal fluid.
These particular singularities may be visualised in physical terms. As an example
we have already seen that a source may be imagined as a porous tube from which
fluid flows in all directions. However, this is not a necessary requirement for a valid
potential function. A wide range of singularities exist which represent no simple
physically understandable flow. Yet they still represent a valid flow of an ideal fluid
provided that the potential function satisfies Laplace's equation (2.22). Multipoles
fall into this category. The stream and potehtial functions for multipoles aligned with
the x axis are defined as

\jl

M cos (me)
"'"..,.m

metres 2/second

m = 1,2,3, ...

(2.37)

metres 2/second

(2.38)

~,,

M sin (me)
21Trm

<l>

1,2,3, ...

and the corresponding functions for multipoles aligned with the y axis are

\jl

M sin (me)
21Trm

metres 2/second

1,2,3, ...

(2.39)

Sec. 2.8]

49

Some simple flows

<!>

M cos (m9)

2rrrm

metres 2/second

1,2,3, ...

(2.40)

If m = 1 the multipoles become the dipoles already discussed in section2.8.3. Some


examples of streamlines and equipotential contours for various values of m are
shown in Fig. 2.14.

r --- -

'

',

IL---

_____ j
m=3

m=1 (DIPOLE)

--,

: j '

I__ _

________ _j
m=4

m=2 (QUADRUPOLE)

Fig. 2.14- Multipoles: --streamlines; - - - equipotentials.

'

2.8.5

Flow around a circular cylinder in a uniform stream

The stream and potential functions for a uniform stream of velocity U metres/second
parallel to the x axis are, from equations (2.27) and (2.28),

'l' = Uy

metres 2/second

[Ch.2

Fluid dynamics

50

<j>

= Ux metres2/second

These may be combined with a doublet to give

\jl

U
My
y - 21T(x2 + y2)

<1>

Ux+

Mx
21T (xz + yz)

metres 2/second

(2.41)

metres 2/second

(2.42)

and these functions are plotted in Fig. 2.15.

..

<!>INCREASING

I
I

:JJ

)>

(fl

G)

=: :
..,

Fig. 2.15- Potential flow around a circular cylinder in a uniform stream.

Now we have seen that the lines on which the stream function is a constant are
streamlines. Since in steady flow the fluid flows along these lines, any one of them
~ould be replaced by a solid boundary of the same shape without altering the
characteristics of the rest of the flow. If we choose the particular streamline

Some simple flows

Sec. 2.8]

"' =

51

metres2/second

from the set of streamlines defined in equation (2.41), we find it reduces to

metres

(2.43)

or

M
metres 2
21TU

x 2+y 2

= --

(2.44)

Equation (2.43) is the x axis and equation (2.44) is the equation of a circle of radius

~ C:u)

(2.45)

metres

and we can see that equations (2.41) and (2.42) represent the flow around a circular
cylinder in a uniform stream. For simplicity only streamlines and equipotential
contours outside the circle are shown in the illustration of the flow in Fig. 2.15. The
velocities are obtained by differentiating equation (2.41) or (2.42):

o\jl
oy

o<j>
ox

M ( x

= u- 21T (x2 + y2)2

metres/second

(2.46)

'll

0\j/

':\x
"

o<j>
ay

Mxy
( 2 2) 2 metres/second
1T x + y
,
I

(2.47)

At the points Az(- r, 0) and Bz(r, 0) the velocities are zero, and these points at
the front and rear of the cylinder are called stagnation points. As explained in syctipn
2.5 the pressure is then equal to the stagnation pressure of the fluid.
'
At Cz(O, - r) and Dz(O, r) on the top and bottom of the cylinder the velocities are

= 2U metres/second
v = 0 metres/second

So the velocity at these points is twice the freestream velocity and, as would be
expected, parallel to the x axis. The pressure at these points is a minimum given by
equation(2.21):

Fluid dynamics

52
2.9

[Ch.2

CONFORMAL TRANSFORMATIONS

The solution for the potential flow around a circular cylinder in a uniform stream is
but one example of the many ways in which complicated flows can be synthesised by
adding the stream and potential functions for elementary flows. Many such solutions
can be built up by suitable combinations of sources, sinks and uniform streams.
A powerful additional technique allows a wide range of further flow patterns to
be derived from these basic synthesised solutions. The method involves mapping the
streamlines and equipotential contours of a known flow into the streamlines and
equipotential contours of the required flow using some suitable mapping function.
In order to exploit the method we redefine the x-y plane of our known flow
solutions (for example, the flow around the cylinder in the uniform stream) as the
complex plane z such that
z

= x + iy

metres

where i = V(- 1).


We then define a complex potentia/function w(z) by combining the potential and
stream functions to give

w(z) = w(x + iy) = <Pz + i'Vz metres 2/second


This often results in a considerable algebraic simplification. For example, the
complex potential functions of the elementary flows already considered are simply
(a) uniform stream at an angle

w(z) = Uz sin

x to the y axis

x- iUz cos x

metres 2/second

(2.48)

(b) source at the origin

w(z) =

m
2

loge

1T

lzl

metres 2/second

(2.49)

(c) sink at the origin

w(z) = -

m
1T loge
2

lzl

metres 2 /second

(2.50)

(d) doublet aligned with the x axis at the origin

w(z)

M
21TZ

metres 2/second

(2.51)

Conformal transformations

Sec. 2.9]

53

(e) circular cylinder in a uniform stream parallel to the x axis


M
Uz +2rrz

w(z)

The velocities
function:

uz

and

Vz

metres 2/second

(2.52)

are obtained by differentiating the complex potential

o<Pz + 1. -0\jfz

dw(z)
dz

ox

ox

metres/second

or
dw(z)
dz

a (<Pz + i'Jfz ) = - 1
. o<J>z- +0\jfz
ay ay

- i

Uz - ivz

~y

metres/second

metres/second

(2.53)

(see equations (2.10) and (2.24)).


Suppose that we require to find the flow pattern (i.e. the streamlines and the
equipotential contours) around some arbitrary shape which is defined in another
complex plane such that

s = Xsz + ix

83

'!'I

metres

(see Fig. 2.16).

- ---------

.----~------------~~~-
~=f(z)

Z Plane

__. _...

..:::

~Plane

z=F(~)

ixs3

iy

Fig. 2.16- Mapping a shape in the z plane into another shape in the L, plane.

Fluid dynamics

54

[Ch. 2

Then we need to find a mapping function


~

= f(z)

metres

and its inverse

F(~)

metres

relating all points in the z plane to corresponding points in the~ plane. We require
the function to map all points on the surface of the body in our known solution in the
z plane (for example, the circular cylinder) onto the surface of the body about which
the flow is required in the ~ plane. Finding a suitable function to map a simple
geometric shape in the z plane into an arbitrary shape in the ~ plane may be a
formidable problem. However, a wide range of shapes can be produced and it is
often possible to achieve a reasonable approximation to a desired shape using
reiativeiy simpie mapping functions.
Let us suppose that a suitable mapping function has been found. Then for any
point in the ~ plane we may calculate the numerical value of the complex potential
w(z) at the corresponding point in the z plane. Let us call this W(~) where

W(Xsz + ixs3)

W(~)

w(z)

<Pz + i'Vz metres 2/second


Then
dW(~) _
-d- Z

-~uX 82

(.+.
. ) _
'+'z + 1\jfz -

-I-~ux 83

(,~.

'+'z

. )

1\j/2

d
metres1secon

and equating real and imaginary parts we obtain


-

aq,z

-~-

metres/second

(2.54)

uXB3

Comparing equations (2.54) with equations (2.25) it is evident that the complex
potential w(z) calculated in the z plane must also represent some valid fluid flow in
the~ plane.
Consider the value of the complex potential on any streamline (including the
surface of the body) in the z plane. Since the stream function is constant everywhere
along a streamline the complex potential must be of the form

w(z)

real number+ i x constant

Since w(z) has the same value at corresponding points in the~ plane it is clear that the

Conformal transformations

Sec. 2.9]

55

streamlines in the z plane must map into streamlines in the plane. By a similar
argument equipotential contours in the z plane must also map into equipotential
contours in the plane.
The velocities in the plane are obtained by differentiating the complex potential
with respect to

s
s:

dz dw(z)
dz

dw(z)

---

ds

ds

F'(s) (uz- ivz)

ur, -

ivr,

metres/second

(2.55)

As an example consider the mapping function

metres

(2.56)

where the coefficients a0 , a1 and a3 are real. Choosing different values for these
coefficients allows the flow around the circular cylinder in the z plane to be mapped
into the flow around a wide variety of shapes in the plane. We shall meet this
mapping function again in Chapter 11 where we shall see that it is used to obtain the
solution to the flow around shapes of ship-like cross-section. The coefficients are
then required to have particular values to ensure that the mapped shape in the plane
approximates to the required hull cross-section. However, f9 r the time being let us
choose the values
'

Then the mapping function becomes

s = z + alz

On the surface of the cylinder we have

z = ia e-;e metres

z=

ia e;e metres

I,

Fluid dynamics

56

lzl = a

[Ch.2

metres

where

~ (2~U)

metres

(equation (2.45))

and the circular cylinder maps into an ellipse whose equation is

(a+ a~) cos +(a -a~) sin


8

8 metres

and whose major and minor axes are

at

a+-

metres.

a '

Fig. 2.17 shows the streamlines and equipotential contours obtained from this
mapping function for the particular case of

The major and minor axes are then

4a
3 '

2a
3

metres

The velocities are obtained from equations (2.53) and (2.55):


dz dw(z)
d~
dz
2

M)

2
d ( Vz+-2 - -

a 1 dz

21Tz

57

Conformal transformations

Sec. 2.9]

<1>

-=

Increasing

Fig. 2.17- Potential flow around an elliptical cylinder in a uniform stream.

/:zu (1- a:)


z -a 2

metres/second

'

(2.57)

1!

The front and rear stagnation points Az and Bz on the circular cylinder map into the
corresponding points As and Bs on the ellipitical cylinder (Fig. 2.17). Putting z = a
in equation (2.57) yields us- ivs = 0 metres/second, so that these points are ,also
stagnation points in the ~ plane. Similarly the points Cz and Dz on the circular
cylinder, which experience the maximum velocity 2U metres/second in the z plane,
map into the corresponding points Cs and Ds on the elliptical cylinder. Putting
z = ia metres in equation (2.57) and equating real and imaginary parts gives

u =
v

3U

metres/second

0 metres/second

for the velocities at Cs and Ds. The velocities at any point in the~ plane may be found
by this method.

Fluid dynamics

58
2.10

[Ch. 2

VISCOSITY

Although neglect of viscosity can often be justified, it cannot always be ignored. In


particular, viscosity has important effects very close to the surface of a body in a fluid
and is responsible for the existence of the 'boundary layer', a thin layer of slow
moving fluid immediately adjacent to the body surface.
Consider the flow around the body shown in Fig. 2.18. Immediately behind the

Fig. 2.18- Growth and separation of boundary layer (boundary layer thickness exaggerated).

front stagnation point the boundary layer will be laminar with a smooth well ordered
structure and the velocity profile shown in Fig. 2.19. (x andy are here taken as

0.8

0.2

0.4

0.6

0.8

1.0

.!!_

u'

Fig. 2.19- Laminar and turbulent velocity profiles.

Sec. 2.11]

Lifting surface characteristics

59

parallel and normal to the local body surface.) At the transition point eddies will
begin to grow in the boundary layer and the velocity distribution changes to the
turbulent profile also shown in Fig. 2.19.
The shear stress applied by one layer of fluid moving over another is given by

't

du

f..tw dy

kN/metre 2

(2.58)

where f..tw is the coefficient of viscosity. At the surface this appears as a frictional force
on the body. Clearly the force applied to the body by the turbulent boundary layer is
much greater than that applied by the laminar boundary layer.
Outside the boundary layer the velocity gradient is very small and the viscous
forces are negligible. So potential flow methods can be applied provided that the
boundary layer is relatively thin. This is generally true over the forward portion of
the body where there is a favourable (falling) pressure gradient. This helps to
minimise the growth of the boundary layer and keeps the flow firmly attached to the
body surface. However, the pressure gradient beyond the maximum diameter of the
body is adverse (rising) which tends to slow the flow and leads to a rapid thickening of
the boundary layer. At some stage the velocity gradient at the surface may become
zero as shown in Fig. 2.18. Beyond this point flow reversal occurs and the boundary
layer is said to separate. Large-scale eddies which are not predicted by potential flow
methods will then occur. Predictions of forces based on potential flow may then be in
error.
Flow separation may occur whenever there is an adverse pressure gradient on
bodies with tapering tails. The more rapid the taper, the more likely is separation. In
particular, separation is virtualiy guaranteed at any discontiJ,1~ity or sharp corner on
the body surface.
'

2.11

LIFTING SURFACE CHARACTERISTICS

Ship hulls are usually fitted with appendages such as rudders, propyller shaft brackets
and roll stabiliser fins. These can influence the behaviour of the ship in rough weather
and we shall require a method of estimating the forces developed by them. These
forces may be estimated using potential flow methods based on the techniques
described above, but results of adequate accuracy can be obtained using the simple
empirical formulae given below.
Consider the typical lifting surface appendage shown mounted on the hull in Fig.
2.20. The geometry of the surface is conveniently defined by the root and tip chords
cr and ct and the outreach b. The mean chord is

metres

Fluid dynamics

60

[Ch. 2

c,

c,

Fig. 2.20- Lifting surface notation.

and the planform area is


A

(2.59)

The aspect ratio, defining the general proportions of the surface, is


a -

2b

4b

(2.60)

When the surface is at an angle of incidence ex to the incident flow it will generate a
lift force L and a drag force D. These forces are respectively normal and parallel to
the direction of the incident flow.
For a given angle of incidence and planform shape the lift and drag are found to
be proportional to the square of the forward speed and the planform area. So the lift
and drag may be expressed in non-dimensional terms as

Lifting surface characteristics

Sec. 2.11]

61

(2.61)
D
!pU 2A

(2.62)

Most surfaces used in ship applications have a streamlined symmetrical section


like that shown in Fig. 2.20. For these sections there is no lift at zero incidence and
the drag is a minimum. Fig. 2.21 shows typical non-dimensional lift characteristics of
such a surface.

Stall

c
Q)

'()

~
Q)
0

.::=

:.::;

Low aspect ratio


'll

Incidence angle a

Fig. 2.21- Typical lifting surface characteristics.


I'

For small angles of incidence the lift coefficient increases more or less linear'ly
with the angle of incidence and we may write the lift coefficient as
(2.63)
where dCdd~ is the lift curve slope.
The slope of the curve diminishes 'as the angle of incidence is increased and
maximum lift occurs at the stall angle ~stall The lift curve slope increases with the
aspect ratio, but surfaces of high aspect ratio stall earlier and more abruptly than
those of low aspect ratio. TQ.e lift characteristics of symmetrical sections are only
weakly dependent on the section shape.

Fluid dynamics

62

[Ch. 2

Whicker and Fehlner (1958) tested a variety of lifting surfaces of low aspect ratio
such as are typically employed on ships and derived an empirical formula for the lift
curve slope as a function of aspect ratio:

l.81Ta
d'
1
1. 8 + y'(a 2 + 4) ra tans-

(2.64)

Fig. 2.22 shows this formula plotted for rectangular lifting surfaces together with

Aspect ration a

Fig. 2.22- Lift curve slope. (After Whicker and Fehlner (1958).)

illustrations of the surface planforms associated with various aspect ratios. Clearly
the lift curve slope increases dramatically with increasing aspect ratio: in other words
long slender lifting surfaces (like a glider~s wings) are much more effective than short
stubby surfaces.
Whicker and Fehlner also reported the stall angles found for their lifting surfaces.
These are given approximately by
1.05- 0.445a + 0.075a 2
0.39

radians

radians

for a>3.0

for a <3.0

(2.65a)
(2.65b)

If the angle of incidence approaches 90 degrees the lift force (normal to the flow

Sec. 2.11]

Lifting surface characteristics

63

direction) becomes zero. The drag force is then very large and acts normal to the
plane of the surface. For this case Hoerner (1965) gives
C0

1.17

(2.66)

'II
r

I,

3
Regular waves

3.1

INTRODUCTION

The waves which influence the behaviour of ships at sea are generally irregular and
more or less random in nature. No two waves have exactly the same height and they
travel across the surface at different speeds and in different directions. Techniques
for coping with the chaotic nature of the real sea surface are described in Chapter 4,
but it is first necessary to discuss the characteristics of ideal regular waves.
Such waves never occur in the real ocean environment although they can be
produced in laboratory towing tanks and form the basis of many seakeeping model
experiments. Of equal importance is the fact that the theory of irregular waves is
based on the assumption that they can be represented by 'superposing' or adding
together a suitable assembly of regular waves. So it is clear that the characteristics of
regular waves have a profound influence on the behaviour of ships in rough weather
even though they are never actually encountered at sea: an understanding of their
nature is one of the vital tools in the study of seakeeping.
Fig. 3.1 shows a train of regular waves advancing across the surface of a body of
water of constant depth d. The waves are two-dimensional: that is, they advance in
the x direction and the crests are perpendicular to the x axis. The crests may be
considered as extending to infinity on either side of the x axis; alternatively the waves
may be imagined to be advancing down a long narrow tank bounded by vertical walls
parallel to the x axis;
The saiient characteristics of the waves are:
~

~0

'A
c

the instantaneous depression of the water surface below the mean level
the wave amplitude or vertical distance from the mean level (y=O) to a crest or
a trough; ~ 0 is always positive
the wave height: twice the wave amplitude
the wave length: the horizontal distance (in thex direction) between one crest
(or trough) and the next
the wave celerity: the velocity of an individual crest in the x direction

Sec. 3.2]

The potential function

65

Fig. 3.1- Regular waves.

cxo
RIA.

the wave period: the time interval between successive crests (or troughs)
passing a fixed point
the instantaneous wave slope: the gradient of the surface profile (in radians)
the maximum wave slope (in radians)
the wave steepness.

'ri

These waves progress across the surface in a regular orderly fashion. Each wave
crest advances at the same steady velocity c so that the waves never overtake each
other and the wavelength A. and period Tremain constant. The shape of each wave
remains the same and the whole wave train appears to advance like a rigid corrugated
sheet. Fortunately for the student of seakeeping, the characteristics and deta~led 1
structure of regular waves are very well predicted by the technique of classical fluid'
mechanics outlined in Chapter 2. In common with the treatment of many other
amenable flows it is necessary to assume that the water is incompressible and inviscid
in order to obtain a workable solution. This does ,not imply that water is actually
incompressible or inviscid: merely that the values of compressibility and viscosity are
such that they have little discernible influence on the characteristics of regular waves.
We shall also assume that the effects of surface tension are negligible. This restricts
the validity of the solutions to wavelengths greater than about 0.1 metres.

3.2 THE POTENTIAL FUNCTION


It is first necessary to find a p,otential function <1> which describes the fluid flow
associated with a regular wave. A large number of potential functions which happen

Regular waves

66

(Ch. 3

to satisfy Laplace's equation (2.22) could be formulated and each would describe
some flow of an ideal in viscid incompressible fluid. The choice of a potential function
to describe some particular flow is a matter of considerable mathematical skill and
insight coupled, no doubt, with a good deal of trial and error. Lamb (1932) showed
that the potential function
.-~. _g(, 0 cosh(k(d- y)]
(k _ )
ro
cosh( kd)
cos x rot

'+' -

(3.1)

is appropriate to the case of the two-dimensional regular wave of amplitude (, 0


propagating across the surface of a body of fluid of any constant depth d as illustrated
in Fig. 3.1; k and ro are constants whose physical meaning will be derived in Section
3.3.
The potential function satisfies Laplace's equation (2.22) so it is confirmed as a
valid representation of some ideal fluid flow. If we assume that the bottom boundary
(the sea bed) at y = dis impervious or waterproof there should be no flow through it.
So
( v )y = d

=0

metres/second

and hence, from equation (2.10b),

( ~<P)
y

=0

metres/second

y=d

and the chosen velocity potential also satisfies this condition.


3.3

PRESSURE CONTOURS AND THE SURFACE PROFILE

Bernoulli's equation (2.19) for the unsteady motion of an ideal fluid


2

q
2

+ a<j)- n + ~ = 0

at

metres 2 /second 2

(3.2)

must appiy everywhere and can be used to find the surface profile associated with the
velocity potential given by equation (3.1). The only force applied externally to any
fluid particle is gravity. Hence, from equation (2.12),

X= an = 0 metres/second 2

ax

an = g

Y =-

ay

metres/second2

67

Pressure contours and the surface profile

Sec. 3.3]

Hence
Q=

gy metres 2/second 2

and equation (3.2) becomes

q2 aq,
P
- + - - gy + - = 0 metres 2/second 2
2

at

(3.3)

In calm water the pressure at depth Yp metres is

P= pgyP kN/metre 2
and a constant pressure contour is a horizontal straight line. Under regular waves this
contour is distorted as shown in Fig. 3.2. The depth of a point on this contour is
y

= yP + Sp metres

where Sp is the depression of the contour below the depth Yp

Surface profile

I
Constant pressure contour

~P~ogy,

y
I'

Fig. 3.2- Constant pressure cpntour beneath a regular wave.

Since the pressure everywhere along the contour and the depth yP are both
constant, the quantity

Regular waves

68

J~ (~-gyP) dt

[Ch.3

metres /second

will be a constant on the contour at any given timet. It may be added to the potential
without affecting the velocities in any way (since they are functions of the potential
gradients and not of the potential itself). So we may define a new velocity potential

<j>'

= <j> +

J~ (~-gyP) dt

metres 2/second

so that

and equation (3.3) becomes


z aq,'
~ + at - g Sp = 0 metres 2/second 2

and the prime may now of course be omitted.


If we now assume that the velocity is small (tantamount to assuming that the wave
amplitude so is small compared with the wave length) we may neglect q 2 so that the
depression of the constant pressure surface is

or, since

Sp is small,
(3.4)

Substituting the expression for the velocity potential (equation (3 .1)) yields the
equation for the constant pressure contour at depth metres:

yP

Sp- so

cosh[k(d- Yp)] .
_
cosh(kd)
sm (kx rot)

metres

(3.5)

69

Pressure contours and the surface profile

Sec. 3.3]

This and many other expressions which follow can be simplified by using the
following approximations:
(a) for water depth greater than about half the wave length
cosh[k(d- Yp)] = sinh[k(d- Yp)] = cosh[k(d- Yp)] = sinh[k(d- Yp)] =ex ( _ k )
cosh(kd)
cosh(kd)
sinh(kd)
sinh(kd)
p
Yp
(3.6)
tanh(kd) = 1.0
(b) for water depth less than about 0.03 times the wavelength
cosh[k(d- Yp)] =
10
cosh(kd)
'

cosh[k(d- Yp)] _ _!_


sinh(kd)
- kd

sinh[k(d- Yp)] = k(d _ ),


cosh(kd)
Yp

(3.7)

sinh[k(d- Yp)] _ d- Yp
sinh(kd)
d

tanh( kd) = kd
So the constant pressure contour in deep water is given by
Sp

=so exp(- kyp) sin (kx- rot)

metres

'tl

(3.8)

'I

and in shallow water by


Sp

=so sin (kx- rot)

metres

(3.9)!,
I

As an illustration of these equations Fig. 3.3 shows typical pressure contours


beneath a regular wave of length 100 metres in three different depths of water. These
results have been obtained by setting t = 0 in equations (3.5), (3.8) and (3.9).
In very shallow water (d = 2 metres) equation (3.9) applies and the amplitude of
the pressure contour is everywhere the same. In deeper water (d = 20 metres)
equation (3.5) is appropriate and the amplitude decreases as the bottom is
approached. In very deep water (d = 100 metres) equation (3.8) applies and the
amplitude of the pressure contour then becomes very small at the bottom.
The surface profile is one of these constant pressure contours (with the pressure
equal to the atmospheric press~re). It is obtained by setting
Yp = 0 metres

70

Regular waves

[Ch. 3

K/1\A ld\f\o,

~p

I
I

' l'

VVlI

:
I

~0.25

"-.../

/"- /'.I
v
v lI
I

Vvl

I0.5
I

r-----------., 0.75

'-----------11.0
Depth 10m

Depth 2m

Depth 100m

Fig. 3.3- Constant pressure contours under a 100 metre long wave in various depths of water.

in equation (3.5) to give

s=so sin (kx- rot)

(3.10)

metres

so

which is the equation of a regular wave of small amplitude metres advancing across
the fluid surface. This result is independent of water depth. So the chosen velocity
potential function does indeed represent the fluid flow associated with a regular
wave.
Equation (3.10) is illustrated in Fig. 3.4. Consider first the wave shape in the
geographical or spatial sense. This is tantamount to fixing time at some instant t as,
for example, when taking a photograph. If for simplicity we choose t = 0, equation
(3.10) is reduced to

s=so sin(kx)

metres

which represents a simple sine wave starting at the x origin.


The wavelength is A metres and the 'wave number' k is now seen to be
k

= 2rr metres- 1
A

(3.11)

71

Pressure contours and the surface profile

Sec. 3.3]

t=t2

Wave shapes at successive time intervals

x,

),

kx,
6)

so

t, t2

Time history of elevation at x =x,

Fig. 3.4- Regular waves pictured in space and time.

A second photograph taken a short while later at t = t 1 ~econds would reveal


exactly the same wave profile with the same amplitude and wavglength moved along
the x axis a distance rot/k metres. This can be seen by recasting equation (3.10) as

! '

and the term rot1/k (metres) can be recognised as a phase 'lag' which governs the
location of the wave along the x axis. As time passes, the lag increases and the wave
advances steadily away from the origin with velocity
(l)

c =k

metres/second

(3.12)

c is the wave celerity or phase velocity.


An alternative view of events can be obtained by fixing the distance x and
allowing time to proceed. Physically this can be considered as recording the time
history of the rise and fall of the water surface at some fixed point x = x 1 metres. The

72

Regular waves

[Ch. 3

resulting sine wave of amplitude /,; 0 metres and period T seconds is also illustrated in
Fig. 3.4 and the frequency ffi is related to the period by

ffi

2
;

radians/second

(3.13)

The time history is given by recasting equation (3.10) as

/,;= -/,;0 sin[ ffi (t-

k~ 1 ) J

(3.14)

metres

and the term kx 11ffi (seconds) may now be recognised as a phase 'lag' which governs
the temporal location of the sine wave on the taxis.

3.4

WAVE SLOPE

It is sometimes convenient to quantify the effects of the waves in terms of their slope
rather than their elevation or depression below the mean level. The slope of the
pressure contours may be obtained by differentiating equations (3.5), (3.8) and (3.9)
as appropriate.
For any depth yP
-~a:Yr-

dx - k/,; 0

cosh(d-yp)
_
cosh(kd) cos(kx ffit)

radians

(3.15)

For deep water


a:Yr

= k/,;0 exp(- kyp)

cos(kx- ffit)

radians

(3.16)

For shallow water


a:Y

= k/,; 0

cos(kx - ffit)

radians

(3.17)

The wave slope at the surface is obtained by setting yP = 0 metres to give, for any
water depth,

ex= cx0 cos(kx- ffit)

radians

(3.18)

where the wave slope amplitude is


O:o

= k/,; 0

radians

(3.19)

So the wave slope varies sinusoidally in both time and space in much the same way as

73

Regular wave characteristics

Sec. 3.5]

the surface elevation. Fig. 3.5 shows the surface profile and the corresponding wave
slope at some timet= t1 The wave slope is a maximum when the surface elevation is
zero and vice versa.
(J)t,

~\

1\

~0

\ "-__/

'

Fig. 3.5- Wave depression and wave slope profiles at t = t 1


'11I

The time history of the wave slope at a certain location f = x 1 is obtained by


recasting equation (3 .18) as
1'

I,

and this is compared with the corresponding surface elevation time history (equation
(3.14)) in Fig. 3.6. The wave slope lags the surface depression by a quarter of a wave
period.

3.5

REGULAR WAVE CHARACTERISTICS

At the suiface the vertical velocity of the water is

74

Regular waves

[Ch. 3

kxl

"'

.. t

Fig. 3.6- Wave slope time history at x = x 1

Since the wave elevation is assumed to be small this may be written

(see equation (3.4)). Substituting equation (3.1) for the velocity potential we obtain
the relationship between the wave frequency and the wave number for regular waves
of small amplitude 1;,0 in water of any (constant) depth d:
co= y[gk tanh (kd)]

radians/second

(3.20)

Combining equations (3.12) and (3.20) gives for the ceierity

c=

~ (f tanh (kd))

metres/second

(3.21)

Combining equations (3.5)-(3.21) in various ways leads to a multitude of relationships between wave frequency, period, number, length and celerity, and these are
listed for easy reference in Table 3.1. The deep water approximations are usually
considered adequate for most practical purposes.

Particle orbits

Sec. 3.6]

75

Figs 3.7 and 3.8 show how the wave period and celerity vary with wavelength for
various water depths. As might have been expected, long waves have very low
frequencies and vice versa. As if to compensate for this the celerity increases with
wave length: for example, the celerity of a 1000 metre wave is almost 40 metres/
second (over 75 knots) in deep water compared with only about 4 metres/second
(about 8 knots) for a 10 metre wave. This dependency of celerity on wavelength
distinguishes surface waves from some other forms of wave motion (notably
electromagnetic radiation) and we shall see that it is responsible for some peculiar
properties of waves on the sea surface.
In shallow water the wave celerity is reduced and the 1000 metre wave's deep
water celerity is almost halved in water of 50 metres depth. In very shallow water the
celerity becomes independent of wavelength and depends only on the water depth:

c=\l(gd)

3.6 PARTICLE ORBITS


According to equation (2.10) the velocities at any point under the wave can be found
by differentiating the potential function given by equation (3.1). This gives

o<!>
.
u = ax = - u0 sm (kx- rot)
a<j>
V = oy = - Vo

COS

metres/second

(kx- rot) metres/second

(3.22)

'ilI

(3.23)

where the velocity amplitudes are

u0 = gk'C,o cosh[ k (d - y)]


ro
cosh (kd)

metres/second

(3.24)
I'

gk'(, sinh[k (d- y)]


metres/second
ro
cosh (kd)

v0 = --0

1.

(3.25)

These relationships may again be simplified for the two special cases of deep and
shallow water. For deep water
!fo = Vo = ro'C,o exp ( - ky)

metres/second

(3.26)

and for shallow water

u0 = ro'(, 0; v0 = ro'(,ok( d- y) metres/second

(3.27)

Regular waves

76

[Ch.3

-o
c
0

(.)

Q)

"'c

();
.~

-o
~

>
(.)
c

Q)

:::l

0"

Q)

>
ctl

0
1

8 10

20

40

60

200

1000

400

Wave length (metres)


Fig. 3.7- Wave frequency.

40

c=

~[~tanh(2~d)J

'0
c
0

(.)
Q)

"'~

a;
E
~

20m

Q)

10m
5m

8 10

20

40

60

100

Wave length (metres)


Fig. 3.8- Wave celerity.

200

Table j . l - Kegular wave relatiOnships

Quantity

In terms
of

Any depth

Deep water
(d > O.SA)

Shallow water
(d<0.03A)

C/l
~

!"">
(.;J

(J)

21T
T

21T
T

y'[gk] tanh (kd)]

y'(gk)

k y'(gd)

~[ Z~g tanh ( 2~d)]

~e~g)

21T y'(gd)
A

21T
T

(J)

..,"'=

21T

21T

Ill

21T

(J)

(J)

(J)

21T
y'[gk tanh(kd)]

21T
y'(gk)

21T
k y'(gd)

~e;A)

A
y'(gd)

r.;
;-

..,

s:

"'

27T'lt
A

~ g tanh e~d)

2c

1Tg

-.l
-.l

Table 3.1 (continued)


-...)

Quantity

In terms
of

Any depth

Deep water
(d > 0.5A)
(!)2

(!)

Shallow water
(d<0.03A)
(!)

V(gd)

41T2
gT2

21T
T\f(gd)

21T
A

21T
A

00

21T
A

(JQ

=
S'
..,

g
c

c2

~
D:l
~

..,

(!)

21Tg
0)2

gT2
21T

21T
k

21T
k

21TC 2
g

21rV(gd)
(!)

T\l(gd)
21T
k

Q
(jJ

Table 3.1 (continued)

Quantity

In terms
of

Any depth

Deep water
(d>0.5A)

Shallow water
(d<0.03A)

Vl
(b

(')

V)

(l)

(l)

gT
21T

~(~tanh (kd))

~(f)

~[~~tanh (2 ~d) J

~(~~)
2

mso
y(gd)

41Tzso
gTz

21Tso
Ty(gd)

ffi

(l)

e.

y(gd)

<Xo

.,..,

--

So
g

kso

kso

kso

21Tso
A

21Tso
A

21Tso
A

gso

a.
tD

..,
Q

8:
.....

"'

-.I
\0

Table 3.1 (contmued)

Quantity

Uo

In terms
of

Deep water
(d > 0.5A)

Any depth

Shallow water
(d < 0.03A)

00
0

soro exp ( - ;zy)

(J)

21Tso
( - 41T Y)
T
exp ----gf2

soV(gk) cosh [k(d-y)]


V[sinh(kd) cosh(kd)]

soV(gk) exp(- ky)


~
~

IJQ

e..,

2
so (21Tg) 112 cosh ( 1T( dA- y))

~ [A sinh ( 2~d) cosh ( 2~d) J

s0

~ (21Tg)
exp (-21Ty)
A
A

(J)

<

"'

so~ ~~xp ( ~~y)

so8
c

So~(~)
soro exp ( - :Zy)
2

Ill

Vo

Ill

21Tso
( - 41T Y)
T
exp ----gf2

soro (d- y)

n::r

21Tso (d- y)
Td

t.N

Table 3.1 (continued)


C/J

Quantity

Vo

Any depth

In terms
of

so y(gk) sinh[k(d- y)]


y[sinh(kd) cosh(kd)]

Shallow water
(d<0.03A)

Deep water
(d>0.5A)

So V(gk) exp(-

ky)

(l)

!"">
v:>

sok ~ (~) (d- y)

2
so y(2rrg) cosh( rr(dA- y))
A

[ . (T2rrd) cosh e7Td)]


T

so~( 2~g) exp( -~rry)

2:so

~ (~) (d _y)

y A smh

.,""=
Col

e.

a.

gy)

.,

so exp(- 2
c
c

~
.....

"'

Xo

so exp(- :

0)

y)

--

---:rg

So exp (- 4rr y)

s cosh[k(d- y)]
0
sinh(kd)

so exp(-

ky)

: ~(~)
~; ~(~)
So

kd

00

......

Table 3.1 (continued)

Quantity

Any depth

In terms
of

Xo

Deep water
(d>O.SA.)

Shallow water
(d<0.03A.)

00
N

s0 cos he7T(dA. y))

21Ty)
soexp (-A.-

A.
. h(27Td)
SID
-

-so"-

27T

A.

s 0 1~xp (

7-gy)
2

Yo

co

so e:xp(- : y)

47T2y)
soexp (-T
-

~
~

IJQ

.,=e.
~

so sinh[k(d- y)]
sinh (kd)

k
s

=
~

"'

so exp(- ky)

. h(27T (d-y))
A

0 SID

A.

So

21Ty)
e~xp (-A-

. h(27Td)
SID
-

nP"

A.

so exp( -Jy)

.I itUIII:: .J .I

Quantity

In terms
of

~ ~UllllllUt:U)

Any depth

Deep water
(d>0.5A.)

Shallow water
(d<0.03A.)

C/l
(b

!'1
v.l

Yo

Ua

(!)

So (d- y)
d

2ffi

gT

41T

'"I
=.

k
A.

~(:k tanh(kd)) (1+ sin~~:kd))


~[~! tanhe~d) J

+A sinh(~~d)
4wd

12.

~(:k)

'"I

s:.....
"'

vg"'

81T

~(g:)
00
v.l

Regular waves

84

[Ch.3

Fig. 3.9 illustrates these formulae for shallow and deep water as well as for the
general case. The vertical velocity is always zero at the sea bed (since the bottom is
assumed to be impervious) and the motion is therefore purely horizontal at this
point. The horizontal velocity is generally greater than the vertical velocity except in
very deep water where the two amplitudes are everywhere the same. In shallow
water the horizontal velocity amplitude is constant but it generally increases with
height above the sea bed and assumes an exponential variation in very deep water.
Following our physical interpretation of the characteristics of the wave profile we
shall now adopt the same techniques for examining the geometry of the structure of
the flow at a particular location by fixing x andy and allowing time to proceed. Since
the wave amplitude is assumed to be small the velocity amplitudes given by equations
(3 .24) and (3 .25) must also be small and it follows that a particle of water oscillating
about some point (x, y) will never stray very far from that point. The path of the
particle can therefore be calculated approximately by assuming that it is always
subject to the velocities calculated for the point (x, y). With this assumption the
particle's trajectory is obtained by integrating equations (3.22) and (3.23) to give
Ax= - x 0 cos (kx- rot)
~y

=Yo sin (kx- rot)

metres

(3.28)

metres

(3.29)

where Ax and ~yare the deviations of the particle from its datum position (x, y) and
the amplitudes of its displacements are
x = u0 = gks 0 cosh[k(d- y)] =
0

Yo

ro

ro 2

s cosh[k(d- y)]
0

cosh(kd)

= v0 = gks 0 sinh[k(d- y)] =


ro
ro 2
cosh(kd)

sinh(kd)

s sinh[k(d- y)]
0

sinh(kd)

metres

(3.30)

metres

(3.31)

Again the deep and shallow water approximations may be applied. In deep water

Xo =Yo= so exp(- ky)

metres

(3.32)

and in shallow water

Xo =

~~'

Yo= so (d- y)

metres

(3.33)

Equations (3.28) and (3.29) represent an elliptical orbit with major (horizontal) axis
2x0 and minor (vertical) axis 2y 0 . Individual particles of water under a wave will
therefore follow elliptical paths as shown in Fig. 3.10. At the surface (y = 0) the
minor axis is always twice the wave amplitude but the orbits become flattened as the

85

Particle orbits

Sec. 3.6]

Orbital velocity amplitude (metres/second)

0.5
I

1.0
I

1.5
I

2.0

2.5

0.5

1.0

1.5

0.5

1Vo

0.25

Depth
100m

0.75

1.0

Fig. 3.9- Orbital velocity amplitudes;

Depth 2m

y!d

0.5

Uo

Uo

Depth
2m

1.0

A.= 100m, so= 1m.

Depth 10m

Depth 100m

ffi---EB~

cp
y

'tl

'I

~
.

'

_____...___________ 4=
__L __________ l

Bottom
111111111
ill
I.

Fig. 3.10- Orbit shapes under a 100 metre wave.

sea bed is approached. At the bottom the particles merely oscillate to and fro with no
vertical displacement as the wave passes overhead. The major (horizontal) axes of
the orbits decrease as the water depth increases until the orbits become circular when
the water is very deep. In this case the orbit radius decreases very rapidly (exponentially) with depth below the surface: at halfthe sea depth the radius is only about 4%
of its surface value. Fig. 3.11 shows the relationship between the profile of a deep
water wave and the circular orbit of a particle at the surface.

86

Regular waves

[Ch. 3

f\~~Zt
Fig. 3.11 -Orbit of a particle at the wave surface.

3. 7

PRESSURE FLUCTUATIONS UNDER A WAVE

The pressure at any point under a regular wave may be found from Bernoulli's
equation (3.3). If we assume that the velocity is small we obtain

P=pgy+P
so the pressure at any depth y metres oscillates around the steady hydrostatic
pressure pgy kN/metre 2 The fluctuating part of the pressure is
cosh[k(d- y)] .
P = - pg(,0 cosh(kd) sm (kx- rot)

kN/metre

(3.34)

In deep water this becomes

P = - P8so exp (- ky) sin(kx- rot) kN/metre 2

(3.35)

and in shallow water

P = - P8so sin (kx- rot) kN/metre 2

(3.36)

Fig. 3.12 shows the variation of the pressure amplitude beneath a 100 metre long
wave for three different depths. In very shallow water the pressure amplitude is
constant and everywhere equal to the hydrostatic pressure associated with the
surface wave amplitude. In deeper water the pressure amplitude decreases with
depth and becomes negligible at the bottom in very deep water.

3.8

ENERGY OF A REGULAR WAVE

The energy associated with a train of regular waves includes contributions from both
potential and kinetic energy. Consider a small length of the regular wave shown in
Fig. 3.13. The surface depression(, is given by equation (3.10) and the mass of water
over the length is approximately - p(, per unit width of the wave. The centre of
gravity of this mass is approximately - S/2 metres above the undisturbed surface
level

ox

ox

ox

Energy of a regular wave

Sec. 3.8]

87

Pressure amplitude (kN/metre 2 )

10

yld

Depth
2m

Depth
20m

'-----'----11.0

Fig. 3.12- Pressure amplitudes; A.= 100m, 1;,0 = 1 m.

Mass/unit

width=-Q~

bx

Fig. 3.13- Potential energy in a wave.

ox

and its potential energy relative to the undisturbed (calm water) state is lpgl,? per
unit width of the wave. If we now allow
to become infinitesimally small we may
integrate to obtain the total potential energy summed over a single wave length:

ox

Regular waves

88

EP =

J"!pg~2
0

pg~ii"-

d.x =

pg~6
J" sin
2

[Ch.3

(kx- rot) d.x =

joules per metre width of wave

(3.37)

Consider now a small element of fluid beneath a wave as shown in Fig. 3.14. The

...

I.
I

~------------------~------r-----------_.--~~x

~~M

l
d

-l:,J~u
.;.,<.~....., - - - - - - .

YI~------------~xv------------.~v

Bottom

Fig. 3.14- Kinetic energy in a wave.

mass of the element per unit width of the wave is p ox oy and it has a total velocity q
given by
(3.38)
So the kinetic energy of the particle is
!pq 2

ox oy

joules per metre width of wave

U we now allow ox and oy to become infinitesimally small we may integrate to obtain

Energy transmission and group velocity

Sec. 3.9]

89

the total kinetic energy of the fluid in one wave length between the surface and the
bottom:

Ek = lp

J:(Jr~ q dy) dx
2

joules/metre

Substituting equations (3.22)-(3.25) in equation (3.38) gives

and, after some manipulation, the kinetic energy is found to be

Ek = pgdijA. joules per metre width of wave

(3.39)

So the potential and kinetic energies are equal and the total energy in one wavelength
is
~2A.

E = pg 0 joules/metre
2

(3.40)

which leads to the remarkable result that the average energr per square metre of sea
surface is independent of the wave frequency and depedds only on the wave
amplitude:

E = pg~6 joules/metre2
2

3.9

(3.41)

ENERGY TRANSMISSION AND GROUP VELOCITY

The energy associated with a sequence of regular waves is transmitted along the
direction of their propagation. The rate of energy transmission can be found by
considering the energy flux across the plane AA in Fig. 3.15. We begin by calculating
the rate at which the fluid on the left of~ small element of height oy is doing work on
the fluid on the right of the element.
Since 'the element is small the p-ressure and velocity acting on its face may be
regarded as constant (at a given time) and the force exerted by the fluid on the left is
P oy kN per metre width of the element. The work done by the fluid on the left i:;
uP oy joules per second per metre width. If we now allow oy to become infinitesimal
the total rate of transmission of energy across the plane AA is obtained by integrating

90

[Ch. 3

Regular waves

A
X

u
y

Bottom

AI.-

..

'._,. ':.::.~.5:>-
:::'~..<.:~~<~
..,
.

--=-~-~.

.~,..

. ' . .-:::..;~~
..... .

Fig. 3.15- Energy transmission in a wave.

over the depth of the fluid. Neglecting the small contribution due to the portion of
fluid above the undisturbed surface level (y = 0), the rate of transmission of energy is

E = Jd

uP dy

joules/second per metre width of wave

(3.42)

()

Using equations (3.22) and (3.35) we find that the energy is transmitted at a rate

E = pg2L,5k sin2(k.x- rot)Jd cosh2[k(d- )] d


ro cosh 2(kd)

_ pg

2
s_o_

k___
sinL(kx,_ J-, rot) {d Y dy
1

W l:U:SH ~K-U)

J ()

joules/(metre second)

After some reduction this becomes

. _ pg(,~c (
2kd ) . 2
E- -2- 1 +sinh (2kd) sm (kx- rot)
pg2L,okd2
.
ro cosh (kd) sm(kx- rot)
2

joules/(metre second)

Energy transmission and group velocity

Sec. 3.9]

91

The rate of transmission of energy evidently fluctuates with time but we are
concerned with its mean value. Over a long period of time (or an integral number of
wave periods) the mean value of sin 2 (kx- rot) is! and the mean value of sin (kx- rot)
is zero. So the mean rate of energy transmission is

-'-

pgs6c ( 1
2kd )
+ sinh(2kd)

E =-4-

joules/(metre second)

(3.43)

Now the total energy is given by equation (3.41) and this energy is transmitted at a
mean velocity given by

;.,t 0

c(

E
2kd )
= E = 2 1 + sinh ( 2kd)

metres/second

This relationship is plotted in Fig. 3.16. In deep water

0~--~--_.--~~--~~--~_._.~
0.01
0.02
0.04 0.06- 0.1
0.2

d/i,

Fig. 3.16- Group velocity.

(3.44a)

92

Regular waves

2kd

[Ch.3

=0

sinh (2kd)
and

u0 = - metres/second
2

(3.44b)

In shallow water
2kd
sinh (2kd)

1.0

and
u0 = c metres/second

(3.44c)

For deep water we may interpret this result by considering the progression of a
group of regular waves down a laboratory tank. If the energy associated with each
wave length is E joules/metre the amplitude of the waves is, from equation 3.40,

so=~(:~)

metres

(3.45)

Each individual wave within the group is propagating forward at a velocity c


metres/second but the energy is only propagating at c/2 metres/second. So after one
wave period each wave will have moved forward one wave length, taking half its
associated energy with it. It follows that half the energy of each wave must be left
behind to be added to the energy brought forward by the next wave. In this way the

total energy per square metre within the group is kept constant.
At the leading edge of the group the first wave will be propagating into calm
water. So this orderly exchange of energy from wave to wave is interrupted and after
one wave period the energy of the leading wave is halved. The wave amplitude is
reduced and this process continues as the leading edge of the wave train propagates
down the tank at the wave celerity.
The leading edge of the group proper (defined as the position of the first wave of
full amplitude given by equation (3.45)) propagates down the tank at u0 metres/
second and this velocity is called the group velocity. Individual waves within the
group propagate at the wave celerity c, which in deep water is twice the group
velocity (see equation (3 .44b)).

4
Ocean waves

4.1

WAVEGENERATION

As explained in the previous chapter, a knowledge of the characteristics of regular


waves is an important asset to the naval architect. However, such waves do not occur
in the real ocean environment and this chapter is concerned with the characteristics
of naturally occurring 'real' waves.
The waves which are of most concern are those which arise in the ocean through
the action of the natural wind. Other wave generation mechanisms exist but are of
little practical importance except in special circumstances.
The mechanism by which wind-driven waves are formeQ is not wholly understood
or of particular interest to the average reader of this book~l Suffice it to say that a
steady wind blowing over an open stretch of calm water will create ripples which will
travel across the surface in more or less the same direction as the wind (see Fig. 4.1).
Wind

c>

0
~

(a) calm surface

(b) Ripples

(d) Ripples develop on growing waves

Fig. 4.1- Wind-generated ~~ves.

~
(c) Small waves

94

Ocean waves

[Ch. 4

If the wind continues to blow for long enough and sufficient length of water or 'fetch'
is available, the ripples will advance and grow in length and height until they can
more properly be called waves. At the same time, the wind generates new ripples on
the surface of the growing waves and these ripples will eventually grow into waves
themselves. The process is of course continuous and the observed waves at any
particular place and time will consist of a mixture of wavelengths and heights
superimposed on each other.
The individual wave components apparently still behave in the same way as they
would in ideal conditions, uncontaminated by waves of other lengths. Thus the fast
moving long waves continually overtake the slow moving short waves and the shape
of the surface is changing all the time as the waves progress through each other.
Clearly the waves are absorbing energy from the wind. This energy absorption is
countered by two principal decay mechanisms: wave breaking and viscosity. If the
wind continues to blow at constant velocity for long enough and sufficient fetch is
available, the rate at which energy is absorbed by the waves will eventually be exactly
balanced by the rate of energy dissipation and a steady state 'fully developed' wave
system will be achieved. Such wave systems are rare because the required steady
conditions do not often presist for long enough and the fetch may be limited by the
local geography.
If the wind ceases to blow, the wave system it has created will gradually decay.
Since wave breaking is a relatively powerful decay mechanism, the short steep
waves, which are more likely to break, decay first, leaving the longer waves to be
dissipated by the relatively weak forces of viscosity. This decay process may last
several days, during which these fast moving long waves may travel several thousand
kilometeres and be recognised at some distant location as a swell. Swells are
generally of long period and comparatively regular. Locally generated wave systems
may therefore be contaminated by swells generated elsewhere. These swells will of
course bear no relationship to the local wind.

4.2

STATISTICAL ANALYSIS OF TIME HISTORIES OF IRREGULAR


WAVES

Whatever the complexities of local geography and the vagaries of the wind, an
observer at sea will see a confused (and confusing) pattern of ever changing wave
crests and troughs travelling in different directions. For many years this apparent
chaos (and the resulting unpredictable nature of ship motions) provided an insur_.......,. .... ...,..f.....,.\....1...,. ..-..1-..n+.n..-.1..-..

UlUUUlaUH;;;

VU;)la\,J~

4-,..., -..--rr...-.o.C'Oco ~ ...... +).,a.

ha.l....J ..-...f roonlrooT"\~T"liT

LV }J.lVt;lV.:l.:) JJJ Ul\,.; 11\..-lU VJ. .:)\.;Un."-''-'1'1115

"LJ"\.'1.'1.70'1.7Q'I"

.L..I.VVY\,.IV\,.;.1'

~n

rOI"'OT"li-

J.H .l\.1\,.;\.;UL

ua.r:arco

J""U.l.:>

considerable progress has been made in the application of statistical methods to


quantify the characteristics of the waves on the sea surface, and these methods form
one of the foundations of the modern science of seakeeping.
Fig. 4.2 shows part of a typical record of wave elevation obtained from a wavesensing device in the ocean. As expected, the record is irregular in nature and no
coherent pattern is obvious.
It is customary to define four basic measurements:
wave amplitude

Sa (metres)

the vertical distance from the mean water level to a


peak or a trough (a peak below the mean level or a

Sec. 4.2]

Statistical analysis of time histories of irregular waves

95

TH

1-.-----------------------~~------------------------~~

T,
~(metres)

*Negative amplitude

'rl'I
Fig. 4.2- Typical wave record: analysis of peaks and troughs.

wave height Ha (metres)


wave period TP (seconds)
wave period Tz (seconds)

trough above the mean level gives a negative amplitude; otherwise amplitudes are always positive)
the vertical distance from a trough to a succeediog
peak or vice versa (always positive)
the time between two successive peaks
the time between two successive upward or downward zero crossings.

These individual measurements are unique to the particular part of the record
chosen for analysis and are of little use for characterising the whole time history. So it
is customary to describe the general ch;uacteristics of the complete time history in
terms of the mean values of these quantities:
~a

mean value of many measurements of ~a (metres)

Ha
TP

mean value of many measurements of Ha (metres)


mean value of many measurements of TP (seconds)

96

Ocean waves

Tz

[Ch. 4

mean value of many measurements of Tz (seconds)


Two additional quantities are also used:

~113

significant single amplitude: mean value of highest third of many measure-

H 113

significant wave height: mean value of highest third of many measurements of

ments of Sa (metres)

Ha (metres).
They are related as follows:
H 113

= 2.0 ~113 metres

In addition to the statistical measures associated with peaks, troughs and zero
crossings there exists another class of measurements used to quantify the characteristics of an irregular wave record. Here the time history is sampled at discrete (short)
intervals of time to obtain successive measurements of the surface depression ~n
relative to some arbitrary datum as shown in Fig. 4.3. For a typical irregular wave
'
record an appropriate time interval would be 0.5 or 1.0 second.

Timet (seconds)

l;, (metres)

Fig. 4.3- Typical wave record: analysis at successive time intervals.

These measurements enable three important quantities to be derived:


~

mean surface depression,

97

Fourier analysis

Sec. 4.3]
N

L~

(4.1)

metres

n=l

(where N is the number of observations of surface depression)

m0 = variance of surface depression relative to the mean,


N

L (sn- ~)

n=l

metres 2

cr0

(4.2)

standard deviation or root mean square (rms)


depression relative to the mean

= ym0 metres

(4.3)

In passing it should be noted that a sensible analysis using either of these


techniques requires a wave record containing at least 100 pairs of peaks and troughs.
Such a record will be typically of about 20-30 minutes duration. Shorter records run
the risk of yielding unreliable results because they may, by chance, be unusually
severe or unusually moderate. The record illustrated in Figs 4.2 and 4.3 would
therefore not be of adequate length for analysis.

Very long records of, say, several hours should also be avoided. This is because it
is quite likely that real changes in the wave statistics would occur in this time due to
changes in the wind speed or the arrival of swells from distant 1~torms.

"

4.3

FOURIER ANALYSIS

The coptinuous process of wave generation (and the typical form of an irregular wave
record) suggests that any given time history of length, say, T H seconds might
reasonably be represented by the Fourier series
I

s(t)

~ + LAn cos (ront) + Bn sin (ront) metres

(4.4)

n=l

where the frequencies are given by

(n

= 1, 2, 3, ... , oo)

The coefficients are given by

radians/second

(4.5)

Ocean waves

98

[Ch. 4

An

JTH
0 s(t) cos ((J)n t) dt

metres

(4.6)

Bn

= T2

JTH s(t) sin ((J)n t) dt

metres

(4.7)

TH

Equation (4.4) may be written as

~+

L Sno cos (ront +en)

metres

(4.8)

n=l

where the coefficients are


(4.9)
and the phase angles are given by
(4.10)
In physical terms equation (4.8) may be interpreted as representing the irregular
wave record by the sum of an infinite number of sine waves of amplitude Sno and
frequency ron- These frequencies have been chosen (equation (4.5)) so that one cycle
or period of the lowest frequency ro 1 corresponds to the length T H of the record.
Similarly the record length corresponds to two cycles of the second sine wave, three
of the third and so on. The interval between the frequencies is

21T

TH

radians/second

(4.11)

and oro becomes very small as TH becomes very large.


The individual sine waves are staggered with respect to each other. The phase
angles en which define the stagger are related to the time origin. ~' the mean value of
the record, is often made zero by judicious choice of the datum level of the
measurements.
Any given time history could be analysed in this way if it were actually possible to
calculate the required infinite number of sine wave amplitudes, frequencies and
phase angles from equations (4.5), (4.9) and (4.10). The resulting synthesised time
history (equation (4.8)) would match the original exactly over the time interval T H
There would, however, be no guarantee that the synthesised time history would
match the original outside the range of the recording (indeed it would be most

The wave energy spectrum

Sec. 4.4]

99

surprising if this were the case). The synthesised time history will repeat itself at
intervals ofTH whereas the original (real) time history will never repeat itself.

4.4

THE WAVE ENERGY SPECTRUM

It is customary to express the relative importance of the component sine waves


making up an irregular wave time history (equation (4.9)) in terms of a wave
amplitude energy density spectrum (usually abbreviated to the more easily managed
wave energy spectrum). The energy per square metre of the sea surface of the nth
wave component is ipg/;~0 (equation (3.41)): the wave amplitude energy density
spectrum is defined so that the area bounded by a frequency range (say roa to rob as
shown in Fig. 4.4) is proportional to the total energy (per square metre of sea surface)

""0

c
0

QJ

.~
""0

;;:
(/)

~
Q)

I.

Fig. 4.4- Definition of wave energy spectrum.

of all the wave components within that range of frequencies. It follows that the total
area enclosed by the spectrum is proportional to the total energy per square metre of
the complete wave system.
If we set

Ocean waves

100

ffia

[Ch. 4.

8ro
(J)-- radians/second
n
2

and
(J)b

8ro

(J)n

+2

radians/second

there will be only one component frequency ron in the range IDa to rob. Actually a real
irregular wave would also contain components at other frequencies within this range,
but the Fourier analysis technique does not identify them explicitly. Instead their
effects are amalgamated in the single frequency identified by the Fourier analysis.
The wave amplitude spectral density ordinate corresponding to this frequency ron is
given by

so that the spectral ordinate is

~~0 metres2J(radian/second)
28ro

(4.12)

A wave energy spectrum corresponding to any irregular wave time history can be
derived in this way and a typical example is shown in Fig. 4.5(a). The spectrum is
discontinuous and consists of a series of rectangles of width oro. The area of each
rectangle is proportional to the energy attributed to that frequency band and
represented by the corresponding single sine wave component.
If the wave energy spectrum is known it is possible to reverse this process and
generate a corresponding time history by adding a large number of component sine
waves according to equation (4.8). In this case the sine wave amplitudes are obtained
by rearranging equation (4.12) to give
(4.13)
It is also necessary to specify the phase angles en and these should be chosen at
random. An infinite number of choices is possible and each will give a different time
history. Nevertheless, all the time histories will have the same wave energy
spectrum. One possible choice would be the phase angles corresponding to the
original wave time history from which the spectrum was derived. Only in this case
will the synthesised time history be identical to the original time history.
In principle an infinite number of sine wave components are required but
acceptable results can be obtained with a limited number. The form of the wave
energy spectrum can be used as a guide to choosing an appropriate range of

Spectral moments

Sec. 4.5]

101

3
(b)

=cc:
0

(.)

Q)

"'

'0

lfffllllllllllffl

N(/)

Q)

E
8

;;;
0
Wave frequency''' (radians/secon9)

Fig. 4.5- Wave energy spectra: (a) typical spectrum from Fourier analysis of irregular wave
time history; (b) typical line spectrum corresponding to time history synthesised by summing
sine waves.

frequencies. Clearly components corresponding to large spectraJ ordinates must be


included but little will be lost by omitting very high and very low frequencies if their
contributions to the spectrum are small.
The use of a limited number of component sine waves may give an apparer;ttly
1
acceptable time history but it should be remembered that the energy spectrum
actually being realised is not the original spectrum but a series of infi:o.itely high spikes
of infinitesimal width at each of the chosen frequencies as shown in Fig. 4.5(b ). The
synthesised time history contains no energy at frequencies between those chosen for
the synthesis.
4.5

SPECTRAL MOMENTS

The definition of variance given in equation (4.2) can be .written as

(4.14)

Ocean waves

102

[Ch. 4

if the time history has a zero mean and the number of observations is very large. A
time history represented by equation (4.8) therefore has a variance

JTH ( 2: sno cos (ron t + en) )2 dt


00

mo =

T1
H

n=l

metres 2

(4.15)

Since the frequencies are chosen in accordance with equation (4.5) this reduces to

2: s~o

n=I

metres 2

(4.16)

and, from equation (4.13),

2: s~ (())) oro

mo

n=I

(4.17)
So the variance of the irregular wave time history is equal to the area under the wave
amplitude energy density spectrum.
The time history given by equation (4.8) can be differentiated to obtain the
vertical velocity and acceleration of the sea surface:

~(t)

~(t)

2: - snoron sin (ront+ en)

metres/second

(4.18)

metres/second 2

(4.19)

n=l

n=I

-sno<D~ cos

(ront+en)

These can be regarded as irregular wave time histories in their own right and can be
analysed to obtain statistics of velocity and acceleration in exactly the same way
as for surface elevation. The amplitudes of the component sine waves are now
Sno<Dn metres/second and Sno<D~ metres/second 2 respectively.
These velocity and acceleration time histories can be analysed to produce
corresponding velocity and acceleration energy spectra. By analogy with equation
(4.12) the spectral density ordinates are

Spectral moments

Sec. 4.5]

S~(ro)

4r2
Wn'onO

S~(ro)

28ro

103

ro~S t; (ro)

metres 2/second 2/(radianlsecond)

(4.20)

ro~St; (ro)

metres2fsecond 4f(radian/second)

(4.21)

So the velocity and acceleration spectral densities can be obtained by multiplying the
amplitude spectral density by appropriate powers of the frequency.
By analogy with equation (4.17) it is clear that the area under the velocity and
acceleration spectra must be equivalent to the variances of velocity and acceleration
respectively. The variance of velocity is

J: s~

m2

(ro)dro

(4.22)
and the variance of acceleration is

J:

',)
'I

ro St; (ro) dro

metres /second

(4.23)

m 2 and m 4 are called spectral moments since they can be considered as moments of
area of the amplitude spectrum about the vertical axis. In general

(4.24)
and n may take any positive integer value (n = 0,1,2, ... ).
The average frequency can be found by determining the centre of area of the
spectrum from

(!)

ml

radians/second

and the corresponding average period is

(4.25)

Ocean waves

104
-T = 21Tmo

seconds

[Ch.4

(4.26)

It can also be shown (Ochi and Bolton (1973)) that the mean period of the peaks is

TP =

21T

~ (::)

seconds

(4.27)

and the mean zero-crossing period is

Tz =

27T

~ (::)

seconds

(4.28)

Strictly speaking equations (4.27) and (4.28) are valid only if the surface depression
measured at equal intervals of time is normally distributed (see Chapter 17). In
practice this assumption is invariably true for real ocean waves.
Fig. 4.6 shows two irregular wave time histories, and sketches of the corresponding wave energy spectra are shown in Fig. 4.7. The 'narrow band' time history of Fig.
4.6(a) could loosely be described as a sine wave of varying amplitude, and the origin
of the terminology is clear from the appearance of the spectrum: the wave energy is
concentrated in a narrow band of frequencies and little or no energy is present at
other frequencies. One property of this form of time history is that any peak is almost
invariably followed in orderly succession by a downward zero crossing, a trough, an
upward zero crossing and another peak. Peaks below the datum level are very rare
and it follows that the average period of the peaks is almost the same as the average
zero-crossing period.
The 'wide band' time history contains energy over a wider band of frequencies as
shown in Fig. 4.7(b). In this case there are many peaks and troughs which are not
immediately followed by zero crossings and the average period of the peaks is very
much less than the average zero-crossing period. There are many peaks below the
datum level and many troughs above the datum level.
The ratio between the average period of the peaks and the average zero-crossing
period can be regarded as a measure of the 'narrow handedness' of the time history
and its wave energy spectrum. The 'bandwidth parameter' is defined by

(4.29)
and values of E lie in the range 0 to 1: E = 0 corresponds to a very narrow banded
spectrum and E = 1 corresponds to a very wide banded spectrum.
It has been shown (Cartright and Longuet-Higgins (1956)) that the significant

Sec. 4.5]

Spectral moments

105

t (seconds)
L,

(metres)

~-----=------.-It

(seconds)

(metres)

T,

'1'1

'I

Fig. 4.6- (a) Narrow and (b) wide band time histories.

wave amplitude can be related to the area under the wave energy spectrum. In terms
of significant wave height the relationship is
,. 1

H 113

= 4.00

~ (1-;) Vm

metres

(4.30)

so that
metres

1f s

= 1 (wide band spectrum)

(4.31)

4.00ym 0 metres

if s

= 0 (narrow band spectrum)

(4.32)

2.83 Vm 0
and

Ocean waves

106

[Ch. 4

(a)

E=O

1'1

I\
OJ

._

(radians/seconds)

(b)

c){

t> (radians/second)

Fig. 4.7- (a) Narrow and (b) wide band spectra.

Thus, as expected from the general appearance of Fig. 4.6(b), the significant wave
height of a wide banded time history is relatively small.
,
It is often convenient to assume that e = 0 for real wave systems, and equation
(4.32) is assumed to apply so that the significant wave height can readily be estimated
by integrating the wave energy spectrum. In fact e is usually of the order of 0.5 and
this practice results in an overestimate of the significant wave height.
4.6

IDEALISED WAVE ENERGY SPECTRA

In general the wave energy spectrum derived from an analysis of an irregular wave
record obtained at a particular place and time in the ocean will be a unique result that
will never be repeated. Although it may be a useful guide to likely wave conditions,
its use for ship design purposes is strictly limited and it is customary to rely instead on
families of idealised wave spectra. Current practice is to use different formulae for
open ocean and coastal (limited fetch) conditions.

Idealised wave energy spectra

Sec. 4.6]

107

The Bretschneider or ITTCt (two-parameter) wave energy spectrum formula is


appropriate to open ocean wave conditions and is given by

SB~;(ro)

= : 5 exp

~~)

metres2f(radian/second)

(4.33)

where

= 691
f4

lil
1'

172.75

metres 2/second4

(4.34)

second- 4

(4.35)

The 'two parameters' are the characteristic wave height H 1 and the average period T
(equation (4 .26)). We shall see that the characteristic wave height is often assumed to
be the same as the significant wave height.
The spectral moments of the Bretschneider spectrum are
00

m0 =

exp (

=- =

4B

~~) dro

0.0625Hl metres 2

ooA

(4.36)

'I

2.916

'fl

ro exp

(-B)
(1)4

= ~r(O)
where r is the gamma function.
Hence, from equation (4.36),
t International Towing Tank Conference.

lil
fZ

metres 2/second 2

(4.37)

dro
oo _ metres 2/second 4

(4.38)

Ocean waves

108
H1

4.00 ym0

[Ch.4

metres

(4.39)

and the characteristic wave height may be related to the area under the spectrum.
The mean zero-crossing period (equation (4.28)) is

-Tz = 21T \jf(mo)


mz

21T

= 4 V(1TB) = 0.92T seconds

(4.40)

and the mean period of the peaks is (equation (4.27))

TP = 21T ~ (::) =

0 seconds

(4.41)

The bandwidth parameter (equation (4.29)) is


E

1.0

(4.42)

and the Bretschneider spectrum is therefore extremely broad banded. These results
imply that a true realisation of a time history corresponding to a Bretschneider
spectrum (including all frequencies up to infinity) would have countless tiny ripples
of infinitesimal period superimposed on the more visually obvious large-scale wave
structure as shown in Fig. 4.6(b). These ripples are responsible for reducing the
significant wave height from the narrow banded value (4.0 vmo) to the wide banded
value (2.83 vmo)
While this result is mathematically correct, the infinitesimal ripples which reduce
the significant wave height have no discernible effect on the large-scale visual
appearance of the waves or the ship motions which are caused by them. In practice a
time history realisation of a Bretschneider spectrum covering a finite, but,adequate,
range of frequencies of practical importance would not include these tiny ripples, and
a practical analysis of the record to find the significant wave height would include
only visually obvious peaks and troughs. Such a synthesised time history would
invariably have a narrow banded appearance with TP = Tz so that the significant
wave height derived from the record would be much more neariy given by equation
(4.32) than by equation (4.31).
Comparing equations (4.32) and (4.39) it is seen that
(4.43)
and indeed the characteristic wave height for the Bretschneider spectrum is often
loosely referred to as the significant wave height.
The modal period T 0 of the spectrum corresponds to the frequency ro 0 of the

Idealised wave energy spectra

Sec. 4.6]

109

peak, which may be obtained by differentiating equation (4.33) and setting the result
to zero. It is found that

4.849

radians/second

(4.44)

and the modal period is

21T
roo

l.296T

= 1.41Tz seconds

(4.45)

(see equation (4.40)). The corresponding peak spectral density ordinate is


S8 ~;(ro) 0

= 0.01846H{T
= 0.01425H[T0 metres 2/(radian/second)

(4.46)

Equations (4.40) and (4.45) may now be used to define the constants A and B
more fully:

li[

R[

172.75 T 4

487.3 Tti

li[

123.8 'f4

metres2/second 4

(4.47)

' d

1949

691

Tti

'f4

495

'I

-="4 secondsTz

(4.48)

It should be emphasised that the relationships between the periods (equations


(4.40) and (4.45)) are not general and apply only to the special case pf the
Bretschneider spectrum. Fig. 4.8 shows some specimen Bretschneider wave energy
spectra for a characteristic wave height of 4 metres and various model periods. As
expected from equation (4.39), the area under each spectrum is the same since the
characteristic wave height is the same in each case. The position and height of the
peaks depend on the modal period. Fig. 4.9 shows Bretschneider spectra for a modal
period of 10 seconds and various characteristic wave heights.
In coastal waters where the fetch may be limited the JONSWAP (Joint North Sea
Wave Project) spectrum is used:
S1 ~;(ro)

0.658CS 8 ~;(ro)

metres:?f(radianlsecond)

(4.49)

where S8 ~;( ro) is the Bretschneider wave spectral density ordinate (equation (4.33)).
The factor C is given by
-

Ocean waves

110

[Ch.4

5P-------~~-------r--------~------~

20 sec

4
"0

c
0

(J
Q)
(/)

15 sec

.~
"0

(/)

Q)

U./

.s
"'c

~
0

t5Q)

a.
(/)

Q)

>

s"'

Frequency w (radians/second)

Fig. 4.8- Bretschneider wave energy spectra; characteristic wave height 4 metres.

exp

[-=--!
(roTo -1)
2y
21T
2

c = 3.3
where

2
]

(4.50)

Sec. 4.6]

Idealised wave energy spectra

0.07

for

0.09

for ro

-ac

(J)

>

111

21T
To

(4.51)

<21T
To

10

(.)

Modal frequency

Ql
(/)

cco

1:l

~
N

(/)

~
Q)

E
.,
(/)

u;
c

Ql

1:l

~
(.)

Ql

c.
(/)
Ql

>

co

2.0
Frequency w (radians/second)

Fig. 4.9- Bretschneider wave energy spectra; modal period T0 = 10 seconds.

The JONSWAP spectrum is thus a distortion of the Bretschneider spectrum


specified in terms of the characteristic wave height and the modal period. Fig. 4.10
'!'I
'I

T0 =10sec

:::::
1:l
c
0

(.)

3l

~JONSWAP

-~

1:l

(/)

~
Q)

...
c}{

0
, Frequency tJ (radians/second)

Fig. 4.10- JONSWAP and Bretschneider spectra; significant wave height 4 metres.

I,

Ocean waves

112

[Ch.4

shows a comparison between the two spectra for a characteristic wave height of 4
metres and a modal period of 10 seconds. The effect of the additional factors in the
JONSWAP formula is to increase the height of the peak of the spectrum. There is,
however, a corresponding reduction in the spectral ordinates, on either side of the
peak and the areas enclosed by each spectrum are the same since the characteristic
wave heights are the same.
4. 7

WAVE SLOPE SPECTRA

When considering the effects of waves on the angular motions of ships (pitch, roll,
yaw, etc.) it is often convenient to express the energy of the wave system in terms of a
wave slope spectrum rather than the conventional wave amplitude spectrum already
discussed.
We have seen in Chapter 3 that the slope of the surface of a regular sine wave also
varies sinusoidally (equation (3.18)) and that the wave slope amplitude is

cx0 =

k~ 0

(equation (3.19))

radians

In deep water the wave number


(1)2

metres- 1

so that the wave slope amplitude of the nth component sine wave becomes
radians

(4.52)

The time history of the slope of an irregular wave is also an irregular time history
and can be represented by the sum of an infinite number of sine waves in a manner
analogous to equation (4.8). The amplitudes of these wave slope components are
given by equation (4.52). The wave slope irregular time history has its q.wn energy
spectrum and the wave slope spectral ordinates are given by
(1)4

= -1
S~(ro)
g

radians 2/(radians/second)

(4.53)

(see equation (4.12)).


So the wave slope energy spectrum can be obtained by multiplying the wave
amplitude spectral ordinates by ro~/g 2 The Bretschneider and JONSWAP wave
slope spectra are

Ss"'( ro) =

-A
2

rog

exp

(-B)
-4

ro

radians 2/(radianlsecond)

(4.54)

Sec. 4.8]

Wave spreading

S1 cx(ro)

113

0.658CSBcx(ro) radians 2/(radian/second)

(4.55)

Fig. 4.11 shows examples of these wave slope spectra. The JONSW AP spectrum

T0 =10 sec

............... JONSWAP

"D

c
0

(J

Q)

VJ

c
"'
"D
~

Frequency <> (radians/second)

'tl
'I

Fig. 4.11- Wave slope spectra; significant wave height 4 metres.

is very sharply peaked but the most striking comparison with the shapes of the
corresponding wave amplitude spectra (Fig. 4.10) is the much greater comparative
importance of high wave frequencies. This corresponds with practical observatiorls:
short high-frequency waves are often very steep even though their amplitudes are
very small.
All the relationships derived for wave amplitude spectra (equations
(4.12)-(4.30)) have analogous relationships for wave slope spectra. Thus, for
example, the variance of wave slope can be obtained by integrating the wave slope
spectrum.
r

4.8

WAVE SPREADING

In ideal conditionsin the open ocean the waves might all be expected to travel in the
same direction. However, these 'long crested' waves in: which the infinitely long wave
crests remain straight and parallel are never experienced outside the artificial

Ocean waves

114

[Ch.4

confines of the laboratory towing tank, although approximations may occasionally


be found at sea. It is, however, much more likely that the real waves in the ocean will
be travelling in many different directions, although an easily recognised 'primary'
direction, often more or less aligned with the wind, may be discernible. Changes in
wind direction, the influence of coastlines and bottom topography and the presence
of wave systems originally generated elsewhere will all conspire to ensure that the
true long crested wave system is at least a rarity and probably a myth.
The presence of more than one long crestal wave system results in alternate
enhancement and cancellation of wave crests and troughs, and this phenomenon
gives rise to the term 'short crested' to describe the appearance of a wave system with
a spread of wave directions. So a wave energy spectrum derived from a record of
surface elevations obtained at a particular point in the ocean will invariably contain
contributions from several different wave directions. It is often convenient to ignore
this fact and assume that the wave system is long crested, and for many purposes this
may give acceptable results. However, the degree of wave spreading does have a
profound influence on some ship motions (particularly roll) and its effects cannot
always be ignored.
The amount of wave spreading at any particular time and place is, like the other
characteristics of the wave system, dependent on the immediate past history of the
wind as well as on geographical factors. An infinite number of possibilities exist, but
for design purposes it is usual to assume that if the primary wave direction is JJrelative to some fixed datum (Fig. 4.12), the secondary wave directions v are
distributed in the range - Vmax < v- fJ- < Vmax The directional wave spectrum is
defined such that the quantity pg S~;(ro, v) oro ov is equivalent to the wave energy
contained in the frequency band oro and the directbn band ov as shown in Fig. 4.13.
Hence the directional spectral density ordinate, by analogy with equation (4.12),
is given by

= ~~i0

2 oro ov

metres 2/(radian/second) per radian

(4.56)

where ~njo is' now the amplitude of the component sine wave appropriate to the nth
frequency and the jth direction. For ship design purposes it is assumed that the
directional wave spectral ordinates are related to the ordinates of the equivalent total
wave energy spectrum S~;(ro) by

metres 2/(radianlsecond) per radian

(4.57)

where D is a constant and m is a positive integer. Since the total wave energy is
assumed to be distributed over the range of directions from - Vmax to Vmax it follows
that

Wave spreading

Sec. 4.8]

115

Limit of
speading

Secondary
wave
direction

Fig. 4.12- Primary and secondary wave directions.


',I
'I

J~~~ Ss(co,v) d(v- JL)

metres ?(radian/second)

Putting

v'

we obtain

;ax frr/2

2v

-rr

D cosm(v') S~(co) dv' ~etres 2/(radian/second)

12

(4.58)

116

Ocean waves

[Ch.4

SJ<n,v)

QgSJ<n, v) ll11) llv =energy


in frequency band /)(1)
and direction band bv

(I)

(I)

Primary wave direction

Fig. 4.13- Typical directional wave spectrum.

and

1T

4vm~

f" ,

(4.59)

cosm( v') dv'

()

1
'1 ..
..Vmax

1T

4vmax
1

Vmax
In general

form

form

form

(4.60)

Wave spreading

Sec. 4.8]

1.3.5.7 .... .m
1r
2.4.6.8 .... .(m -1) 4vmax

form odd

2.4.6.8 .... .m
1r
1.3.5.7 .... .(m -1) 2Ymax

form even

117

Fig. 4.14 shows the spreading function D cosm (v') for various values of the
2.0

-----,-----r----"T""-----,
20

1.5

10

1.0

Cn
0

Cl

0.5

Relative Heading V

Fig. 4.14- Wave energy spectrum spreading functions.

Ocean waves

118

[Ch. 4

spreading index; m = 0 represents uniform spreading with equal contributions to the


wave energy from all directions. As m is increased, the energy becomes progressively
more concentrated around the primary wave direction v- /.L = 0. A nearly long
crested sea can therefore be represented either by choosing a high value for m or a
small value for vmax For ship design purposes the most common practice is to use
m = 2 and vmax = 90 to represent short crested seas, and trials evidence (Cummins
and Bales (1980)) suggests that this 'cosine squared' spreading is appropriate for
typically occurring conditions in the open ocean. However, spreading angles as low
as 60 or as high as 120 may frequently be found. For cosine squared spreading with
vmax = 90 equation (4.57) becomes
2
- cos 2(v- /.L) S ~ffi)
1T

metres ~(radian/second) per radian


(4.61)

Equations ( 4.57) and (4.60) are of little direct use in practical computations of
ship motions in short crested seas. These calculations (see Chapter 14) require the
spread wave spectrum to be represented by a discrete contribution from each of a
finite number of secondary wave directions within the range of the spreading. Each
contribution is essentially a scaled down version of the total wave energy spectrum as
shown in Fig. 4.15. If the secondary wave directions are spaced at intervals of ()v the
appropriate wave energy spectrum at each secondary direction is given by

where the weighting factor W is

D cosm ( 1T- (v2Vmax

I.L)) bv

Table 4.1 lists the weighting factors for intervals of bv


angles and indices.

(4.62)

= 15 and various spreading

Sec. 4.8]

Wave spreading

0.083SJw)

-90

Primary wave direction

Fig. 4.15- Representation of directional spectrum at discrete heading intervals of 15; cosine
squared spreading over 90.

119

120

Ocean waves

[Ch. 4

Table 4.1- Weighting factors for calculations of ship motions in short crested seas;
ov = 15
m

V-/1-

(degrees)
Vmax

10

20

0.063
0.125
0.125
0.125
0.125
0.125
0.125
0.125
0.063

0
O.Q75
0.139
0.181
0.196
0.181
0.139
O.Q75
0

0
0.037
0.125
0.213
0.250
0.213
0.125
0.037
0

0
0.003
0.065
0.248
0.368
0.248
0.065
0.003
0

0
0
0.016
0.230
0.508
0.230
0.016
0
0

0
0
0.001
0.146
0.710
0.146
0.001
0
0

0.042
0.083
0.083
0.083
0.083
0.083
0.083
0.083
0.083
0.083
0.083
0.083
0.042

0
0.034
0.066
0.093
0.113
0.127
0.131
0.127
0.113
0.093
0.066
0.034
0

0
0.011
0.042
0.083
0.125
0.156
0.167
0.156
0.125
0.083
0.042
0.011
0

0
0
0.008
0.043
0.119
0.206
0.245
0.206
0.119
0.043
0.008
0
0

0
0
0
0.011
0.080
0.239
0.338
0.239
0.080
0.011
0
0
0

0
0
0
0
0.027
0.236
0.473
0.236
0.027
0
0
0
0

0.031
0.063
0.063
0.063
0.063
0.063
0.063
0.063
0.063
0.063
0.063
0.063
0.063
0.063
0.063
0.063
0.031

0
0.019
0.038
0.054
0.069
0.081
0.091
0.096
O.Q98
0.096
0.091
0.081
0.069
0.054
0.038
0.019
0

0
0.005
0.018
0.039
0.063
0.086
0.107
0.120
0.125
0.120
0.107
0.086
0.063
0.039
O.D18
0.005
0

0
0
0.002
0.010
0.033
0.073
0.124
0.167
0.184
0.167
0.124
0.073
0.033
0.010
0.002
0
0

0
0
0
0.001
0.008
0.040
0.115
0.209
0.254
0.209
0.115
0.040
0.008
0.001
0
0
0

0
0
0
0
0
0.009
0.073
0.241
0.355
0.241
0.073
0.009
0
0
0
0
0

= 90o
-90
-75
-60
-45
-30
-15
0
15
30
45
60
75
90

Vmax

= 60o
-60
-45
-30
-15
0
15
30
45
60

Vmax

= 120o
-120
-105
-90
-75
-60
-45
-30
-15
0
15
30
45
60
75
90
105
120

5
Ocean wave statistics

5.1

INTRODUCTION

Chapter 4 described how an idealised wave energy spectrum may be defined in terms
of the significant wave height and various measures of the average wave period. This
allows representative spectra to be constructed for any point in the ocean provided
that these quantities are known. Of course many different combinations of significant wave height and average period may occur at any- particular point. For practical
ship design purposes we need to choose appropriate values for the sea areas and
seasons in which the ship is expected to operate.
This chapter reviews the available sources of wave data.

'I!

5.2 VISUAL OBSERVATIONS


5.2.1 Sea state code
The description of the mechanism of wave generation in Chapter 4 shows that thefe
can be no unique correlation between wind and wave height. Nevertheless, mariners
have traditionally used the visual appearance or 'state' of the sea :as an indication of
the local wind speed. This led to the concept of a numerical scaie of sea state as a
measure of the severity of the waves and differeQ.t scales were evolved by different
national authorities. These scales have often been used to report sea conditions in
preference to more precise estimates of wave height and period.
In 1970 the World Meteorological Organisation(WMO) agreed the standard sea
state code given in Table 5.1. Each sea state number corresponds to a range of
significant wave heights and there is noindication of period. As such, the sea state
can be regarded at best as a rather- vague and indeterminate indication of wave
conditions which is of only limited use in reporting sea conditions. Nevertheless, its
use is so well established and widespread in the seafaring community that naval
architects and oceanographers must sometimes tole_Eate its deficiencies.

Ocean wave statistics

122

[Ch. 5

Table 5.1- World Meteorological Organisation sea state code


Sea state code

0
1
2
3
4
5
6
7
8
9

Significant wave
height (metres)
Mean
Range
0
0-0.1
0.1-0.5
0.5-1.25
1.25-2.5
2.5-4.0
4.0-6.0
6.0-9.0
9.0-14.0
Over 14.0

0
0.05
0.3
0.875
1.875
3.25
5.0
7.5
11.5
Over 14.0

Description

Calm (glassy)
Calm (rippled)
Smooth (wavelets)
Slight
Moderate
Rough
Very rough
High
Very high
Phenomenal

5.2.2 Visual observations of wave height and period


With encouragement from oceanographers, sailors now often report sea conditions
in more detail by estimating the 'average' wave height and period. Hogben and
Lumb (1967) compared these visual observations with values measured by wave
buoys and other suitable instruments and obtained the following approximate
relationships:

H 113 = 1.06 Hobs metres

(5.1)

Tz = 0. 73

'fobs

seconds

(5.2)

T0 = 1.12

'fobs

seconds

(5.3)

Nordenstrom (1969) derived alternative expressions:


()0 75
H
metres
113 = 1.68 Hobs

(5.4)

Tz = 0 82 ('!'obs)0 96

(5.5)

seconds

If we asume the Bretschneider spectrum period relationships (equations (4.45))


Nordenstrom's period relationship can also be written as

T.0 = 1 16 ('!'obs )0 96

seconds

(5.6)

These relationships are illustrated in Figs 5.1 and 5.2. It may be concluded that
observers' estimates of average wave height correspond reasonably closely to the
significant wave height. Since the true mean wave height must by definition be less

Sec. 5.2]

Visual observations

123

Observed wave height Robs (metres)

Fig. 5.1- Significant and observed wave heights.

(ij

16

""0

c
0

(.)

QJ

.!!!.

"'

It-:

""0

""0

(.)
Q)

.:::

QJ

c.

OJ

""0
0

u;
2

-~

<J)

c.
co

(.)

""0

:2:

QJ

co
QJ

Observed period fobs (seconds)

Observed period fobs (seconds)

_,

Fig. 5.2- Measured and observed periods.

124

Ocean wave statistics

[Ch. 5

than the significant wave height, this implies that observers ignore the smaller waves
when making their estimates.
Average visual estimates of wave period apparently agree quite well with the
modal period, but Hogben and Lumb found that individual estimates were often
widely scattered and could not be regarded as reliable.
5.3

WAVEATLASES

5.3.1 Visual observations


A very comprehensive atlas based on over 55 million visual observations from ships
on passage between 1854 and 1984 was published as Global Wave Statistics by
Hogben, Dacunha and Olliver (1986). This superseded the earlier work by Hogben
and Lumb (1967). The new atlas covers virtually the entire globe and gives the
probabilities of occurrence of significant wave heights and zero-crossing periods for
all the sea areas shown in Fig. 5.3.

180

150

120

90

60

30

30

Go

90

120

150

160

150

120

90

60

30

30

60

90

120

150

180

Fig. 5.3- Sea areas used by Hogben, Dacunha and Olliver (1986). (Reproduced by permission
of British Maritime Technology Ltd.)

The reliability of the raw visual observations of wave height was enhanced by
correlating them with simultaneous observations of wind speed. This allowed
unrealistic estimates of wave height to be eliminated from the data base. The
unreliable visual estimates of the wave period were not used at all. Instead, wave
period statistics were constructed from correlations with measured data.
Table 5.2 shows a typical set of data from Area 9 (west of the British Isles) in
winter. The data are subdivided into different wave directions and are presented in
the form of scatter diagrams, giving the joint frequency of occurrence (in parts per
thousand) of particular combinations of significant wave height and zero-crossing

Table 5.2- Wave height and period statistics. (After Hogben, Dacunha and Olliver (1986).) Reproduced by permission of British
Maritime Technology Ltd.)

(/)
~

DECEMBER TO FEBRUARY

AREAS

!'
Ut

NORTH

NORTH WEST
PERCENT AGE OF OBS -

-..!.

73 20& 21!17 232 123

11

287.

41!

14

PERCENTAGE OF 085 ._ 6. 28,.


4 1000

9-1 0

' ''
'' "' ""
" u
' "n u
u
n"
" '
I

7-8
6-7
5--6

,_,
,_,
,_,

17

17
I

1-2
0-1

4-5
<4

IS

8-7

5-6

S-9

7-11

I
I
I

I
I

I
I
I

10-1 I

'' -..! ,
G
TOTAL

2) 11) 241

27!t 198

97

37

11

PERCENTAGE OF OBS
4 1000

''
'

13-1

G 12-1
,,_,

:;;

II

...

:I:

TOTAL

NORTH EAST

13-1

'

'' ''
-I
-, "
,."
'
'
u
' - ""
'
""
"'
'
"',.
'
5

10
\4

.
7

\4

...

I
I
I
I

2-

191
220

I
I

''

12-1
11-1

...
~

' '' '' '


'' ' "" " '
" "" "
'" "" "
" " '
'
I
I

!1-9
7-!1
6-7
5-6

. . ..
. .
7

20

75

0-1

4-5

12-13
TOTAL
11-12
>13

<4

20

7
7

80
50

76
~- ~ ____!~ __!}_
B-7
8-9

5-6

10
20

~ ;:!
~ ~=;
Ill

t0-1t

9-10

I
I

I
I

~ 1~=~ I)

7-B

I
I
I

''
''

''
"
""
7

I
I

76

"'

197
250
208

"

10-11
12-13
TOTAL
9-10
11-12
>13

n~

:!
--'
~
~ ,_, ' t
--- ----- --~
- ~
-- ''

1~1

12-1
11-1
10-1 I

~
~
~

5--6

~
~
~

~~

<4

44 154 265 2511 163

- -- -

--- -- --I

4.

58~

26

II

' "%"'
I

I
I

75
~

7--8

3 1000

-- -- --- --- '


'' ' - -- "'
' '
- .,"
" ' ' --

. .
.

I
I

'' "
u"
,_, '"'

10

. "18

I
I
I

115
187

I
I

'"
'".,

- -I

tn

I
I

11

I?

1"1.

9-10

11-12
ZERO CROSSINC PEqJOO (s)

ZERO CROSSING PERIOO { s)

lf:RO CROSSINC PE!IIOO ( s)

74

- ---- --' '


" "R"
" R

TOTAl

'"

ALL DIRECTIONS
WEST

!:

,
13-1
12-1

.'
I
0

ll-9

...

7-8
6-7

5-6

4-5

J-4

2-3
1-2
0-1

...
"'

"'

'
i '~=~
11-1

42 1511 21!11 270 158

I
I
I

''5
' '
' " """
" " "
' "
____-_ _
,_ "
" "
7
\4

52

72

4-5

<4

18

8-7

5-6

I
I
I

I
I

IS

8-9

''
u"
n
7

I
I
I

il

...

'

13-1
12-1
11-1
10-1 I
0

-,_,

201

, 1-1

''

48

t4

4-5

8-7

5-6

I
I
I

7-8
6-7
5--6

..
18

,_,
,_,

,_,

1-2
0-1

I
4-5

<4

7
17

8-7

..

07
70

18
I

8-9

7-!1

"
""
"'
'"

I
I

I
I

'

'
'
'
,_,
'' "' ""
l-
,_,
'" "" nM ""
'
,, ~' ,_, "

72

199

1-2
0-1

"'

10-11
12-13
TOTAL
9-10
11-12
>13

'
'

u"

7
8

I
I
I

I
I

2-

IU

7
I

...

10-11
12-IJ
TOTAL
5-6
7-8
9-10
11-12
>13
ZERO CROSSlNC PERIOD (s)

,,

~
~
~

1)-1
12-1
1 ,_,

17

91 219 276 214 115

''
'

47

15

2
5

10

5-6

IB

10

27

70

81

1-2
0-1

4-5

11-9

<4

6-7

5~

I
B-9

7-B

17

17

I
I
I

"
""
,

0-11

fll

114

187

PERCENTAGE OF OBS 6 1000

I
I

'
' ''
.
.. .. ".' '''
I

7-8
6-7

I
I

'
''
' " "
,.' " "
' " " ,." " "'
'
I

45

~
~
tD

' '
'' " ''' '
"" "" ''
"u "'

'"
70

12-13

7-8
9-10
11-12
Zf.:RO CROSSINC Pf.:IUOO (s)

41~

10-1 I

9-1

..

~ ,_,
,_,
,_,

16

3 1000

TOTAl
>13

SOUTH EAST

PERCENTAGE OF OBS "" 19.


1

TOTAL

IJ

29

"

13-1
12-1

7-8
6-7
5-6

..

45

7!1

'I
,_,
\1 10-1
,_, 0

118

45

<.4

1000

' '
' ''
'

" ' ' '


'
"58
' "" " u" '
' - "'"
' "" " "45 ""
_,_
"' '
'"
"
"" " "'
"
I
I

I
I
I

39 14!1 261 259 16!1

ZERO CROSS INC PERIOO ( ~)

10-1 I
9-1 0

TOTAL

SOUTH

zes zJo tn

>I I
~ 13-1
12-1

I
I
I

16~

PERCENTAGE OF OBS ,. 6.

6 1000

52.75

1-2
0-1

SOUTH WEST
12

IJ

10-11
12-13
TOTAL
9-10
11-12
>13

PERCENTAGE OF OBS -,; 19. 76?.


1

I
I

'' "
' ''
' H" "R "" "" ''
' q" "n " " " '
'' u" "' " "' '
I

ZERO CROSS INC PERIOO ( s)

TOTAL

20

,_,

"

85

.' .' ''


I
I

,_, '

0-7

"'
"'

44 159 277 285 1511

,_,
,_,

181

'

,,

7-B

50
76

110

27

TOTAL

5 1000

I
I
I

10

16
I

7-8

IS

' ''
''
" "' '''
" '

..
.. .. .. ""'
I
I

82

EAST

PERCENT AGE OF OBS - 100 00,.


(INCLUDING 1 09:'. DIRECTION UNKNO\\t<)

PERCENTAGE OF OBS- 22.04?.

50

17

I
I
I

''
'

il

~
~

""

235

81

"'
'"
"
160

10-11
12-13
TOTAL
9-10
11-12
>13

,,
1)-1

31

133 252 265 179

!1!1

9. 40~
34

11

4 1000

:I : : : : : : : : : : :

:;=:u-10-11
9-10
B-9
7-8
6-7
5-6
4-5
3-4
2-l
1-2
0-1

<4

1
2
4-5

1
4
14
11

1
2

5-6

ZERO cqOSSINC PERIOD (s)

TABULATED PROBABILITIES ARE IN PARTS PE~ THOUSAND OF THE POPULATION IN EACH TABLE

1
4
1)
35
61
18
B-7

1
3
7
19
44

1
1
2
5

1
1
2
4
7
12
1!1
21
1&
5

1
1
1
2
4
5
7
7
4
1

1
1
2
2
2
2
1

3
5I
9
17

7
IJ
1
JJ
18
HI
1
S2
34 31
1 115
82 46
11)6
114 84 48
275
!ll 52
19
238
11
3
1
45
!1-9
10-11
12-13
TOTAl
7-8
i-10
U-12
>13

Zf.:RO CROSSING PERIOD (s)

......

N
Ut

126

Ocean wave statistics

[Ch. 5

period occurring simultaneously. For example, the probability of occurrence of wave


heights from all directions in the range 4-5 metres with periods in the range 9-10
seconds is 58/1000 = 0.058.
The frequency of occurrence of waves from each of the specified directions is
given as 'percentage of obs' at the top of each scatter diagram. Also shown at the
right-hand side of each diagram is the frequency of occurrence of each significant
wave height range for all periods. The frequency of occurrence of each period range
for all wave heights is shown at the top of each diagram. For example, the probability
of waves of any period from the west having a significant wave height falling in the
range 4-5 metres is 14911000 = 0.149.
The probability of the significant wave height exceeding a given level is obtained
by summing all the observations above that level. For example, for waves from the
west the probability of the significant wave height exceeding 7 metres is

P(B113 > 7) =(50+ 32 + 20 + 12 + 7 + 4 + 3 + 4)/1000 = 0.132


Fig. 5.4 shows the probability of exceeding specified significant wave heights for
four different sea areas. Clearly Area 9 has one of the most severe wave environments in the world, closely followed by the North Sea. The Gulf of Mexico is
particularly benign.
A similar wave atlas based on visual observations for the North Pacific, with
particular emphasis on the seas around Japan, was published by Takaishi, Matsumoto and Ohmatsu (1980).
Wave statistics based on visual observations must always be considered less
reliable than direct measurements of wave conditions even if the reliability has been
enhanced as described above. However, a more serious criticism of visually observed
wave data lies in the fact that ships' masters generally try to avoid bad weather and
this is likely to introduce a fair weather bias into the results. So the published tables
are likely to underestimate the probability of extremely severe weather conditions
simply because ships' masters will try to avoid storms. It may be argued that this bias
makes the statistics applicable to ships which have some freedom to avoid bad
weather. Ships which must remain on station, such as warships and offshore
platforms, may well experience bad weather more frequently than these statistics
would imply.
5.3.2 Hindcasting
Bales, Lee and Voelker (1981) published a wave atlas for the North Atiantic, North,
Mediterranean, Black and Baltic Seas. They used measured wind data obtained over
a period of twenty years for the sea areas in question to 'hindcast' the waves which
would have occurred as a result of the measured winds. The prediction technique
used was devised by Pierson, Tick and Baer (1966) and calculations of wave spectra
were made at six-hourly intervals for the period 1959-1969 for the sea areas shown in
Fig. 5.5. A total of over 133 000 wave spectra were calculated and compiled to
produce statistics in much the same way as shown in Table 5.2.
In addition to data on wave heights and periods, Bales' atlas also gives information on wave direction and wind speed and direction. The atlas also contains

Wave atlases

Sec. 5.3]

127

Area 9 : North East


Atlantic

Area 11 : North Sea

Q)

c:

"'

""0
Q)
Q)

Q)

Area 18:
Sea of Japan

.~
:.0

"'0

..Q

a':

Area 32 : Gulf of
Mexico
0.001

12

Significant wave height R, 3 (metres)


Fig. 5.4- Probability of exceeding significant wave heights. (After Hogben, Dacunha and
Olliver (1986)).

limited data on visibility, cloud cover, precipitation, relative humidity, air and water
temperatures, sea level pressure and ice. Table 5.3 shows an example of a scatter
diagram for wind speed and significant wave height for the entire North Atlantic.
Lee, Bales and Sowby (1985) have also published a similar atlas for the Pacific
Ocean.
The hindcast technique a'.(oids the problems of accuracy and fair weather bias
associated with visually observed wave data but depends, of course, on the accuracy
and reliability of the mathematical model used to predict the wave conditions.

Ocean wave statistics

128

[Ch. 5

Fig. 5.5- Sea areas in the North Atlantic. (After Bales, Lee and Voelker (1981)).

Table 5.3 - Annual wind speed and wave height statistics for the North Atlantic;
probabilities in parts per thousand. (After Bales, Lee and Voelker (1981))

Fl 113 (m)
>24
20-24
16-20
14-16
12-14
10-12
9-10
8-9
7-8
6-7
5-6
4-5
3-4
2-3
1-2
0-1

+
+
+
+
+
1
2
5
17
41

+
+
+
+
+
1
2
4
12
34
68

Wind speed (knots)


17 22 28 34

11

+
+
+
1
1
4
10
26
69
92

+
+
+
1
2
5
12
28
66
23
24

+ Indicates less than one part per thousand.

+
+
+
+
1
3
7
16
37
67
16

+
+
+
1
2
7
14
30
43
10

+
+

+
+
+
1
2
5
11
21
15
2

+
+

41

+
+
1
2
5
8
8
3
1

+
+

48

55

+
+
+
2
2
2
1

+
+
+
+

1
1

+
+
+
+
+

Total

+
+
+
+
+
+

+
1
4
6
11
18
32
52
80
125
187
259
225

Wave atlases

Sec. 5.3]

129

5.3.3 Measured wave data


If both visually estimated and hindcast wave data are subject to uncertainties, direct

measurements of wave spectra must provide the most reliable data of all. However,
measuring wave data over a protracted period (years) is an expensive and complicated undertaking and few attempts at systematic data collection have been made.
Probably the most comprehensive is that organised by the US National Oceanographic and Atmospheric Administration (NOAA) and publishd by Gilhousen et al.
(1983). Fig. 5.6 shows the locations of the wave-measuring buoys around the United

Fig. 5.6- Locations of NOAA wave bouys. (After Gilhousen et al. (1983)).

, II

States. All of the buoys were deployed for at least three years and some have been in
continuous operation for as long as nine years. The buoys recorded information on
air and sea temperature and atmospheric pressure as well as wihd and waves. The
waves were sampled every three hours and a wave spectrum derived from the
recorded time history. The significant wave height and mean zero-crossing period
were derived using equations (4.32) and (4.28) and scatter diagrams similar to those1
shown in Table 5.2 prepared.
,
Many other measurements of wave conditions have of course been made for
specific purposes at various locations throughoutc the world. Typically these are
relatively short-term studies intended to provide data on the local environment for
use in research or specific projects such as ship seakeeping trials or the design of
offshore or harbour installations. Much of the data have been acquired by commercial organisations who regard them as proprietary information not available to the
general public. However, in 1982 the United Kingdo~ Marine Information and
Advisory Service (MIAS) published a catalogue listing the data sources open to
general use. Over 1350 entries were catalogued and their locations are shown in Fig.
5. 7. The majority of the measur~ments have been made in the coastal waters around
the British Isles and in the North Sea (see Fig. 5.8), but a.significant quantity of data
are also available for North American and Australian\vaters.

130

Ocean wave statistics

[Ch.5

Fig. 5.7- Availability of measured wave data. (From Marine Information Advisory Service
Catalogue of Wave Data (1982).)

10W

131

Wave atlases

Sec. 5.3]

0E
I

5W

60N

5E
60N

...

J$..:'

10W

5W
REGIONAL MAPS

0E

5E

BRITISH ISLES AND ADJACENT COASTS

Fig. 5.8- Sites of measured wave data around the British Isles. (Reproduced by permission of
Institute of Oceanographic Sciences.)

, II

6
The spring-mass system

6.1

INTRODUCTION

We shall see in Chapter 8 that the behaviour of a ship in rough weather is


fundamentally similar to the oscillatory response of the classical damped springmass system illustrated in Fig. 6.1. So an understanding of the characteristics of the
spring-mass system is a good basis for the study of ship motions .

. J.

Mass a

Force F

Stiffness c

Displacement

x,...

Fig. 6.1 -Damped spring-mass system.

The spring-mass system consists of a mass a which is connected to a fixed rigid


base (say the Earth) through a dashpot and a spring. If the system is not disturbed it
will adopt an equilibrium position which we shall define as a datum displacement

x= 0 metres

Harmonic response

Sec. 6.2]

133

A varying force F is applied at the free end and we require to find the resulting
(varying) displacement x of the mass relative to its undisturbed equilibrium position.
Each of the three components of the system absorbs a proportion of the applied force
so that, at any instant of time,
spring force + dashpot force + mass force

=F

kN

We assume that the spring has no mass, contributes no damping and obeys
Hooke's law so that the spring force is directly proportional to the displacement x.
Similarly we assume that the dashpot has no mass and no stiffness and that the
dashpot force is directly proportional to the velocity x. Finally we assume that the
mass contributes only inertia to the system so that the mass force is directly
proportional to the acceleration .X. So at any instant of time

ai+bx+cx=F kN

(6.1)

where a is the mass in tonnes (or kN/(metre/second 2 )), i.e. the force required, to
accelerate the mass at 1 metre/second 2 to the right; b is the dashpot damping in
kN/(metre/second), i.e. the force required to extend the dashpot at a rate of 1 metre/
second; and cis the spring stiffness in kN/metre, i.e. the force required to extend the
spring by 1 metre.
This kind of system is known as a 'second-order linear system', 'Second-order'
implies that equation (6.1) contains terms up to the second derivative (.X) but nothing
of higher order. 'Linear' means that each of the component forces is directly
proportional to the appropriate derivative of x. In other words there are no terms
involving powers like x 2 , x 3 , etc.
' II
Real systems may not be truly linear: for example, it is possible to have a stiffness
which increases progressively as the spring is extended as shOWijl in Fig. 6.2. Another
common form of non-linearity is the hard limit caused by mechanical stops which
limit the spring's extension. Non-linear damping is also possible and, in particular,
the dashpot may be fitted with a relief valve to limit the damping force at ~Ofiile
predetermined velocity.
These inconvenient properties complicate the behaviour of spring-mass systems
and it is usual to assume that real systems are linear where this can reasonably be
justified. Equation (6.1) can then be used to examine the behaviour of the system.
Fortunately the linear assumption can be justified for many problems of ship motions
(with certain well defined and specific exceptions) and the results we shall obtain are
directly relevant to the study of ship behaviour in rough weather.

6.2

HARMONIC RESPONSE

Let us suppose that the force F varies in a sinusoidal manner with amplitude F0 kN
and frequency co radians/second:

F = F0 sin(co t)

kN

(6.2)

The spring-mass system

134

[Ch.6
Non-linear

Linear

Hard
limit

Progressive
increase in
stiffness

Displacement x (metres)

.:.!

(a) Spring stiffness


Q)

LL.

Slope=b
(kN/metre/second)

Velocity

'Relief valve'
limit

x (metres/second)

(b) Damping
Fig. 6.2- Linear and non-linear characteristics.

We might expect that the resulting motion of the mass would also be sinusoidal and it
is indeed found that

x = x0 sin (rot+ e) metres

(6.3)

is a solution of equations (6.1) and (6.2). x 0 is the motion amplitude in metres and sis
a phase angle in radians. In other words a sinusoidally varying force applied to a
linear damped spring-mass system will result in a sinusoidally varying displacement
at the same frequency. In practice the phase angle e is found to be negative so that the
displacement sine wave lags the force sine wave as shown in Fig. 6.3. The maximum
(positive) displacementx0 occurs s/ro seconds after the maximum (positive) force F0
Differentiating equation (6.3) gives the velocity and acceleration of the mass:

Sec. 6.2]

135

Harmonic response

(metres)

t (seconds)

T=2rr
(0

Fig. 6.3 -Sinusoidal response to a sinusoidal force acting on a linear damped spring-mass
system.

x= x ro cos (rot + c)
0

.X =

metres/second

(6.4)

- x0 ro2 sin (rot + c) metres/second2

Substituting in equation (6.1) gives

(6.5)

'd

x 0 sin (rot) (- aro cos c- bro sin c + c cos c)+


x0 cos (rot)- aro 2 sin c + bro cos c + c sin c)= F0 sin (cvt) kN
The motion x thus has two components: the terms in sin (rot) are in phase with the
force and the terms in cos (rot) are out of phase, or in quadrature, with the force.
1
These two components may be separated to give two distinct equations:
'

x0

( -

aro 2 cos c - bro sin c + c cos c) = F0 kN

x0

( -

aro2 sin c + bro cos c + c sin c) = 0 kN

After some reduction these two equations yield for the motion amplitude
(6.6)
and for the phase

The spring-mass system

136

tan e =

- bro
c- aro 2

[Ch.6

(6.7)

Let us assume for the time being that there is no damping so that b = 0. Then the
motion amplitude becomes infinite at the undamped natural frequency ro. given by

~ ( ~)

ro* =

radians/second

(6.8)

Equations (6.6) and (6.7) may now be written

(6.9)

and

(6.10)

where the non-dimensional frequency or tuning factor is

(6.11)

and the non-dimensional damping or decay coefficient is


b

(6.12)

Fig. 6.4 shows the amplitude and phase responses for a second-order linear
damped spring-mass system according to equations (6.9) and (6.10) for various
values of the decay coefficient 11 At zero frequency the applied force is steady and
the damping and inertia have no effect because there is no velocity or acceleration.
The displacement is governed only by the spring stiffness:

Xo

F0 c

metres/kN

At A= 1.0 the force due to the spring stiffness exactly balances the force due to the
inertia of the mass. The amplitude response is then

Harmonic response

Sec. 6.2]

137

Q)
CJ)

c
co

Q)
(/)

co

0:

-150-

-200~--------------~--------------~

Fig. 6.4- Response of a second-order lin~ar spring-mass system.

(X)
Fo

=1-

_Q

A=l.O

metres/kN~

(6.13)

2T}C

or 11211 times the zero-frequency response. When there is no damping the amplitude
becomes infinite at the -undamped natural frequency, as we have already seen.
However, for finite damping the maximum amplituge occurs at a lower frequency,
called the damped natural frequency, given by

The spring-mass system

138

[Ch.6
(6.14)

or

co0 =co. y'(1- 2TJ 2 )

radians/second

(6.15)

and the maximum amplitude response is

(Xo)
Fo

max-

1
2TJC y'(1- TJ 2 )

metres/kN

(6.16)

When 11 is small the damped and undamped natural frequencies are almost the same,
as shown in Table 6.1. For larger values of TJ the differences become more

Table 6.1 - Peak responses of a spring-mass system

TJ
0
0.05
0.10
0.25
0.50
0.707

Ao
1.0
0.997
0.990
0.935
0.707
0.0

(Xo)
Fo

A=LO

(Xo)
Fo

max

00

00

10/c
5/c
2/c
1/c
0.707/c

10.01/c
5.03/c
2.07/c
1.15/c
1.00/c

appreciable until the maximum response occurs at zero frequency when 11 exceeds
0.707. The system is said to be critically damped when TJ = 0.707.
_
At higher frequencies the amplitude response falls towards zero regardless of the
decay coefficient or the spring stiffness. Physically this corresponds to the situation
where the oscillation is so rapid that the system has insufficient time to respond
appreciably.
Fig. 6.4 also shows the phase response of the system. At very low frequencies the
phase is nearly zero and the displacement x is almost in phase with the applied force
F. In other words the system responds more or less instantaneously to the slowly
varying force. As the frequency is increased, the displacement begins to lag behind
the force and the phase 10 becomes negative. As might have been expected, the lag
increases with damping, showing that a well damped system responds sluggishly to
the applied force. The phase is always - 90 at the undamped natural frequency
regardless of the damping. At higher frequencies the lag increases still further and
tends to 180 at infinite frequency.

Free decay

Sec. 6.3]

6.3

139

FREE DECAY

Let us now suppose that the spring-mass system is deflected to some initial
displacement x 00 and then released. We require to examine the subsequent motion.
Since there is no applied force after the system is released, F= 0 and equation (6.1)
becomes

ai+bx+cx=O

(6.17)

kN

We might expect that the resulting oscillation would resemble a 'sine' wave with a
continually decreasing amplitude. In fact the response

= x 00 exp

~ t) cos (rod t)

(6.18)

metres

is a solution of equation (6.17). Substituting equation (6.18) in equation (6.17) and


separating 'in-phase' and 'quadrature' components as before leads, after some
reduction, to

c-

~'t + a (_!_'t2 - ro~) = 0

kN/metre

(6.19)

and the time constant is given by


I

2a
1
't = - = - seconds
b 11ro*

II
(6.20)

Combining these two equations gives for the oscillation frequency


(6.21)
So the frequency of the decaying oscillation, lies between the damped and
undamped natural frequencies (see equation (6.15)). In practice, if 11 is small the
differences are negligible.
Fig. 6.5 shows the free decay of a linear damped spring-mass system for various
values of the decay coefficient 11 When 11 = 0 there is no damping and the oscillation
continues indefinitely with no loss of amplitude because there is no mechanism for
energy dissipation. As 11 increases, the oscillations decay more rapidly until they
effectively disappear after only a single cycle when 11 is greater than about 0.5.
The free decay of an oscillaJion may be used to estimate the decay coefficient of
the system. If 11 is small the decaying oscillation frequency rod is almost the same as
the undamped natural frequency ro., and free decays are often used to obtain an
estimate of the system's natural frequency. So the peaks and troughs occur at

140

[Ch.6

The spring-mass system

~ [\ [\ [\ [\ (j [\_, "

V VlJVV V\-

~Ac,? .,~o e
x~4-I-(\+V-tA~\J~c=>.~. . . ,.,. -.=- - -" '-= -~- ~-~

- n=0.1

l "

4-v---~-~=--=------------~- 1]=0.25

L\-'-./~-------------,- 1]=0.5

lL~===---------------,1-

1]=0.707

Fig. 6.5- Decay of oscillations in a linear damped spring-mass system.

intervals of rr/roct
time

Jrr

tJ = -

:!;

rr/ro* seconds and the Jth maximum (peak or trough) occurs at

seconds

(j)*

(see Fig. 6.6). The corresponding peak amplitude is (from equation (6.18))

XoJ

= x00 exp

-lrr)
(--cro*

metres

= X 00 exp

(- lrrTJ)

metres

Now the ratio between the Jth and the (J + l)th amplitude is
XoJ
Xo(J+l)

exp (- lrrTJ)
exp [- (J + 1)rrTJ] = exp(rrTJ)

System with no stiffness

Sec. 6.4]

141

1
( - XQJ )
l1 =-lo
:rr 9e Xo{J+1)

Xoo

J:r

Fig. 6.6- Estimation of decay coefficient.

and the decay coefficient is given by

1
( -XOJ-)
11 =-log
rr

(6.22)

Xa(l+l)

'II
The decay coefficient can therefore be estimated from the decaying oscillation by
determining the ratio between any pair of successive amplitude~. When the damping
is very small and the oscillation decays very slowly, several estimates of the decay
coefficient can be obtained from a single record. The method is not really practical
when 11 is much greater than about 0.2 and is in any case strictly valid only for ~m'\ll
values of 11.

6.4 SYSTEM WITH NO STIFFNESS


Fig. 6. 7 shows a related system in which there is no spring stiffness so that the system
only has damping and inertia. The equation of motion is now

ai+bx=F kN

(6.23)

and we again require to find the response of the system to a sinusoidally varying force

F= F0 sin (rot)
A solution is

kN'

The spring-mass system

142

[Ch. 6

Mass a
Force F

Displacement x

Fig. 6.7- Damped system with no stiffness.

x = x 0 sin (rot + s)

metres

so that the application of a sinusoidally varying force again results in a sinusoidally


varying displacement. Following the procedure used for the system with stiffness
leads to the amplitude response
Xo

F0

b 2 ro' y(1 + ro' 2 )

metres/kN

(6.24)

and the phase is given by

1
tans=----,
.
0)

(6.25)

where the non-dimensional frequency is


,

aro

ro=b

(6.26)

Fig. 6.8 shows the amplitude and phase response of the zero-stiffness system
according to equations (6.24) and (6.25). The responses are quite different from
those of the spring-mass system. The absence of stiffness means that there is no
natural frequency and the amplitude rises steadily as the frequency approaches zero.
At zero frequency the steady force F0 is resisted only by the damping force bx (since
the acceleration, after an initial transient, is zero) and the mass moves at a steady
velocity given by
x. = -Fo metresIsecond

System with no stiffness

Sec. 6.4]

143

5a

b2r-~r------------r---------------,

w'

or---------------.----------------,
Iii
Q)
~

Ol
Q)

:5'.
w

Q)

"'

(0

.I::

c._

-200~--------------~--------------~
0
2
w'

Fig. 6.8- Response of a second-order linear system with no stiffness.

Since this velocity continues indefinitely, the amplitude x 0 is infinite at zero


frequency. In practice this ideal amplitude may not be achieved because of mechanical constraints such as limits on the dashpot travel.
The phase is - 90 at zero frequency. In other words the displacement lags the
force by 90. This phase lag increases as the frequency increases and tends to - 180
at infinite frequency.

7
Heading and encounter frequency

7.1

HEADING

The ship's heading is defined with reference to the direction of propagation of the
waves. The convention chosen is shown in Figs 7.1 and 7 .2. The ship is assumed to be

Fig. 7.1- Definition of heading angle.

attempting to maintain a straight line track at a constant speed U metres/second


across the sea surface. The waves will cause deviations from the intended course and
track, but a directionally stable ship in the hands of an experienced helmsman (or an
autopilot) will usually be able to follow a sensibly straight mean course so that the
heading angle p, can be readily defined as the angle between the intended track of the
ship and the direction of wave propagation.

Encounter frequency

Sec. 7.2]

145

Following
[l=0

Starboard 1---+1
beam
[t=90

t----1 Port beam


[l=270

Head
[t=180

Fig. 7.2- Headings.

With this definition:


J.1- =
J.1-

= 90

J.1- =
J.1-

oo
270

= 180

corresponds to following waves with the waves and the ship travelling in
the same direction
corresponds to beam waves with the waves approaching the ship from the
starboard side
' II
corresponds to beam waves with the waves approaching the ship from the
port side
corresponds to head waves with the waves travelling in the opposite
direction to the ship.
1

Quartering waves are defined as heading angles between oo and 90 (or 270 and
360). Bow seas are defined as heading angles between 90 and :180 (or 180 and
270).

7.2

ENCOUNTER FREQUENCY

The characteristics of regular waves were discussed in Chapter 3 and it was shown
that the wave frequency ro, with which a train of regular waves would pass a fixed
point in the ocean, is one of the most important wave parameters. Now although this
wave frequency has some direct influence on ship motions, they are also critically
dependent on the frequency wi}h which a moving ship ";ould encounter these regular
waves.

146

Heading and encounter frequency

[Ch. 7

Fig. 7.1 shows a ship heading at an angle JL relative to the direction of propagation
of a train of regular waves. The component velocity of the ship in the direction of
wave propagation is
U cos JL

metres/second

and the waves will overtake the ship with a relative velocity

c - U cos JL metres/second
Since the wave crests are A metres apart, a crest will meet the ship once every Te
seconds, where the encounter period is given by

A
Te = - - - - seconds
c- U cos JL

(7.1)

The corresponding encounter frequency roe is defined as


2rr

2rr
-::;: (c- U cos p,)

radians/second

(7.2)

In deep water this reduces to the fundamental relationship

ro- kU cos JL

ro 2 U
ro - - - cos JL radians/second
g

(7.3)

and this is illustrated in Fig. 7 .3.


In seas forward of the beam (90 < JL < 270) cos JL is always negative and the
encounter frequency is always greater than the wave frequency. In beam waves
JL = 90 and cos JL = 0. The encounter frequency is then equal to the wave frequency
and is unaffected by ship speed.
The situation is much more complicated when the heading lies abaft the beam
(0 < JL < 90 or 270 < JL < 360) cos JL is then always positive and the encounter
frequency now has a maximum value

(l)e max

g
4U cos JL

-----'='---

radians/second

(7.4)

which occurs when the wave frequency is


(I)

2roe max radians/second

(7.5)

Encounter frequency

Sec. 7.2]

147

Fig. 7.3- Encounter frequency and heading.

The encounter frequency is zero when

ro=

---='----cos J.L

= 4roemax

radians/second

, II
(7.6)

Since ro = g/c in deep water this corresponds to the condition when the componeht
velocity is equal to the wave celerity:

U cos J.L

c metres/second

The encounter frequency is negative for higher values of ro. These high-frequency
waves advance only slowly and the negative encounter frequency means that the ship
is overtaking the waves. More precisely.~ a negative encounter frequency means that
the waves are being encountered on their trailing faces, while positive encounter
frequencies mean that the waves are being encountered on their leading faces. This is
of course self-evident on headings forward of the beam where the encounter
frequency is always positive. In following and quartering waves a positive encounter
frequency means that the waves are overtaking the ship.

148

[Ch. 7

Heading and encounter frequency

In following and quartering waves a given (absolute) value of encounter frequency may be experienced in three different wave systems (if Iroe I < roe max) as
shown in Fig. 7. 3. Two of these wave systems will give positive encounter frequencies
and the third will give a negative encounter frequency. If lroel > ffiemax the given
encounter frequency will arise in only one wave system and roe will be negative.
The corresponding wave frequencies can be obtained by rearranging equation
(7 .3) to give

ro =

2U

~os 11- [ 1 ~ (1- 4;e U cos 11-) J

radians/second

(7.7)

The physical interpretation of this phenomenon is best illustrated with a numerical example. Consider a ship steaming at 20 knots (U = 10.3 metres/second) in
regular following waves. Suppose that the ship encounters the wave system at a
frequency

lroel = 0.2

radians/second

It is required to find the wave systems which could be responsible. Possible results,
obtained from equation (7.7), are given in Table 7.1 together with corresponding
celerities and wave lengths.

Table 7.1- Regular wave systems giving lroel = 0.2 radians/second; ship speed 20
knots; following waves

Wave
no.

(J)e

(J)

(rad/sec)

(rad/sec)

1
2
3
(4)

0.2
0.2
-0.2
(- 0.2)

0.285
0.667
1.12
(- 0.17)

c
(m/sec)

A.
(m)

34.4
14.7
8.8

759
139
' 49

Wave no. 1 has crests about 0.75 kilometres apart, but its celerity is very high and
it overtakes the ship with a relative velocity of nearly 47 knots. So the high celerity
compensates for the distant crests and results in the required encounter frequency.
Wave no. 2 is much shorter and slower and overtakes the ship with a relative velocity
of only about 8.5 knots. However, the closer crests compensate for the lower relative
velocity and the wave again gives the required encounter frequency. Wave no. 3 is
very short and the celerity is only 8.8 metres/second. So the ship overtakes this wave
system with a relative velocity of about 3 knots, giving the required encounter
frequency. Again the very low relative velocity compensates for the short wave
length. Wave no. 4 is a trivial result: negative wave frequencies have no physical
meaning.

Sec. 7.2]

Encounter frequency

149

The relationship between encounter frequency and wave length may be further
examined by rearranging equation (7.3) to give

U cos JJ.

2
2~ ~ ~8) roe J
[

metres/second

(7.8)

and this is plotted in Fig. 7 .4. The diagram may be used to find the wavelength

1000

Head

~seas

Beam
seas

U cos J.1 (metres/second)

Fig. 7.4- Encounter frequency and wave length; deep water.

Following
seas ---+

150

Heading and encounter frequency

[Ch. 7

corresponding to any given encounter frequency for a particular speed and heading.
As already demonstrated, a particular encounter frequency is experienced at only
one wavelength (or wave frequency) in head or bow waves; but in following and
quartering waves up to three different wave systems will yield the same absolute
encounter frequency. Fig. 7.4 also demonstrates another peculiar property of regular
waves. In following and quartering waves a wide range of wavelengths may produce
virtually the same encounter frequency. For example, a ship steaming at 20 knots in
quartering waves (JL = 45) has a component velocity of about 7.0 metres/second. In
this condition all the wavelengths from about 50 metres to about 400 metres yield an
encounter frequency close to about 0.3 radians/second. We shall see in Chapter 13
that this phenomenon can have profound implications for roll motions in quartering
seas.

8
Basic equations for ship motions
in regular waves

8.1

INTRODUCTION

Ships do not, in the normal course of events, experience regular waves at sea. So the
study of ship motions in regular waves appears at first sight to be an academic
exercise of no practical significance. Yet it is an essential first step in the calculation of
ship motions in a realistic irregular seaway; moreover; an appreciation of regular
wave motions will give the reader an insight which will prove invaluable in
understanding the general nature of the motions of ships in rough weather.
I

II

8.2 AXES AND SHIP MOTION DEFINITIONS


I

A ship in rough weather experiences a complex sequence of motions as it twists and


turns its way across the ocean surface. The motions seem to defy any rational
analysis, particularly by those who suffer from their effects on board the s\Jip,
Nevertheless, it is possible to make some observations on the characteristics of ship
motions which will help to clarify their nature and will form a basis for the modern
theory of seakeeping.
Let us suppose that the ship is attempting to, maintain a straight course at a
constant speed U metres/,second as shown in Fig. 8.1. Waves continually cause the
ship to deviate from its course and track and the helmsman may find it necessary to
take corrective action. In addition the ship will rise and fall in response to the
changing water level and the deck will eldom be truly horizontal. The ship will
generally follow some kind of spiral path which is more or less aligned with the
intended course. Finally the ship's speed will be continually varying around the
nominal speed U metres/second as the ship surges along its track in response to the
waves.
Any particular ships's track and motion time hi~_tciry can be represented by a

152

Basic equations for ship motions in regular waves

[Ch. 8

Fig. 8.1 -Typical path of a ship in waves.

combination of the time histories of three lineart and three angular displacements.
These six displacements are defined using the right-handed axis system shown in Fig.
8.2.
The axis system Exy has its origin fixed at E at the mean water level and regular
waves propagate along the Ex axis. A second axis system ExE 1 xEzY also has its origin
atE but is rotated through the heading angle JJ- so that ExE 1 coincides with the mean
track of the ship.
A point 0, lying at the mean water level, moves along ExEl at the mean speed of
the ship, U metres/second. (This speed is approximately the same as the ship would
achieve at the same power in calm water. In head waves the speed will be slightly
reduced and in following waves it may be increased.) The mean position of the ship's
centre of gravity G 0 lies vertically above 0 and is taken as the origin of a third axis
system G 0 x 1 x 2 x3 . At any instant of time the position of the ship's centre of gravity G
relative to the moving origin G0 is defined by three linear displacements:
surge x 1 metres: positive forward
sway x 2 metres: positive to starboard
heave x 3 metres: positive down.
The attitude of the ship is defined by three angular rotations about the axes G 0 x 1 ,
G 0 x 2 and G 0 x3 :
t 'Linear' here means a displacement along an axis as opposed to a rotation about an axis. There is no
necessary implication that the motion responses are linear in the sense that they are directly proportional
to a force.

Sec. 8.2]

Axes and ship motion definitions

153

Body axis system

Fig. 8.2- Axes and ship motion definitions.

roll x 4 radians: positive starboard side down


pitch x 5 radians: positive bow up
yaw x 6 radians: positive to starboard.
Most ships have port/starboard symmetry and so surge, heave and pitch, which lie
in the plane of symmetry, are called vertical plane or symmetric motions. Sway, roll
and yaw are termed lateral plane or antisymmetric motions. The motions are often
' (/
referred to as 'degrees of freedom'.
Another right-handed set of axes Gx 81 x82 x83 is fixed in the ship and is used to
define locations on (or in) the ship's structure. The origin is at tqe (moving) centre of
gravity G and the axes rotate as the ship rolls, pitches and yaws. Locations are
defined as:
x81 metres: positive forward
x 82 metres: positive starboard
x83 metres: positive down.
The wave depression at any point xis, according to equation (3.10),
~

~0

sin (k.x - rot)

metres

where the timet may be measured from an arbitrary dat~m. Transforming to the axis
system aligned with the ship's track we find that the wave depression at any point
(xEl, XEz) is
~ = ~0 Sin (k.xEl COS/)-- kXEz Sin/)-- rot)

-c

metreS

Basic equations for ship motions in regular waves

154

[Ch. 8

If we choose a datum time such that the moving origin 0 is at E at time

kU

1T

cos JL

seconds

the moving and Earth-fixed frames of reference are related by

xE 1

XEz

= x2

x 1

+U

(t-

kU ;os

J.L)

metres

metres

y = x3 - OG0

metres

Then the wave depression at (x 1 , x2 ) in the moving frame of reference is


(8.1)

8.3

GENERAL EQUATIONS FOR SHIP MOTIONS IN REGULAR WAVES

8.3.1 Basic equations of motion


The structure of the ship may be regarded as being composed of a large number of
very small masses Dm tonnes. Fig. 8.3 shows one of these masses located at

Fig. 8.3- Accelerations experienced by elemental mass llm.

xB 2 , xB 3 ) relative to the centre of gravity of the ship. If the ship has linear
accelerations i 1 , i 2 and i 3 metres/second 2 and angular accelerations i 4 , i 5 and i 6
radians/second2 the mass Dm will have linear accelerations

(xBu

metres/second 2

forward

General equations for ship motions in regular waves

Sec. 8.3]

i~ = i

i~

2 -

xB3 i 4 + xB 1 i 6 metres/second 2

= i 3 + XBz i 4 -

xB 1

is metres/second

155

to starboard
downwards

The forces and moments necessary to sustain these accelerations are

oFl

om

i~

kN

forward

oF2

om

.X~

kN

to starboard

oF3

om

.X~

kN

downwards

oF4

om XBz X~- om XB3 X~

kN metres

roll moment to starboard

OFs

om XB3 X~- om XBl X~

kN metres

pitch moment bow up

oF6

om XBl X~- om XBz X~

kN metres

yaw moment to starboard

The forces and moments required to sustain the linear and angular accelerations
of the whole ship are obtained by allowing om to approach zero and integrating over
the volume of the ship. Bearing in mind that, by definition of the centre of gravity,

JxBz dm

JxB3 dm

we obtain

mi;

= F;

kN

(i

I44 i 4 - I 4s is - I46 i 6

1, 3)

= F4

- Is 4 i 4 +Iss is - Is 6 i 6
- I 64 i 4 - I6s is + I66 i 6

II

kN metres

= Fs
= F6

kN metres
kN metres

where F; (i = 1, 3) are the surge, sway and heave forces and F; (i = 4, 6) are the roll,
pitch and yaw moments reHuired to sustain the accelerations of the ship. m is the total
mass in tonnes and I 44 , Iss and I 66 are the mass moments of inertia of the ship defined
by

I44

Jcx~z +x~3) dm

tonne metre 2

about the

Iss

Jcx~l +x~3 ) dm

tonne metre 2

about the xBz axis

xB 1

axis

[Ch.8

Basic equations for ship motions in regular waves

156

about the xB 3 axis

The product moments of inertia are defined by

/45

/46

]56

/54

xBJ XBz dm

tonne metre 2

/64

JxBl xB3 dm

tonne metre 2

165

JXBz XB3 dm

tonne metre 2

For conventional ships the product moments of inertia are usually small and are
invariably neglected. The equations of motion then reduce to

mi;
I;;i;

= F;
= F;

kN

(i

kN metres

1, 3)
(i

4, 6)

(8.2)

8.3.2 Motions in regular waves


The forces and moments in equations (8.2) may be applied by any external means,
but we are here concerned with the forces and moments applied to the ship by a train
of regular waves.
For a given hull shape at a particular speed and heading in a given regular wave
system (i.e. a particular wave length) the F; would be expected to be functions of the
displacement, velocity and acceleration of the surface depression and the six possible
motions. So we may write

F; = F; {1;;, ~'

s, (x; .X; i ;(i =

1, 6)}

kN or kN metres

(8.3)

If the wave amplitude is small compared with the wave and ship lengths motions will
also be small and we may use a Taylor series expansion to obtain a linear
approximation to equations (8.3):

L (- a;jij- b;Jj- cijx)


j=!

kN or kN metres

(i

1, 6)

(8.4)

Sec. 8.3]

General equations for ship motions in regular waves

157

where the coefficients a;, a;j, etc are functions of the wavelength (or wave frequency),
ship speed and hull form; a;, b; and C; are also functions of the heading angle.
Substituting equations (8.4) into the six equations (8.2) we obtain six general
linearised equations for small-amplitude motions in regular waves:
6

L (A;jij + b;ij + C;jxj) = Fw;

1, 6)

(i

kN or kN metres

(8.5)

j=l

where

(} = 1, 6; i = 1, 6; j

* i)

and

Aij

= m+a;j

A;j

I;j+ aij

1 3 i = 1 3 ,. = i)
' '
' '
(} = 4 6 i = 4 6 ,. = i)
' '
' '
(} =

and the exciting forces and moments due to the waves are

Fw;

a;~+ b,~ + c,(,

kN or kN metres

(i

= 1, 6)

(8.6)

It is customary to relate all the regular wave ship motions, t9 the wave depression
experienced at the moving origin 0 in Fig. 8.2. In practice tl\e waves here will be
distorted by the presence of the hull (indeed 0 will often be within the hull so that the
surface depression cannot be defined at that point). So, to be precise, the motions are
related to the waves which would have been observed at 0 in the absence of the hull.
Settingx 1 =x2 = 0 metres in equation (8.1) gives for the surface depression at 0

s = so sin (roet)

(8.7)

metres

and the velocity and acceleration of the sea surface perceived by an observer on the
ship at 0 are
~

and

= roeso cos ( roet)

~=

ro~ so sin (roet)

(8.8)

metres/second
metres/second 2

(8.9)

Substituting equations (8. 7.)-(8.9) in equation (8.6) we obtain

Fwi

= Fwio sin (roet + y;)

kN or kN metres

(i

1, 6)

(8.10)

[Ch. 8

Basic equations for ship motions in regular waves

158

where the excitation amplitudes are


(i

1, 6)

(8.11)

and the phases are given by

(8.12)

tan Y;

The equations of motion (8.5) may now be written as


6

L (Aij ij +

b;jxj

+ c;jx)

(i

1, 6)

j=l

(8.13)
Solutions to these equations have the form
X;

X;o

sin (roet + 8;)

metres or radians

(i

= 1, 6)

(8.14)

So small-amplitude regular sine waves impose sinusoidal exciting forces and


moments on the ship and these result in sinusoidal motion time histories, The motion
amplitudes are directly proportional to the excitation amplitudes, which are in turn
proportional to the wave amplitude. The phase angles Y; and 8; relate the excitation
and motion time histories to the time history of the wave depression at 0 as shown in
Fig. 8.4. The peak excitation occurs y/roe seconds before the maximum wave
depression. Similarly the peak (positive) motion occurs 8/roe seconds before the
maximum wave depression. Equations (8.13) are a generalised form of the equations
for the spring-mass system discussed in Chapter 6.
8.4

COEFFICIENTS IN THE EQUATIONS OF MOTION

The theoretical methods outlined in Chapter 9 are usually used to determine the
coefficients in the equations of motion, but these quantities may also be determined
by experiment. It is instructive to examine the techniques involved since this throws
some light on the physical meanings and characteristics of each coefficient.
Full-scale experiments are impractical but experiments with models are a viable
alternative. The coefficients can be measured in forced oscillation experiments in
which the forces and moments applied by the waves are replaced by forces imd
moments applied by some suitable mechanism while the model is towed in calm
water. Fig. 8.5 shows a typical experiment rig in which the model is mounted on two
vertical struts spaced equally about the centre of gravity. If the struts are oscillated in
unison so that the strut motions s3 a and s3 t (positive down) are the same,

Sec. 8.4]

159

Coefficients in the equations of motion

Wave depression
at 0

X;

o;
/w.

Motion

Time

Fig. 8.4- Time histories of wave depression, exciting force and motipn in regular waves.

(8.15)
and the model executes a sinusoidal heave motion. All other motions are restra'ine'd
and the forces necessary to impose the heave oscillation are measured by transducers
at the ends of the struts and recorded on suitable apparatus. The aft transducer is
fitted with a swinging link and this ensures that the longitudinal force is measured in
its entirety by the forward transducer.
The longitudinal force consists of a steady force required to tow the model at the
constant speed U metres/second and an additional oscillatory component due to the
oscillatory motions of the model. We may write for the oscillatory part

where F 10 is the force ampljtude and the heave motionJeads the force by e3 radians.
For an arbitrary shaped hull form the transducers ~ill also experience vertical and

Basic equations for ship motions in regular waves

160

[Ch.8

Fig. 8.5- Forced oscillation experiment.

lateral forces and roll moments. If the motions are small the total forces and
moments imposed on the model may be obtained by appropriate addition and
subtraction of the forces measured by the individual transducers:
Fz = Fza + Fzf = F20 sin (roet)

kN

F3 = F3a + F3f

F30 sin (roet)

kN

F4a + F4f

F40 sin (roe t)

kN metres

F4

(F3a- F3f)

Fs
F6

Xr

- (Fza- Fzf)

= Fso sin (roet)


Xr

kN metres

F6o sin (roet)

kN metres

The motions of the model are related to these forces and mOI~ents by six
equations analogous to equations (8.13):
6

L (A;jij + b;A + C;jx)

F;0 sin (roet)

kN or kN metres

j=l

All the motions except heave are zero and


fori *3

Equations (8.16) then reduce to six much simpler equations

(i

1, 6)

(8.16)

Coefficients in the equations of motion

Sec. 8.4]

161

(i

1, 6)

(8.17)

and these may be recognised as the equations of motion of six second-order linear
damped spring-mass systems with sinusoidal excitation (see equation (6.1)). The
motion response of each system is given by equation (8.15) and the amplitude and
phase response are, from equations (6.6) and (6.7),
1

[y'(ci3- A ;3 COe) 2 + b 5 co~]


(i = 1, 6)

metres/kN or metres/(kN metre)

(i

= 1, 6)

(8.19)

Combining these equations yields for the in-phase and quadrature components of the

six applied forces and moments

Fa

___!.._

cos e; kN/metre or kN metres/metre

(i

X;o

1,6)

(8.20)

and

- F;o sin e;
X;o

kN/metre or kN metres/metre

(i

= 1, 6)
!I

(8.21)

The components of the six applied forces and moments which are in phase with the
heave motion are therefore associated with the stiffness and inertia coefficients,
while the quadrature components are associated with damping.
The coefficients which are of most interest in the heave oscillation experiments
are a 33 , b33 and c33 which relate the heave motion to the applied heave force. Fig. 8.6
shows the physical mechanisms responsible for these coefficients. At zero frequency
the model has no heave velocity or acceleration and the heave forceJis related only to
the heave displacement through the coefficient c33 This arises because a steady
downward heave displacement produces an additional displaced volume and a
steady upward restoring force due to buoyancy. A typical relationship between the
heave displacement and the force is shown in Fig. 8.7(a). Provided that the heave
displacement is small, this may be approximated by a straight line whose slope is c33 .
c33 is specifically defined as the gradient oithe curve as it passes through the origin.
At higher frequencies the in-ph!!.se component of the applied heave force
includes a contribution from heave inertia A 33 This is made up of contributions from
the so called 'added mass' a 33 as well as the real mass m ofthe ship. The former arises
because the accelerating hull causes changes in the fluid velocities adjacent to its
surface as shown in Fig. 8.6(a). The additional force re_quired to accelerate this water

162

Basic equations for ship motions in regular waves

[Ch. 8

(a) Heave acceleration

1J

Fluid

accelerates

'~*

---

Waves radiate
outwards
(b) Heave velocity .,.__

(c) Heave displacement

Fig. 8.6- Effects of heave motions.

as well as the hull is included in the inertia coefficient and the ship behaves as though
it has an increased mass. A 33 is sometimes called the 'heave virtual mass'.
Fig. 8.8 shows the results of some heave oscillation experiments by Smith (1967)
on a model of the Dutch Friesland Class destroyer. The added mass a 33 is about the
same as the mass of the ship over much of the frequency range and rises to even
higher values at low frequencies.
The heave damping b 33 arises because the oscillating ship generates waves which
racliate outward and dissipate energy as shown in Fig. 8.6(b ). Energy is also
dissipated by friction but these effects are very small. Smith also measured b 33 and
some results are shown in Fig. 8.8.
The pitch moment F50 sin (roet) measured in the heave oscillation experiments
will yield estimates of the coefficients a53 , b 53 and c53 These describe the influence of

Coefficients in the equations of motion

Sec. 8.4]

163

(a)

Keel emerges

Heave displacementx3

~
/

Q)Q)

CE

(b)

~0

~E

"''Q)O
mu
_J ...

Slope=O

-Q)

.E
Vertical plane motion

(c)

Lateral plane

moti~nd

Fig. 8.7- Typical force/motion relationships.

heave on pitch in the equations of motion (8.13) and they occur because local inertia,
damping and stiffness forces everywhere along the hull exert pitching moments
about the centre of gravity. If the ship has fore and aft symmetry like a canoe, the
moments arising from the forces on the forward half of the ship will almost exactly
balance those arising from the after half of the ship, and these 'coupling' coefficients
will be very smallt; but for more orthodoxforms residual moments will remain which
may not be negligible.
For arbitrary shaped hull forms forces and moments in the other four degrees of
freedom (surge, sway, roll and yaw) will also be required to sustain a pure heave
~

t They will be exactly zero at zero forward speed.

164

Basic equations for ship motions in regular waves

[Ch.8

Frequency w. (radians/second)

2.5
Frequency

(1) 0

(radians/second)

Fig. 8.8- Heave added mass and damping; Friesland Class destroyer. (After Smith (1967).)

oscillation. However, most practical ship forms have port/starboard symmetry and
all lateral plane excitations associated with motions in the vertical plane are zero. In
other words the relationships between the lateral plane forces and moments and
vertical plane motions have the form shown in Fig. 8.7(b) and all the associated
coefficients are zero.
The apparatus shown in Fig. 8.5 may also be used to induce a pitch oscillation by
oscillating the struts in opposition. Provided that the motions are small the pitch is
given by

Coefficients in the equations of motion

Sec. 8.4]
Xs

Xso sin ( ffiet

+ Bs)

165

radians

where

x 50

===

2s3 a =
Xr

2s3 f radians
Xr

The analysis proceeds along lines exactly similar to those used in the heave oscillation
experiment and yields estimates of the terms A 55 , b55 , c55 , a53 , b53 and c53 Again all
the lateral plane forces and moments and associated coefficients are zero if the hull
has port/starboard symmetry.
The pitch oscillations cause local vertical motions everywhere along the hull so
that each section of the hull experiences local inertia, damping and stiffness forces
analogous to those experienced by the whole model in the heave oscillation
experiment. These forces exert moments about the centre of gravity and are
responsible for the coefficients a55 , b55 and c55 The local forces distributed over the
forward part of the hull oppose those on the after part of the hull so that the residual
heave forces associated with the coefficients a53 , b53 and c53 are usually small. Indeed
they would be zero on a hull with fore and aft symmetry at zero speed.
The pitch virtual inertia coefficient A 55 includes contributions from the so called
added mass moment of inertia as well as the true mass moment of inertia of the ship's
structure. This is analogous to the heave added mas~,already discussed. The true
mass moment of inertia may be expressed as

155

mkg tonne metre 2

(8.22)
'1'/

where the longitudinal radius of gyration about a transverse axis through the centre
1
of gravity is usually in the range
0.2Ls < k 5 < 0.25L5

metres

and Ls is the waterline length of the ship in metres.


Smith (1967) determined the coefficients a55 and b55 for the Friesland Class
destroyer and some of his results are shown in Fig. 8.9. The added mass moment of
inertia is about the sam,e as the true mass moment of inertia over much of the
frequency range tested but rises to higher levels at low frequencies.
Lateral plane forced oscillation experiments may be used to determine the
remaining coefficients in the equations~of motion. Vugts (1968) used these techniques to measure the hydrodynamic characteristics of number of cylinders of shiplike cross-section at zero forward speed. Some of his results are shown in Figs 8.10
and 8.11 for a nearly rectangular cylinder.
The sway added mass (Fig.,. 8.10) exceeds the true:mass at low frequencies but
falls to much lower levels as the frequency increases. The roll added mass moment of
inertia is usually less than about half the true mass moment of inertia.

166

Basic equations for ship motions in regular waves

Frequency

We

[Ch. 8

(radians/second)

Frequency w. (radians/second)

Fig. 8.9- Pitch added moment of inertia and damping; Friesland Class destroyer.
'
(After Smith (1967).)

In most cases the coefficients are found to be nearly independent of the motion
amplitudes used in the experiment and this justifies the assumption of linearity in the
derivation of the equations of motion. Coefficients associated with roll motion,
however, are an exception to this general rule as shown in Fig. 8.11. Roll added mass
moment of inertia decreases and roll damping increases with roll amplitude. So the
assumption of linearity may not be justified in this case.
In general, vertical plane forces and moments will always be required to sustain
motions in the lateral plane even for ships with port/starboard symmetry. However,
the relationship between the vertical plane excitation and the lateral plane motion
for ships with lateral symmetry will have the symmetrical U-shaped form shown in

167

Coefficients in the equations of motion

Sec. 8.4]

Equivalent wave length/beam 1..18

2.0
<le V(B/2g)

2.0 '

VleY(B/2g)

Fig. 8.10- Sway added mass and damping. (After Vugts(1~68).)

Fig. 8.7(c). In other words the vertical plane excitation will have the same magnitude
and direction regardless of the direction of the lateral plane motion. Since we are
concerned only with small motions and our linearisation requires the coefficients to
be determined from the slope at the origin, all such coefficients are zero.
A third category of coefficients is always zero regardless of the ship's shape.
These are all stiffness coefficients associated with 'the ship's geographical location
with respect to the origin G 0 No forces or moments are required to sustain surge and
sway displacements x 1 and x2 so that
cil

c;2

(i

1, 6)

Table 8.1lists 60 coefficients (out of a total of 108) which are zero for a ship with
port/starboard symmetry. A further 12 coefficients are usually negligible and are
invariably neglected.
-

Basic equations for ship motions in regular waves

168

[Ch. 8

Equivalent wave length/beam A./8

100 20 10

54 3

(1)

Y(B/2g)

Fig. 8.11- Roll added mass moment of inertia and damping. (After Vugts (1968).)

This results in six much simpler equations for small-amplitude motions of a ship
with lateral symmetry:
(8.23)
sway:

(m + a 22 )

Xz

+ bzzXz + az4X4 + b24X4

+ a 26 .X 6 + b 26 x6 + c 26 x 6 = Fw2o sin (roet + y2) kN


heave:

(m + a33) .i3 + b33X3 + c3~3

+ a35 .X5 + b35 x5 + c35 x5 = Fw3 o sin (met+ y3) kN


roll:

(8.24)

a 42 .X 2

(8.25)

+ b 42 x2 + (/44 + a44 ) i 4 + b 44 x4 + C44X4

+a 46 .X6 +b46 6 +c46 x 6

Fw4o sin (roet+y4) kN metres

(8.26)

Sec. 8.4]

Coefficients in the equations of motion

169

Table 8.1- Zero value terms in the equations of motion


2

Motionsj

Motionsj

1
2
3
4
5
6
1
2
3
4
5
6

02
03

s
s

s
s

03

02

02

02
02

03
02

03

02

02
01
01
02

03

03

03

03

03

03

01
01

01
01

03
01
01
02

03

02

02
01
01

03
02

03

01
01
02
02

03

Surge
Sway
Heave
Roll
Pitch
Yaw
Surge
Sway
Heave
Roll
Pitch
Yaw

Added inertia a;and damping b;1D40

Stiffness c;1

Surge Sway Heave Roll Pitch Yaw-

Forces and moments i


Key: 0, zero; S, small; 1, by position; 2, by symmetry; 3, for small motions.

pitch:

as3X3 + bs3X3 + Cs3X3 + (Iss +ass) is + bssis

=
yaw:

Fwso sin (roet + y s)

+CssXs

kN metres

a6zXz + b6zXz + a64X4 + b64X4 + (/66 + a66) f6 + b66i~ 4- c66x6


= Fw 6 o sin (roet + y 6 ) kN metres

(8.27)

(8.28)

The vertical plane motion equations (heave and pitch) are coupled as we have
already seen. In other words the heave equation includes terms dependent on pitch
so that heave is influenced by pitch and vice versa. However, the surge equation is
uncoupled and independent of the other motions. The lateral plai).e motions are also
coupled so that these motions are affected by each other. There is, however, no
coupling between the vertical plane motions and the lateral plane motions. So heave,
pitch and surge are not affected by events in the lateral plane; neither are sway, roll
or yaw motions affected 'by heave, pitch or surge. t This allows the vertical and lateral
plane motions to be considered separately.

t This is not necessarily true for large-amplitude motions.

9
Strip theory

9.1

INTRODUCTION

Solving the equations of motion (8.23)-(8.28) requires the evaluation of the


coefficients and the excitation amplitudes and phases. These may be determined by
experiment but this method is laborious and hardly practical for routine calculations.
In any case, if experimental methods are used, it is much more appropriate to
measure ship motions directly as described in Chapter 16.
Considerable effort has therefore been devoted to developing theoretical methods of determining the coefficients and excitations to allow ship motions to be
calculated without recourse to experiment. Various authors, including Tasai (1959),
Gerritsma and Benkelman (1967), Salvesen, Tuck and Faltinsen (1970) and
Schmitke (1978), have made significant contributions. Their theories are generally
similar, differing only in detail and mathematical rigour. They are complicated and a
complete description is beyond the scope of this book. This chapter is intended to
give an abbreviated presentation of the main features of strip theory in general and is
largely based on the methods proposed by Gerritsma and Benkelman.
All the theories assume that:
(a) The ship is slender (i.e. the length is much greater than the beam or the draught
and the beam is much less than the wave length).
(b) The hull is rigid so that no flexure of the structure occurs.
(c) The speed is moderate so there is no appreciable planing lift.
(d) The motions are small.
(e) The ship hull sections are wall-sided.
(f) The water depth is much greater than the wave length so that deep water wave
approximations may be applied.
(g) The presence of the hull has no effect on the waves (the so called Froude-Kriloff
hypothesis).
The theories are grouped under the general heading of 'strip theory' since they all

Sec. 9.2]

Strip motions

171

represent the three-dimensional underwater hull form by a series of two-dimensional


slices or strips as shown in Fig. 9.1. Each strip is of length oxB 1 metres and oxB 1 is
assumed to be small.

Fig. 9.1 -Representation of underwater hull section shape by an infinite cyclinder.

Each strip has associated local hydrodynamic properties such as added mass,
damping and stiffness which contribute to the coefficients for'tije complete hull in the
equations of motion. Similarly the wave excitations experienced by the hull are
composed of contributions from all the strips.
.
Strip theory assumes that these local hydrodynamic propefties are the same as
would be experienced if the strip were part of an infinitely long cylinder of the same
cross-sectional shape as shown in Fig. 9 .1. In other words three-dimensional effects,
such as mutual interference between the strips, flow leakage around the ends of the
ship and effects due to changes in the shape of the strip over the length oxB 1 , are
ignored.

9.2 STRIP MOTIONS


Let us first suppose that the ship is undergoing a generalised forced oscillation in all
degrees of freedom except surge. If the pitch and yaw oscillations are small the
motions of each strip will be essentially confined to the plane of the strip. If the strip is
locatedxBi metres forward of the centre of gravity the motions of a point on the GxB 1
axis will be

172

Strip theory

x~ 0

x4a

= x3 - xB 1 sin x5
== x3 - x81x5 metres

= x4

radians

[Ch.9

to starboard

(9.1)

downwards

(9.2)

to starboard

(9.3)

Consider an observer stationed at some fixed point alongside the ExEI axis in Fig.
8.2. The oscillating ship passes him at a steady velocity U metres/second. At some
instant of time a certain strip is opposite the observer and his perception of its lateral
velocity is given by the total differential of equation (9 .1):

Now the distance x81 from the strip to the approaching centre of gravity is
diminishing at the rate

x81 = -

U metres/second

(9.4)

Hence the lateral velocity perceived by the observer is


(9.5a)
and by a similar argument the perceived lateral acceleration is
(9.5b)
Similarly the perceived vertical velocity and acceleration are
(9.6a)
(9.6b)
and the roll velocity and acceleration are simply

i4o =

x4

radians/second

(9.7)

.X4o =

x4

radians/second 2

(9.8)

An origin 0 in the waterplane, shown in Fig. 9.2, is chosen for calculations of the

Hydrodynamic coefficients

Sec. 9.3)

r---

'
I

173

--1

x~f:G_.....,...,.._.,..
~
x;G

:1

I Waterline

Fig. 9.2- Velocities of a strip.

local hydrodynamic properties of each strip. The velocities and acclerations of this
point are then
x~

= x~ 0 -

OG x~

x~

OG x~

x~ 0

metres/second
metres/second

(9.9)
2

(9.10)

Vertical and roll motions are unaffected by this change of origin:

x3
'I

X3a
'I

metres/second

(9.11)
I

1/

x3a

metres/second 2

(9.12)

x~ = X~a

radians/second

(9.13)

x~ = X~a

radians/second 2

(9.14)

x3

9.3 HYDRODYNAMIC COEFFICIENTS


9.3.1 General considerations
Consider a strip embedded in an infinitely long cylinder lying in calm water as shown
in Fig. 9 .1. The excitations required to sustain the motions in the three possible
degrees of freedom may be obtained from equations analogous to the sway, heave
and roll equations for the complete_ ship (equations (-8.24)-(8.26)). We are here
concerned only with that part of the excitation necessary to oppose the hydrodynamic reactions to the motions. In other words the concentrations of the real mass and
inertia of the strip are excluded from the calculation. ~or a strip with port/starboard
symmetry we have

Strip theory

174

[Ch. 9
(9.15)
(9.16)
(9.17)

where all the motions and excitations are referred to an origin 0 in the waterplane.
The coefficients a22 , b22 , etc. are all local values (per metre length of strip) analogous
to the corresponding coefficients for the complete hull (with due allowance for the
change of origin). These local coefficients are functions of the size and shape of the
strip in question and may be determined by the methods described in Chapter 11.
9.3.2 Coefficients in the heave and pitch equations
Consider a ship undergoing simultaneous forced heave and pitch motions in calm
water. The momentum of the surrounding water in the plane of the strip is
M~

a~ 3 x~

tonne metres/second per metre length of strip

where the vertical velocity is given by equation (9 .11). The force required to oppose
the hydrodynamic reactions and sustain the motions of the strip is composed of the
rate of change of this momentum together with contributions from the damping and
stiffness:

As successive strips pass the observer stationed alongside the ExEl axis he will
perceive a changing local added mass a~ 3 So the rate of change of momentum is

gt (M~) = a~3 x~ + a3J,x~

kN/metre length of strip

Now we may write for the rate of change of added mass

'I

a33

= dxd

a33 XB1

Bl

= - Udxd
Bl

(a~ 3 )

tonne metres/second per metre length of strip

(see equation (9.4)). Using equations (9.11) and (9.12) the downward vertical force
on the strip becomes

Hydrodynamic coefficients

Sec. 9.3]

(b '

33-

175

uda~3)(
.,
uXs )
dxst x3-Xs1Xs+
(9.18)

The total heave force and pitch moment required to balance the hydrodynamic
reactions and sustain the heave and pitch motions of the ship are obtained by
allowing ox81 to approach zero and integrating over the length of the hull:

(9.19)

= -

F5

Jx

81

dF3

(9.20)

kN metres

Comparing these equations with equations (8.25) and (8.27) enables us to obtain
expressions for each coefficient in the equations of motion for pitch and heave. For
conventional ships with zero cross-section area at the forward perpendicular the local
added mass is always zero at the stem. It follows that
I

da~3

dx

dx81

,
a 33 a

II

tonnes/metre

81

where a~ 3 a is the local added mass for v.ertical motions at the stern in tonnes/metre
and x 81 a is the distance from the centre of gravity to the stern (negative).
The coefficients then become

(9.21a)

176

[Ch. 9

Strip theory

b33

Jc~3dxB1

C33

a3s

Jb~3 dxBl - Ua~3a

kN/(metre/second)

kN/metre

-JXB1 a~3 dxBl

(9.21c)

kN/(radian/second 2)

(9.21d)

Ja~3 dxBl- JXB1 b~3 dxBl + UxB1aa33a

b3s

(9.21b)

2U

kN/(radian/second)
(9.21e)

Jb33 dxBl - U2a33a - JXB1 c33 dxBl

C3s

lj

as3

- JxB1a33dxB1

bss

kN metres/(metre/second2)

(9.21f)

(9.22a)

JxB c~ 3 dxB 1

kN metres/metre

(9.22c)

Jx~ 1 a33 dxB 1

kN metres/(radian/second2)

(9.22d)

ass

kN/radian

- 2U XB1 a33 dxBl +


- Ux~ 1 aa3 3

JX~1 b33 dxBl

kN metres/(radian/second)

(9.22e)

kN metres/radian
(9.22f)

177

Hydrodynamic coefficients

Sec. 9.3]

9.3.2 Coefficients in the lateral plane equations


A similar approach is used to determine the coefficients in the lateral plane equations
of motion ((8.24), (8.26) and (8.28)). We now consider a ship undergoing simultaneous sway, roll and yaw motions in calm water and the lateral motions of a strip are
given by equations (9.9), (9.10), (9.13) and (9.14).
The horizontal momentum of the surrounding water in the plane of the strip is
M~

(a~ 2 x~

+ a~ 4 x~) tonne metres/second per metre length of strip

where the second term is the contribution of the roll motion.


The lateral force required to balance the hydrodynamic reactions and sustain the
motions of the strip includes contributions from the rate of change of momentum, the
lateral motion damping and the lateral force required to sustain the roll velocity:

The rate of change of momentum perceived by our stationary observer is

and the lateral force is


I

II

(9.23)
The angular momentum of the water in the plane of the strip includes a contribution
due to the lateral velocity x~:

M~ = a~ 4 x~ + a~2 x~

tonne_ n:etre 2/second per metre length of strip

The roll moment about the axis through G required to balance the hydrodynamic
reactions and sustain the motions includes contributions from the rate of change of
this angular momentum as well as contributions assoc-iated with the roll damping and
the moment required to sustain the lateral velocity. In addition there are contribu-

Strip theory

178

[Ch. 9

tions associated with the roll stiffness c~ and the moment due to the lateral force oF2
acting through 0:

Expanding as before, we obtain

oF4

= [ a~4 x4 + ( b~-

u::)

x4 +

c~4x4

+ a~z(Xz + Xau6- 2Ux6- OG x4)


+ ( b~2- u:::) (iz + XalX6- Ux6- OG x4) JOXal
- OG oF2 kN metres
The total sway force, roll and yaw moments are obtained by allowing
approach zero and integrating over the length of the ship:

(9.24)

ox

81

to

Fz

fdF2 kN

(9.25)

F4

(9.26)

F6

dF4 kN metres

Xa1 dF2

kN metres

(9.27)

Comparing these equations with equations (8.24), (8.26) and (8.28) allows us to
obtain expressions for the lateral plane hydrodynamic coefficients. Again we note
that for conventional ships with zero cross-section area at the forward perpendicular

daiz

dx

dx81

,
a22 a tonnes/metre

(9.28a)

81

(9.28b)

179

Hydrodynamic coefficients

Sec. 9.3]

(9.28c)
where a~2 a is the local added mass for horizontal motions in tonnes/metre at the
stern.
Similar expressions are valid for the coefficients a~4 and a~2 Equating terms in
equations (9.25)-(9.27) and (8.24), (8.26) and (8.28) we obtain

azz

Ja~2

hzz

Jb~2

dxB 1

dxB 1 -

(9.29a)

tonnes

Ua~2 a

(9.29b)

kN/(metre/second)

(9.29c)

Ua~ 4a

(9.29d)

kN/(radian/second)
I

az 6

JxB 1 a~2 dxBl

b26

XBr

b~2

1/

kN/(radian/second 2 )

dxB 1 -

Ua~2a- 2UJa~2

dxB 1

(9.29e)

kN/(radian/second) (9.29f)

(9.29g)

Ja~z

dxB 1 -

OG

a~2

dxBl.

kN metres/(metre/second2 )

+ OG Ua~ 2 a kN metres/( metre/second)

(9.30a)

(9.30b)

[Ch. 9

Strip theory

180

a44 =

Ia~4

+ OG2

b44 =

Ia~z

OG

dxBl -

Ja~2

Ib~4

dxB 1

dxBl-

Ib~4

- OG

dxBl -

a46

Ic~4
I a~z

dxBl

XB1

dxBl

kN metres/(radian/second2 )

Ua~4a + OG Ua~Za- OG b~z

dxBl

Ib~z

Ua~2a

dxBl

+ OG2

+ OG Ua~4a
c44

Ia~4

OG

dxB 1 -

OG

OG

(9.30d)

kN metres/( radian/second)

kN metres/radian

dxBl-

(9.30c)

xB 1 a2z dxB 1

(9 .30e)

kN metres/(radian/second2 )
(9.30f)

I
Ia~2
U a~2a- uIb~z
- OG

+ 20G U

c46

= 2

XB1 b2z dxBl

dxB 1

+ OG XBla Ua~za

kN metres/(radian/second)

dxBl

Ib~z

+ OG u

- OG U2a~2a kN metres/radian

(9.30g)

dxBl

(9.30h)

(9.31a)

(9.31b)

Excitations in regular waves

Sec. 9.4]

181

kN metres/(radianlsecond2 )
(9.31c)

b64

XB1 b24 dxBl - OG XB1 b2z dxBl + OG UxBlaazza

- UxB 1aa24a kN metres/(radianlsecond)


a66 =

b66

Jx~ 1

(9.31d)

a2 2 dxB 1 kN metres/(radianlsecond2 )

(9.31e)

JX~1bzzdxB1- UxBlazza

- 2U xBla22 a dxB 1 kN metres/( radian/second)

(9.31f)

(9.31g)

9.4 EXCITATIONS IN REGULAR WAVES


I

9.4.1 General considerations

!I

The linearisation of the equations of motion in Chapter 8 allows the wave excitations
to be considered independent of any ship motions and to be expressed as functions of
the wave amplitude alone. In other words the wave excitations are assumed to be the
same as the ship would experience if it were rigidly restrained and allowed no
motions at all.
It was shown in Chapter 8 (equation (8.1)) that the wave depression at any point
(x 1 , x2 ) related to the moving origin 0 could be written as
~

= ~ 0 sin(roet- kx 1 cos p. + kx 2 sin p.)

metres

Since the ship is allowed no motions, the centre of gravity remains above 0 and
we may write
~

x1

= xB 1

metres

The wave depression varies across each strip but we ~ssume the ship to be slender
(that is, the waterline beam of all strips is much less-lhan the wavelength) and this
allows us to calculate the wave depression with sufficient accuracy by setting

182

Strip theory

x2

[Ch. 9

= 0 metres

The wave depression experienced at each strip is then


1;, = 1;,0

sin(roet- Q)

metres

(9.32)

where Q = kxB 1 cos p,.


The excitation experienced by each strip is related to the pressures, velocities and
accelerations in the water beneath the wave surface. These quantities vary with
depth (see Table 3.1) and it is usual to simplify the calculation by taking their values
at a mean local draught defined by

= -A
B

metres

(9.33)

where A is the cross-sectional area of the strip in metres 2 and B is the waterline beam
of the strip in metres.
Referring to equation (3.36) we find that the pressure fluctuation at the mean
draught Dis

P=

P8so exp (- kD) sin (kx- rot) kN/metre 2


above the local hydrostatic pressure pgD

and, following a similar procedure to that described above, this can be written as

P= -

P8so exp ( -kD) sin(roet- Q)

- pg exp(- kD)I;,

kN/metre 2

(9.34)

Similarly the vertical velocity of the water at the mean draught is, from equations
(3.23) and (3.26),

v = ro/;,0 exp(- kD) cos(roet- Q) metres/second

(9.35)

and the corresponding vertical acceleration is

v=

- ro2 exp(- kD)

1;,

metres/second 2

(9.36)

The horizontal velocity at the mean draught is, from equations (3.22) and (3.26),

Sec. 9.4]

Excitations in regular waves

-co exp(- kD)

metres/second

183

along the Ex axis

The athwartships component of this horizontal velocity is


u2

co exp(- kD)

s sin fL

metres/second

to starboard

(9.37)

and the corresponding acceleration is

cov sin fL

metres/second 2

(9.38)

The slope of the pressure contour at depth Dis, from equation (3.16),

ks

cx15 = -

exp(- kD) cos(coet- Q)

cov
g

radians

and this has an athwartships component given by

CXDz

. fL
= -cov sm
g

ra d"1ans

(9.39)

The rate of change of the athwartships component of this slope (equivalent to the
average angular velocity of the water) is
' 1/
ixm

= -

kco exp(- kD)

ssin fL

radians/second

(9.40)

and the corresponding angular acceleration is


fxm

= -

kcov sin fL

radians/second 2

(9.41)

9.4.2 Vertical excitation


If we assume that the beam of the strip at the mean draught D is approximately the
same as the waterline beam, the vertical force due to the pressure fluctuation is

In addition there are contributions arising from the rate of change of the vertical
momentum of the water surrounding the strip and a force associated with the vertical
velocity of the water. The total vertical force on the strip is then

Strip theory

184

[Ch. 9

where the vertical momentum is


tonne metres/second per metre length of strip
Our stationary observer perceives the rate of change of momentum as

= -

da~ 3 v + a3' 3v kN/ metre Iength o f stnp

U dx
Bl

Obtaining P, v and
each strip is

8Fw 3

vfrom equations (9.34)-(9.36) we find that the vertical force on

1; 0 exp(- kD)[P 1 sin(roet- Q) + P2 cos(roet- Q)] 8x 81 kN


(9.42)

where

P2 = ro ( b~ 3 - U

da' )
dx::

tonnes/(metre/second2 )

The total heave force and pitch moment are obtained by allowing
approach zero and integrating over the length of the hull:

ox

81

to

The excitation amplitudes are

(9.43)

Sec. 9.4]

Excitations in regular waves

185

and the phases are given by

tan y 3

__ JRzdxs1

(9.45)

__ JxalRzd.xal
tan y 5

(9.46)

where

9.4.3

R1

R2

= (- P 1 sin Q + P 2 cos Q) exp(- kD)'

(P1 cos Q + P 2 sin Q) exp (- kD)

Sway and yaw excitation

1/

Fig. 9.3 shows the hydrostatic force experienced by a restrained strip in waves. The
inclined water surface causes a lateral shift of the centre of buoyancy from B to B 1
and the buoyancy force vector is assumed to act normal to the pressure contour at the
mean draught D. The buoyancy force is
pgA Ox81

kN

and it has a horizontal component


pgA Ox 81 O:vz

kN

to starboard

In addition there are contributions dueto the rate of change of horizontal momentum of the water surrounding the strip and forces associated with the lateral and
rotational velocities of the water. The total horizontal force on the strip is then

186

[Ch.9

Strip theory
M'

T
Pressure
contour at
depth D

Fig. 9.3- Lateral plane hydrostatic force and moment on a restrained strip.

The horizontal momentum is

M'w 2

a~ 2 u 2

+ a~4 &D2

tonne metres/second per metre length,of strip

The stationary observer perceives the rate of change of momentum as

D ( I )
Dt Mwz

da~z
= azzUz- U dx
Uz + az4CXvz
I

,_

81

da~ 4 cxDZ
._
U dx

kN per metre 1engt h o f stnp

Bl

Obtaining u 2 , u2 , etc., from equations (9 .37)-(9 .41) we find that the total
horizontal force on the restrained strip is

Sec. 9.4]

Excitations in regular waves

oFw2 = ~ 0 exp(- kD) sin p, [P3 sin(roet- Q)


+ P4cos(roet- Q)] oxB 1 kN

187

(9.47)

where

The total sway force and yaw moment are obtained by allowing oxB 1 to approach
zero and integrating over the length of the hull:

The excitation amplitudes are

(9.48)

and the phases are given by

(9.50)

(9.51)

188

Strip theory

[Ch. 9

where

(P3 cos Q + P4 sin Q) exp (- kD)

R3

R4

= ( -P3 sin Q+P4 cos Q) exp( -kD)

9.4.4 Roll excitation


If the strip shown in Fig. 9.3 is wall-sided the two triangular wedges ORS and OTU
will have equal volumes and identical shapes. Their centres of buoyancy will be at b 1
and b2 and the strip's centre of buoyancy will move from B to B 1 such that

the line BB 1 is parallel to the line b 1b 2 and the distance BB 1


where oA is the area of each wedge in metres 2
For small values of the pressure contour slope cxv 2 the distance

The area of each wedge is

B2

B2

oA = - ex sin p, = - ex

metres 2

and the centre of buoyancy of each wedge is

B
3

metres

from the centreline.

Hence

and
metres
The buoyancy force exerts a roll moment about 0:

CiA=A
b 1 b2

Excitations in regular waves

Sec. 9.4]

pg (

B3
_
)
- OBA
12

<X02

189

.
kN metres per metre length of strip

The total roll moment exerted on the strip about 0 also includes contributions from
the rate of change of the angular momentum of the water surrounding the strip as
well as contributions associated with the rate of rotation and the horizontal velocity
of the water. The equations of motion (8.23)-(8.28) require the roll moment to be
related to an axis through the centre of gravity. The lateral force Fw 2 acting through 0
exerts a moment about G and this is included to give

DFw4 =

[g/M_',4)+b~4lXoz+pg(~; -AOB )!Xoz + b~zUz] OXBl


- OG oFw 2 kN metres

(9.52)

The rate of change of momentum perceived by the fixed observer is

..

a44!XDz-

uda~4,

IXDz + a4zUz-

dx

Bl

uda~2

dx

Uz

Bl

kN metFes per metre length of strip


and the roll moment becomes

kN metres

(9.53)

where

The total roll excitation on the strip is obtained by allowing oxB 1 to approach zero
and integrating over the length of the hull. We obtain

190

Strip theory

[Ch.9

The excitation amplitude is

and the phase is given by

(9.55)

where
R 5 = (P5 cos Q
R6

P6 sin Q) exp(- kD)

= (- P5 sin Q + P6 cos Q) exp(- kD)

10
Hydrostatic coefficients

10.1

INTRODUCTION

The hydrostatic or stiffness coefficients cij in the strip theory formulae (equations
(9.21), (9.22) and (9.30) are associated with changes in buoyancy forces due to steady
displacements of each strip. This chapter outlines the calculation of these buoyancy
fores to allow determination of the stiffness coefficients.

10.2 VERTICAL PLANE


If the strip is wall-sided a steady downward heave
an increase in displaced volume

displace~At x; metres results in

giving an increase in buoyancy force

pgBx3 oxB1 kN
So the downward heave force required to sustain the steady downward heave
displacement is

and the local heave stiffness coefficient is


c;3 = pgB

kN/metre length of strip

(10.1)

192

Hydrostatic coefficients

[Ch. 10

Now the waterplane area of the complete hull is


(10.2)
The first moment of area of the waterplane about a transverse axis under the centre
of gravity is

(10.3)
The second moment of area of the waterplane about a transverse axis under the
centre of gravity is

(10.4)
We may now simplify the expressions for the vertical plane stiffness coefficients given
for the complete hull in equations (9.21) and (9.22);
c33 = pgAw kN/metre

(10.5a)

(10.5b)
(10.5c)

(10.5d)

10.3 LATERAL PLANE


Although equation (9 .30) gives the roll stiffness c44 in terms of the local roll stiffness
c~ 4 , it is more convenient to calculate c44 by considering the righting moment on the
complete ship as shown in Fig. 10.1. The analysis has much in common with that
already used in calculating the roll excitation in Chapter 9.
The ship is rolled to a small angle x4 radians and we require to find the roll
moment needed to sustain this roll angle. If the ship is wall-sided in the vicinity of the
waterline the two wedges ORS and OTU will have equal .volumes and identical
shapes. Their centres of buoyancy will be at b 1 and b 2 and the ship's centre of
buoyancy will move from B to a new position B 1 such that

Sec. 10.3]

193

Lateral plane

Fig. 10.1- Roll stiffness of complete ship.

the line BB 1 is parallel to the line b 1 b2 and


.
Vthe distance BB 1 = Vb 1 b2 metres
I

!I

where Vis the volume of each wedge in metres 3 and Vis the volume of displacement
of the hull in metres 3
1
The buoyancy force vector now acts through the new centre of buoyancy B 1 and
interesects the plane of symmetry of the ship at the metacentre M. For small angles of
roll the distance

Now the area of a section through each wedge is


B x4

metres

and the centroids of the sections are

194

Hydrostatic coefficients

2B
3

[Ch. 10

metres apart

It follows that

where the second moment of area of the waterplane about the longitudinal axis is

(10.6)

We now have
BM =IT

metres

(10.7)

The buoyancy force


mg

kN

has a lever arm

about the centre of gravity, giving a righting moment

and the roll stiffness is


c44 = mg GMs

kN metres/radian

(10.8)

GM 5 is called the 'solid' metacentric height and is an important measure of the ship's
stability. If GM5 is negative the ship will be unstable, and liable to capsize.
This calculation assumes that there are no free surfaces within the ship. Fig. 10.2
shows the effect of a free surface such as might be found, for example, in a partially
filled fuel or ballast tank. As the ship rolls, the fluid in the tank collects in its low side
and causes a lateral shift in the position of the centre of gravity from G to G 1 . The
righting lever GZ is thereby reduced to G 1Z and this may be col)veniently expressed

Sec. 10.3]

195

Lateral plane

Fig. 10.2- Reduction in GZ due to internal free surfaces.

as an effective reduction in metacentric height from GM5 to GMF. Internal free


surfaces may be responsible for apparent reductions in metacentric height of25% or
more.

!I

11
Local hydrodynamic properties

11.1

INTRODUCTION

Progress with strip theory requires the evaluation of the local added mass and
damping coefficients ajj and bjj in equations (9.21), (9.22), (9.30) and (9.31).
Methods of solution for ship-like forms have been developed by Ursell (1949a,b),
Tasai (1959, 1960), Grim (1959), Porter (1960) and others. The techniques used are
involved and laborious, requiring devious and intricate methods of solution for even
the simplest of cases. Their application is quite impracticable without the aid of a
digital computer.
A full explanation is inappropriate for this book. The presentation which follows
is necessarily superficial but should serve to give a taste of the methods used. The
reader who requires a more detailed knowledge is referred to an excellent paper by
de Jong (1973).
11.2

HYDRODYNAMIC PROPERTIES OF A HEAVING CIRCULAR


CYLINDER

U rsell made the first step towards solving the general problem of calculating the twodimensional potential flow around a cylinder of arbitrary shape floating in a free
surface in 1949. He derived a solution for a heaving circular cylinder as shown in Fig.
11.1.
The water depth was assumed to be infinite and the heaving motion of the
cylinder was taken to be
y =Yo cos( rot+ e)

metres

where

y 0 is the heave amplitude in metres

(11.1)

Sec. 11.2]

Hydrodynamic properties of a heaving circular cylinder

197

y=yo cos(OJt+E)

X
~

Waves radiate
away from the
cylinder

Fig. 11.1- Circular cylinder oscillating in the free surface.


I

1/

ro is the oscillation frequency in radians/second


t is the time in seconds
E is a phase angle in radians.
After initial transients have died away, the oscillating cylinder generates a train of
regular waves which radiate away to infinity on either side of the cylinder, serving as a
mechanism for the dissipation of energy. Ursell showed that the stream and potential
functions describing the flow around the cylinder may be written as

\jl

=!:[ ( + ~1
'l'c

Pzm 'Vzm} cos( rot)

+ ( 'l's + ~ q2m).jlzm)

sin( rot)]

metresYsecond

(11.2)

198

Local hydrodynamic properties

<P=gso[(<Pc+
7TCO

m= 1

+ ( <Ps +

Pzm<Pzm)

[Ch. 11

cos(cot)

1; qzm <Pzm)

sin(cot)]

metres2 /second

(11.3)

where so is the amplitude of the radiated wave at an infinite distance from the
cylinder. The terms
gso['l'c cos(cot)+ 'Vs sin(cot)]
7TCO

gso[ <Pc cos(cot) +


7TCO

<Ps sin(cot)]

represent a pulsating source at the origin. The individual components are


'l'c =

1T

exp(- ky) sin(klxl)

(11.4a)

<Pc =

1T

exp(- ky) cos(kx)

(11.4b)

'l's =

-7r

J exp~-+ ~~l)[v sin(vy) + k cos(vy)] dv


00

exp(- ky) cos(kx) +

(11.4c)

<Ps = 1T exp(- ky) sin(klxl)-

J exp~-+ ~~xl)[v cos(vy)- k sin(vy)] dv


oo

(11.4d)

In these expressions k is the wave number and the two series

represent an infinite number of pulsating multipoles aligned with the y axis (see
equations (2.39) and (2.40)). The individual components are

Hydrodynamic properties of a heaving circular cylinder

Sec. 11.2]

_ zm(sin(2m9)
o/zm- a

..h

,zm

k sin[(2m -1)9])
+ (2m- l)r 2m-1

(11.5a)

k cos[(2m -1)9])
+ (2m -l)r 2m-1

(11.5b)_

zm(cos(2m9)

'!'2m- a

199

,zm

where a is the radius of the cylinder in metres.


The coeffieients Pzm and q2 m weight the contributions of the individual multipoles
and their values are chosen so that the stream and potential functions give the correct
velocities on the surface of the oscillating cylinder. From equation (2.26b) the
velocity normal to the cylinder surface must be

Ur

-1 0\j/
= --;;
metres/second
09

Ur

Now

d: cos 9 = -

Yoffi

sin (rot+ e) cos 9 metres/second

from which we obtain

~: =ayoro sin(rot+e) cos 9


so that

\jl=ay0ro sin(rot+e) sin 9+c(t)

(11.6)

Since the stream function \jl is defined as zero when 9 = 0, the constant of integration
c(t) is zero.
After some manipulation of equations (11.2) and (11.6) Ursell showed that

ay0 ~:sin( rot+ e)= A. cos(rot)+ B. sin(rot)


...

(11.7)

where

~ ~( 1)m-1 2kap2m
- ( )
A o/c 8=7T/2 + .L..J -1
m=l

(11.8a)

Local hydrodynamic properties

200

-('l's) 8=7T/2 + L.J


~
B -

m=l

(-

kaqzm
1)m-l 2"=1:
m

[Ch. 11

(11.8b)

and

'l'c- ('!'c)e=7T/2 sin 9 =

'l's- ('!'s)e=7T/2 sin e =

Pzmfzm

(11.9a)

L:

qzmfzm

(11.9b)

m=l

m=l

where

fzm =

- {sin(2m9) +

2:~ 1 sin[(2m -1)9]- sine sin( (2m -1) ~)}

(11.9c)

and 'l'c and 'l's are evaluated on the surface of the cylinder (r =a).
Ursell obtained an approximate solution for Pzm and q2m by limiting the infinite
series on the right-hand side of equations (11.8) and (11.9) to only six terms. The
equations were then formulated at ten values of e from oo to 90 at intervals of 10.
This gave ten simultaneous equations for p and another ten equations for q. Each
equation had only six unknowns (p 2 , p 4 , p 6 , p 12 and q2 , q4 , q6 , q 12) and the
best approximate solutions were found by a least squares method.
This approach ensured that the stream and potential functions behaved correctly
at the surface ofthe cylinder. We may examine the flow at an infinite distance from
the cylinder by allowing r and x to approach infinity in equation (11.3). The
contributions of the multipoles now become negligible and equation (11.3) reduces
to

<1>

= gl;o exp(- ky) cos(kx- rot)

metres 2/second

(J)

Comparing this with equation (3.1) we see that the potential represents a regular
wave of amplitude 1; 0 metres and frequency ro radians/second in water of infinite
depth. So the potential function has the expected behaviour on the free surface at an
infinite distance from the cylinder.
The added mass and damping coefficients of the oscillating cylinder are obtained
by considering the pressure fluctuations on its surface. From Bernoulli's equation
(2.18) the pressure is

Sec. 11.2]

Hydrodynamic properties of a heaving circular cylinder

P = p(n- a<j>- q

at

201

kN/metre 2

If the velocity q is assumed to be small this reduces to

P = p ( gy -

~~)

kN/metre2

since the force potential gradient due to gravity is

an

ay = g metres/second2

So the variation about the steady hydrostatic pressure pgy is


a<J>
P= - ~

at

kN/metre 2

Consider now a small element of length os on the surface of the cylinder (Fig.
11.2). The force due to the fluctuating pressure acting on the element is

!I

Fig. 11.2- Pressure acting on surface of cylinder.

P os kN per metre length of .-the cylinder


Now
OS = a oe

metres

Local hydrodynamic properties

202

[Ch. 11

and the total vertical force applied to the cylinder by the fluid is
TT/2

P ds cos 9 kN per metre length of the cylinder

-rr/2

So the total force required to oppose these pressures and sustain the oscillation of the
cylinder is

F=

rrt2

-rr/2

P cos e ds = pa

f"'2
-rr/2

a<j>
at
cos e de

= - 2p/~o8 [M0 cos(rot)- N 0 sin(rot)] kN/metre


1T

(11.10)

where

Mo=

No=

"
f
f

6
12
1
( -1)m- q2m
1Tkaq2
o (<J>s)r=acos9d9+ ];1
4m2-1
+-4-

(11.11a)

rr/2
6
( -1)m-1 P2m
1Tkap2
o (<J>s)r=acos9d9+ ];1
4m2-1
+-4-

(11.11b)

The added mass may be defined as


component of force in phase with the acceleration
acceleration amplitude
and the damping coefficient is
component of force in phase with the velocity
velocity amplitude
Now the velocity of the cylinder is, from equations (11.1) and (11.7),

y=

- y0 ro sin(rot+ e)= - g~o [A. cos( rot)+ B. sin( rot)] metres/second


1raro
(11.12)

and the acceleration is

Hydrodynamic properties of a heaving circular cylinder

Sec. 11.2]

ji = gt;o [A. sin(rot)- B. cos(rot)]

1Ta

metres 2/second

203

(11.13)

For the time being we may express the force on the cylinder (equation (11.10)) as

F = Fe cos(rot) + F. sin( rot) kN/metre

(11.14)

where

F = 2 pagt;oMo kN/metre
c
1T
F
s

= - 2 pagt;oNo
1T

(11.15a)

kN/metre

(11.15b)

and the velocity and acceleration (equations (11.12) and (11.13)) are

y = Vc cos(rot) + v s sin(rot)

metres/second

(11.16)

ji = ac cos( rot)+ a. sin( rot)

metres/second 2

(11.17)

where

- gt;oA.
1taro

v
s

= - gt;oB
1taro

a = - gt;oB
c

1Ta

= - gt;oA.
1Ta

metres/second

!I

(11.18a)

metres/second

(11.18b)

metres/second 2

(11.18c)

metres/second 2

(11.18d)

These equations are illustrated in a vector diagram in Fig. 11.3. By using similar
triangles it can be shown that the amplitude of the component of the force in phase
with the acceleration is

(11.19)

Local hydrodynamic properties

204

[Ch. 11

sin(wt)

Fig. 11.3 -Vector diagram showing exciting force, acceleration and velocity for a heaving
cylinder.

so that the force in phase with the acceleration is

2pag~0 (A.No
+ B.Mo) [A. sm(rot)-B.
.
A~+B~
cos(rot)] kN/metre(11.20)

Fa=-1T-

The amplitude of the component of the force in phase with the velocity is
(11.21)
so that the force in phase with the velocity is

Fv=

- 2pag~ 0 (A.M0 + B.N0 )


1T

A~+B~

[A. cos(rot)+B. sm(rot)] kN/metre


(11.22)

Then the added mass is

a' =Fa= 2pa 2


33
ji

(A.No + B.Mo)
A~+ B~

and the damping coefficient is

tonnes/metre length of cylinder (11.23)

Sec. 11.2]

Hydrodynamic properties of a heaving circular cylinder

kN/(metre/second) per metre length of cylinder

205

(11.24)

Equation (11.24) may be simplified by considering the radiation of energy away


from the oscillating cylinder. The energy contained in one wavelength at infinity is,
from equation (3.40),

- pg~6A. - 1Tpg ~6
E - -- - ----o;z joules/metre length of wave crest
2
2

(11.25)

This energy is radiated away from the cylinder at the group velocity u0 given by
equation (3.44). In deep water
c
u0 = Z metres/second

and the energy passing a fixed point in one motion (or wave) period is E/2
joules/metre length of the wave crest. But since waves are generated on both sides of
the cylinder the total energy radiated in one motion cycle is E joules/metre and this is
equal to the work done in oscillating the cylinder through one motion cycle:
2~2

1T

E = pg2
(0

J2rr/ro Fy dt

joules/metre

II

(11.26)

Substituting equations (11.10) and (11.12) this reduces to


(11.27)
and equation (11.24) then becomes
(11.28)
Finally we may obtain the wave amplitude at infinity by considering the value of
the stream function on the surface of the cylinder (r =a) ate= 1rl2. From equation
(11.2) we find that

\jl=g~o
[A. cos(rot)+B. sin(rot)] metres~/second
1T(O

(11.29)

Local hydrodynamic properties

206

[Ch. 11

and from equation (11.6)


'l' = ay 0 ro sin( rot+ c)

metres 2/second

(11.30)

After some manipulation of these two equations the ratio of the wave amplitude at
infinity to the cylinder motion amplitude is found to be
~0
Yo=

11.3

rrka

(11.31)

V(A: + B:)

LEWIS FORMS

Ursell's solution for the flow around a heaving circular cylinder was a useful step
towards creating a viable theoretical method of predicting ship motions in waves.
However, ship hull cross-sections are not usually circular and we require a technique
for predicting the flow and resulting hydrodynamic coefficients for cylinders of
arbitrary, or at least ship-like, cross-section.
The techniques of conformal transformatiQn described in Chapter 2 provide a
method of transforming a circle in the z plane

z=x+iy=re-ie metres

(11.32)

into an arbitrary shape in the ~ plane


~ = XBz

+ ixB 3 metres

(11.33)

as shown in Fig. 2.16. In principle the transformation

metres

(11.34)

will map any point on the circle in the z plane into a corresponding point on any
symmetrical (about the xB 3 axis) shape in the ~ plane provided that appropriate
values of the coefficients a0 , aI> a3 , a5 etc. are chosen. In practice it is usual to set the
radius of the cylinder at

a=r=l.O metre
and to truncate the transformation series to only three terms:

Lewis forms

Sec. 11.3]

207

(11.35)

We shall see that this still allows a wide variety of ship-like cross-sections to be
generated from the unit circle. These forms will not in general be exact replicas of any
given hull cross-section, but the match will usually be sufficiently close to allow
adequate estimates of the hydrodynamic coefficients for ship motion calculation.
The resulting family of forms are known as Lewis forms, after F. M. Lewis who first
proposed their use in 1929.
Substituting equations (11.32) and (11.33) into equation (11.35) and separating
real and imaginary parts we obtain a pair of parametric equations in describing the
shape of the Lewis form in the I; plane:

XBz

= a0 [(1 + a 1)

xB 3

sin

a0 [(1- a 1) cos

e- a 3 sin(39)]

metres

(11.36a)

e+ a 3 cos(39)]

metres

(11.36b)

= 0 corresponds to the bottom of the unit circle and the keel of the Lewis form.
Substituting = 0 in equations (11.36) we obtain

XBz

=0

(11.37a)

metres

(11.37b)
,I I.1

where D is the Lewis form draught in metres.


e= rr/2 corresponds to the intersections of the unit circle and the Lewis form with
the waterplane. Substituting this value in equations (11.36), we obtain

(11.38a)

xB 3

=0

metres

(11.38b)

where B is the waterline beam of the Lewis form in metres. The beam/draught ratio
of the Lewis form is

H = B = 2(1 + a 1 + a3 )
D
1-a1 +a~
The cross-sectional area of the Lewis form is

(11.39)

Local hydrodynamic properties

208

[Ch. 11

B/2

=2

metres 2

x 83 dx 82

(11.40)

Substituting equations (11.36) we obtain

A= 1T;0(1- a[- 3a5) metres2


2

(11.41)

and the section area coefficient is

(11.42)

Explicit equations for the coefficients a1 and a 3 may be obtained by rearranging


equations (11.39) and (11.42) to give

(11.43a)

a3=

3- C+ y(9- 2C)

(11.43b)

where

C= 3+4cr+(l- 4cr) (H-2)


H+2
1T

1T

a0 is simply a scale factor governing the overall size of the Lewis form.
Lewis forms may therefore be defined in terms of their beam/draught ratio Hand
their section area coefficient
Fig.11.4 shows a range of Lewis forms for various
values of Hand t Evidently a wide variety of conventional hull cross-sections may
be represented reasonably well by choosing Lewis forms having the same beam/
draught ratio and section area coefficient. There is no limit to the permissible beam/
draught ratio but only a limited range of section area coefficients are possible.
Clearly the formula for a 3 (equation (11.43b)) becomes invalid when

cr.

cr.

C>i
t

Only the right-hand sides of the symmetrical forms are shown.

Sec. 11.3]

Lewis forms

209

0
H=1

Fig. 11.4- Typical Lewis forms.

and this is prevented if

cr<

6:H (H +20H +4)


2

(11.44)

Lewis forms having section area coefficients greater than this cannot exist. In
practice section area coefficients approaching this limit ltave rather angular shapes of
the type shown in Fig. 11.5(a). They are not repr~~entative of conventional hull

210

Local hydrodynamic properties


0

[Ch. 1l

Xsz

~-----~

H=2
o=1.15

(a) Area coefficient too large

(b) Area coefficient too small

Fig. 11.5- Examples of invalid Lewis forms.

forms: more to the point is the fact that such section shapes would probably
experience flow separation around the sharp bilges, and the potential ,flow techniques to be employed for predicting the hydrodynamic coefficients would not be
expected to give very reliable results. To avoid forms of this nature it is usual to
suggest that the Lewis forms should lie completely within the circumscribing
rectangle so that

If the section area coefficient is too small the Lewis form will adopt physically

impossible shapes with negative values ofxs2 andxs 3 as shown in Fig. 11.5(b). We
therefore also insist that

Hydrodynamic properties of Lewis forms

Sec. 11.4]

211

Applying these limits to equations (11.36) with

o:::::;e:::::;-1T
2

allows permissible ranges for the section area coefficient cr to be determined:

37T
37T
-(4-H) :::::; cr:::::;- (24+H)
64

256

for H:::::;2

These limits are shown, together with the ideal upper limit given by the inequality
(11.44), in Fig. 11.6.
The special case

H=2

1T

cr='
4

yields, from equations (11.43),


I

!I

The Lewis form is then given by equations (11.36) as


Xsz

= ao sin 9,

Xs3

= ao COS 9

metres

and these are the equations of a circle of radius a0 metres in the

11.4

splane.

HYDRODYNAMIC PROPERTIES OF LEWIS FORMS

11.4.1 Introduction
The determination of the hydrodynamicyroperties of Lewis forms oscillating in the
free surface proceeds along lines broadly similar to those pioneered by Ursell for the
heaving circular cylinder. Heave, sway and roll motions are considered in turn: in
each case the stream and potential functions are again given by equations (11.2) and
(11.3) although the components 'l'c, <l>c, 'l's and <l>s and the infinite series of multi poles
are defined differently. In principle it is again necess_'!ry to determine the weighting
coefficients Pzm and q2 m for the infinite number of multipoles. In practice an

212

Local hydrodynamic properties

[Ch. 11

No Lewis forms

1.2

1.0

.....
c

Q)

c::;
~
Q)

0.8

Valid Lewis forms

(.)

co
~

co

0.6

."B
Q)

(/)

0.4

0.2

0
Beam/draught ratio H

Fig. 11.6- Permissible range of Lewis forms.

approximate solution is obtained for a limited number of multipoles by satisfying the


required boundary conditions for the fluid velocity on the surface of the Lewis form
cylinder. Finally the fluctuating pressures are again integrated to find the fluid forces
imposed on the Lewis form, and the hydrodynamic coefficients are determined by
methods similar to those used for the circular cylinder.
The interested reader will find a complete derivation of the appropriate formulae
in de Jong's (1973) paper. For present purposes it suffices to state his results without
proof.

11.4.2 Heaving Lewis form


The added mass and damping coefficients are

(11.45)

pBzro1Tz

b~3 = (A~+ B~)

kN/(metre/second) per metre

(11.46)

Hydrodynamic properties of Lewis forms

Sec. 11.4]

213

where

P2m ( -1)m- 1

k:Q~ 1

metres 2/second

(11.47)

~
kBQ 1
B.='l's(O, 7T/2)+ ~ q2m(-1)m- 1 Q
metres 2/second
2 2

(11.48)

A.= 'l'c(O, 7T/2) +

J
J

Q3
1 [
1rkBQ 5
cl>c(0,9)Q d9+-Q
[p2m(-1)m- 1 Q 4 ] +
Q8 2
2
2 m=1
metres 2/second
(11.50)

Mo=

No=

00

Q3
1 [
1rkBQ 5
1
cl>s(0,9)Q d9+-Q
[q2m(-1)m- Q4]+----g--Q
2
2 m=1
2
metres 2/second
(11.49)

rr/2

rr/2

00

The stream and potential function components are given by


'l'c =

1T

exp(- kxs3) sin(klx82 l)

cl>c =

1T

exp(- kx 83 ) cos(kx82 )

metres2/second

(11.51)
I

oo

oo

J
0

metres 2/second

exp( - vlxs21)
.
[ v sm( VXs3) + k cos( vx 83 )] dv
2 k2
v

II
(11.52)

2
metres /second
(11.53)

exp(- vlxs21)
~.
.
.
[ v cos(vx 83 ) + k sm( vx 83 )] dv
2 k2

v +

2
metres /second
(11.54)

The weighting coefficients Pzm and q 2 m are obtained by approximate solution for
a finite number of unknowns of the simultaneous equations

Local hydrodynamic properties

214

Q6
N
'l'c(1, e)- Q 'l'c(1, 7T/2) = ]; P2m/2m
2

metres 2/second

(11.55)

(11.56)

'l's(1,

[Ch. 11

e)-~: \j/(1, 7T/2) = ];1 q2mf2m

metres /second

where

(11.57)

_ _1_ _ _a_1__ ~
2m+ 1 2m+ 3

(11.58)

Q 1 - 2m- 1

(11.59)
(11.60)

(11.61)

(11.62)
(11.63)

Q
7

= sin[ (2m- 1)e] + a 1 sin[ (2m+ 1)e] _ 3a 3 sin[ (2m + 3)e]


2m-1

2m+1

_ _1_ _ _a_1__ ~
Qs- 2m - 1 2m+ 1 2m + 3

2m+3

(11.64)

(11.65)

<l>c(r, e), <j>.(r, e), etc.' in these formulae imply values calculated at (r, e) in the circle
plane.
11.4.3 Swaying Lewis form
The added mass and damping coefficients for sway motions of a Lewis form are

, pB 2 NoPo + Moqo
a22 = tonnes/metre
2
PZ+

qz

(11.66)

215

Hydrodynamic properties of Lewis forms

Sec. 11.4]

2
b' = proB MoPo- Noqo
22
2
pfi + q{j

kN/(metre/second) per metre

(11.67)

, B 3 PoXR + qoYR
2
a42-kN metres/(metre/second ) per metre
2
pfi + q{j

(11.68)

3
b' _B PoYR -qoXR kN metres/(metre/second) per metre
42- 2
pfi + q{j

(11.69)

where

Q9
3a37f"P2
<Ps(O, 8) -Q d8-~Q ---:-

Q9
3a37f"P2
kB( -1)m- 1
<Pc(O, 8) -Q d8-~Q P2m
Q
Qw(11.71)
2 22
2
2 m=1

rr/2

Mo=-

No=-

rr/2

kB( -1)m-1
L
q2m
Q
Qw(11.70)
2 22
m=1
oo

00

X = Jrrt2,!.. (0 8) Qn d8 + rrkB(alP2- a3P4) + ~ P2m( -1)m+1Q


R
't'c
'
Q2
16Q2
LJ
Q2
12
0
2
2
m=1
2
(11.72)

'rl
= Jrrt2,!.. (0 8) Qn d8 + rrkB(alP2- a3p4) + ~ q2m( -1)m+1Q

y
R

't's

'

Q2
2

16Q2
2

LJ
m=1

Q2
2

12

(11.73)

<Pc(r, 8) and <Ps(r, 8) in these formulae imply evaluation at the polar coordinates (r, 8)
in the circle plane. In cartesian coordinates the stream and potential functions are
'l'c = 7r exp(- kx 83 ) cos(kx83 )

<Pc = -

7r

metres 2/second

exp(- kx 83 ) sin(kx82) metres 2/second

(11.74)
(11.75)

'lfs = 7r exp(- kx 83 ) sin(klxs2lf


oo

exp(- vlxs21)
.
k2 + v~ . [v cos(vx83)- k sm(vx83 )]

Xs2
2
:
dv - - k( 2
2 ) metres /second _"
Xs2+Xs3

(11.76)

Local hydrodynamic properties

216

[Ch. 11

<Ps = 7T exp(- kxB 3) cos(klxB21)

exp( =t vlxB21)
.
k2
[k cos(vxB3) + v sm(vxB 3)] dv
2
+v
o
XB2
2
d
+ k( 2
2 ) metres /secon
XB2 +xB3

oo

(11.77)

The weighting coefficients p 2 m and q2 m are found by approximate solution for a finite
number of unknowns of the simultaneous equations
N

'Vc(O, 9)- 'Vc(O, 7T/2) =

L P2mf2m

(11.78)

m=l

'Vs(O, 9)- \j/.(0, 7T/2) =

q2mf2m

(11.79)

m=l

where
(11.80)

kB

f 2m = cos(2m + 1)9] + 2 Q

Q14 +

kB ( -1)m+l
Q
2 2

Q1s

(11.81)

and the stream function for the multipoles is given by


) _ cos[(2m + 1)9] _ kB cos(2m9)
cos[(2m + 2)9] _
r2m+l
2Q2 2mr2m + a\2m + 2)r2m+2

'V2m ( r, 9 -

_ 3a cos[(2m + 4)9]

3(2m + 4)r2m+4

. )
11 82

The following definitions are used in these formulae:


(11.83)

(11.84)

(11.85)

Sec. 11.4]

Hydrodynamic properties of Lewis forms

2a 1 (1 + a3)
8a3
Q 12 =(2m+ 1) 2 - 4 +(2m+ 1) 2 -16

217

(11.86)
(11.87)

Q _ cos(2m9) + a 1 cos[(2m + 1)9] _ 3a 3 cos[(2m + 4)9]


14
2m
2m + 2
2m + 4

(11.88)

(11.89)
11.4.4 Rolling Lewis form
The added mass and damping coefficients for the rolling motions of a Lewis form are

tonne metres 2/metre

(11.90)

b'44 = proB PoYR + qoXR kN metres/(radian!second) per metre (11.91)


16
PZ + q6
,
a24

pB 3 Moqo + NoPo

PZ + q&

=8

kN metres/( radian/second ) per metre (11.92)

b~ 4 = pro8 MoP~; ~oqo


16

Po

qo

rl

kN metres/( radian/second) per metre (11.93)

XR, Y R, M 0 and P0 are defined in equations (11.70)-(11.73). However,p 2m and q2 m


in these equations are now obtained by approximate solution for a finite number of
terms of the simultaneous equations
N

\jlc(1, 9)- 'Vc(1, rr/2) =

L P2mf2m

(11.94)

m=O

'Vs(1, 9)- 'Vs(1, rr/2) =

2: q2mf2m

(11.95)

m=O

where

(11.96)

Local hydrodynamic properties

218

(m=FO)

[Ch. 11

(11.97)

'Vzm in these formulae is given by equation (11.82).

11.5

MEASUREMENTS OF LOCAL HYDRODYNAMIC PROPERTIES

Accurate calculation of the hydrodynamic properties of cylinders of ship-like crosssection is clearly of paramount importance in the prediction of ship motions in waves.
It is therefore somewhat surprising to find that relatively few experiments to verify
these calculations have been carried out.
Vugts published the results of the classic experiments in this field in 1968. He
tested a number of cylinders in the towing tank at the Delft Shipbuilding Laboratory.
Fig. 11.7 shows, in simplified form, the arrangement he used. The 4.2 metres long

Fig. 11.7- Vugts' oscillation experiments.

cylinders were mounted across the tank and an oscillation mechanism was used to
force heave, sway and roll motions in turn. The waves generated by tile cylinder
motions radiated away and were absorbed by beaches at each end of the tank, some
70 metres from the cylinder. The forces and moments necessary to sustain the
cylinder motions were measured and used to determine the added mass, damping
and cross-coupling coefficients for each motion.
Details of the cylinders are given in Fig. 11.8 and some ofthe results are shown in
Figs 11.9-11.11. Cylinders A, Band Call had

H=2,

7T

cr=4

Cylinder A was circular while cylinders B and Chad ship-like cross-sections. Fig.
11.9 shows that the added mass and damping coefficients were virtually identical,

Sec. 11.5]

219

Measurements of local hydrodynamic properties

Cylinder

11:

Shape

Circle

-=v

Ship

-~

Ship

-=v

Rectangle

Rectangle

Rectangle

1.155

0.5

11:

11:

Triangle

~
-I

!=.-

.v

Fig. 11.8- Vugt's cylinders.


I

.,

'I

confirming that these are essentially functions only of beam/draught ratio and section
area coefficient.
Fig. 11.10 shows how the heave added mass and damping increase with beam/
draught ratio; the opposite trend is shown for the sway coefficients in Fig. 11.11.

Local hydrodynamic properties

220

[Ch. 11

L:ewis form
theory - -

2
w V(BI2g)

Oi
N

til

>
"'"'
::9.
"(

QJ

0.5

0
w Y(B/2g)

Fig. 11.9- Hydrodynamic coefficients for three heaving cylinders; H=2, cr=rr/4. (From
Vugts (1968).)

Measurements of local hydrodynamic properties

Sec. 11.5]

0
2.5

Oi
N

co

>

1.5

<:(

~
rl

:S

0.5

w V(BI2g)
Fig. 11.10 -

Hydrodynamic coefficients for heaving rectangular and triangular cylinders.


(From Vugts (1968).)

221

222

Local hydrodynamic properties

[Ch. 11

0
<u V(B~2g)

0.5

w y(B/2g)
Fig. 11.11 -

Hydrodynamic coefficients for swaying rectangular cylinders. (From Vugts


(1968).)

12
Roll damping

12.1 SOURCES OF ROLL DAMPING


According to strip theory the motion damping arises because the oscillating hull
radiates waves away from the ship. For most motions this constitutes the major
mechanism for the dissipation of energy. So strip theory estimates of motion
damping are generally adequate and reasonable motion predictions are usually
obtained.
Rolling is unfortunately an exception to this general rule. The wave-making
damping b44w predicted for the potential flow around most hull forms is only a small
fraction of the total roll damping which is experienced in reality. Additional
r d
important contributions are illustrated in Fig. 12. L
Hull forms with relatively sharp corners at the bilges and/or at the keel will shed
eddies as the ship rolls. This absorbs a good deal of energy and is a significant source
of additional roll damping. Skin friction forces on the surface of1the rolling hull may
also be significant and any appendages will generate forces which oppose the rolling
motion.
Eddy shedding, skin friction and the appendage forces experienced at low
forward speed arise because of the influence of viscosity which is neglected in strip
theory.

12.2 NON-LINEAR ROLL DAMPING: EQUIVALENT LINEARISATION


Wave-making roll damping and the damping due to the appendage forces at high
forward speed are linear in nature (th~t is, the roll damping moment is directly
proportional to the roll velocity). Viscous roll damping is, however, non-linear and
generally proportional to the square of the roll velocity. This means that the pure
sinusoidal roll response to a sinusoidal wave input (equation (8.14)) given by the
linear theory is no longer valid .Moreover, the linear spectral calculation for motions
in irregular waves, which will be described in Chapt~.r 14, is not applicable.

Roll damping

224

...

--

[Ch. 12

--

......

Wave making

0
Eddies

~J)

-~

"j

0
-~

,;_

Skin friction

~
~

--...,...

')

0
Appendage
forces

Fig. 12.1 - Sources of roll damping.

In order to circumvent these unwelcome problems we may calculate an equivalent linear damping coefficient which allows for the effects of the non-linearities but
is used in a linear way. This allows us to continue to use the linear equations of
motion and the spectral techniques for irregular wave calculations. The equivalent
linear damping coefficient is chosen so that the calculated energy dissipated by this
term in the equation of motion is the same as that which is actually dissipated by the
non-linear effects. In general this means that the equivalent linear damping coefficient depends on the roll motion being experienced and a new value of the damping
must be calculated for every situation.
Since the predominant rolling motions experienced at sea occur at the natural roll

Eddy roll damping

Sec. 12.3]

225

frequency, we may simplify the treatment of roll damping non-linearities by


considering only motions at that frequency. Suppose then that the rolling motion is
given by
(12.1)
Then the roll moment exerted about the centre of gravity by the equivalent linear
damping term will be b44 i 4 kN metres. In one roll cycle the work done and energy
dissipated by this linearised damping term will be the integral of the moment times
the angular distance moved:

1T

ro.4

b44

xl0

kN metres

(12.2)

and the equivalent linearised roll-damping coefficient is therefore related to the


dissipated energy by
b44

E
1T

2
ro.4 X4o

kN metres/(radian/second)

(12.3)

12.3 EDDY ROLL DAMPING

Tanaka (1960) conducted a series of model experiments to determine the eddyshedding roll damping characteristics of a number of different h~ll section shapes as
shown in Fig. 12.2. Schmitke (1978) used these results to develop expressions for the
eddy-making damping coefficients for each type of hull section. He postulated that
the force due to the eddy shedding acts at the relevant sharp corner at a radius rb
metres from the centre of gravity. If the roll velocity is i 4 radians/second the local
flow velocity in the plane of the hull section will be rbi4 metres/second. The local
force resisting the roll motion is expressed in the form
(12.4)
where s and dxa 1 are the girth and length of the hull section and CE is a drag
coefficient depending on the hull form. ~Now for sinusoidal rolling motion (equation(12.1)) the force F exerts a moment about G given by
(12.5)
and the energy dissipated by this moment in one roll-cycle is

RoD damping

226

[Ch. 12

G~

UN

Full

Triangular

Round bilge

Fig. 12.2- Classification of section shapes for eddy calculations.

X40

d.x4

= 3 P r6 xlo

ro~4

s CE dxB 1

kN metres

(12.6)

Using equation (12.3) we see that the equivalent linearised local damping coefficient
for eddy making is

b~E

4pro.4
3
= 3;;:x40 rb s CE dxB 1

kN metres/(radian/second)

(12.7)

The total equivalent linearised roll-damping coefficient for the complete hull is
obtained by integrating along the length of the hull:

kN metres/(radian/second)
(12.8)

Eddy roll damping

Sec. 12.3]

227

It remains to determine the drag coefficient CE which varies along the hull.
Schmitke (1978) gave the following empirical formula:

(12.9)

where Z 1 and Z 2 are given as functions of BjKG, y (the inclination of the hull section
at the waterline) and reiD in Fig. 12.3.; re is the effective radius at the keel given by

z,

4
I

!I

0.20

Fig. 12.3- Z 1 and Z 2 for UN sections. (After Tanaka (1960), Schmitke (1978).)

Roll damping

228

re =

~s

4.12-2.69

~~ + 0.823 ( ~~YJ

[Ch. 12

metres

KG
forB< 2.1
s

KG
for->2.1
Bs

(12.10)

and

14.1-46.7 x 40 + 61.7xl0

(12.11)

with x40 in radians.


Tanaka (1960) found that equations (12.9) and (12.11) also applied to very full
almost rectangular sections (typical of the midship sections of merchant vessels) with
r e now equal to the radius of curvature of the bilge and Z 2 = 1.
For triangular sections at the aft end of cruiser stern ships Schmitke fitted the
following quadratic to Tanaka's data:
CE

0.438-0.449 (BJKG)

+ 0.236 (B

(12.12)

/KG)2

Round bilge sections have negligible eddy-shedding roll damping and


(12.13)
for these forms. Fig. 12.2 shows the definition of rb for each of the classes of section
shape considered.

12.4 SKIN FRICTION ROLL DAMPING


'

The water flowing past the ship's hull exerts frictional forces on the hull surface. It is
usual to express the force acting on a small element of the hull surface in terms of a
non-dimensional local skin friction drag coefficient defined as
frictional force on element
CF = ~--~----~--~------------! p x (local velocity)Z x area of element

(12.14)

Consider a girthwise elementos of length oxB 1 metres as shown in Fig. 12.4. Let
the element be positioned at (xB 2, xB 3), a distance r metres from the centre of gravity.
If the roll velocity is 4 radians/second the velocity at the element will be r 4
metres/second and the component velocity tangential to the surface of the hull will be
r x4sin(el + e2) metres/second. eland e2are the polar location of the element and the
slope of the hull surface so that:

Skin friction roll damping

Sec. 12.4]

229

Fig. 12.4- Roll damping due to skin friction.

Xs3

sin el
sin e2

Xsz

0Xs3
cos e2
OS '

cos el

(12.15)

OXsz
OS

(12.16)

Then the frictional force acting on the element os will be


I

oF = Cp! p [r x4 sin eel+ ezw OS OXsl

kN

II
(12.17)

and the moment about the centre of gravity is


(12.18)
The work done by this moment in a complete roll cycle is
X4Q

oF4

ox:4

kN metres

and using equations (12.1), (12.17) ap.d (12.18) this becomes

(12.19)

Roll damping

230

[Ch. 12

Then the equivalent linearised roll-damping coefficient for skin friction is obtained
from equation (12.3) by allowing 3xB 1 and to approach zero and integrating along
the hull and around the girth:

os

b44F

4 Cp
= "3;;:p ffi4 X40

JL, Is (
O

O r

XB3

2
dxB 2

dxB 3 )

~- XBz ~

ds

dxBl

(12.20)

kN metres/(radian/second)

It remains to determine the local skin friction coefficient Cp. Schmitke (1978)
suggested that the Schoenherr formula for the average skin friction coefficient for
'smooth turbulent' flow used in calculations of ship resistance is appropriate:

Cp

0.0004 + [3.36 log 10 (RN)- 5.6] -z

(12.21)

where the Reynolds number is based on the forward speed and length of the ship:
_ P U Ls
RN-

(12.22)

f.l-w

This is clearly inappropriate if the forward speed is zero, and Kato's (1958)
formula may then be used:
Cp

1.328 R-N 5 + 0.014 R-N 114

(12.23)

where RN is now a Reynolds number based on the average rolling velocity and the
average distance rfrom the centre of gravity:

(12.24)

=-

1T

[(0.887 + 0.145 CB) (1.7 D 8 + Bs CB) + 2(KG- D 8 )]

metres

(12.25)

CB is the block coefficient of the hull defined as

(12.26)

Sec. 12.5]

12.5

231

Appendage roll damping

APPENDAGE ROLL DAMPING

12.5.1 Drag forces on appendages


At zero forward speed the incidence induced on the appendages by the roll motion is
90 and the resulting drag force provides a contribution to the roll damping as
illustrated in Fig. 12.5(a).

(a)

U=O
F=D

(b) U>O

F=L cos a+D sin a


=L

Fig. 12.5- Roll damping due to lifting surfaces.

If the roll velocity is 4 radians/second and the appendage is located at a radius r A


metres (measured from the centre of gravity to the mid-span of the appendage) the
roll motion will impart a transverse velocity r A 4 metres/second to the appendage.
: 11
The resulting drag force on the appendage will be

(12.27)
where C0 is the non-dimensional drag coefficient.
The drag force yields a roll damping moment
(12.28)
and the energy dissipated in one roll cycle is
X40

F4 dx 4

kN metre;;

(12.29)

Using equation (12.27) this becomes


(12.30)

Roll damping

232

[Ch. 12

and equation (12.3) then gives the equivalent linearised damping coefficient due to
appendage drag forces at zero speed as

b44AD

= 4Cn
17' p X4o Oh4 ""'
L.,;
3

3
AA r A

)
kN metres/ ( radtan!second

(12.31)

where the summation is over all appendages. A suitable value for the drag coefficient
is given by equation (2.66).
12.5.2 Lift forces on appendages
If the forward speed is not zero the rolling motion induces an angle of attack on each
appendage as shown in Fig. 12.5b. The angle of attack is

ex

tan-

1('A:4)

radians

(12.32)

radians

and the total velocity experienced at each appendage is


q

y(U 2 + r1 if) metres/second

(12.33)

= U metres/second

The appendage develops lift and drag forces which are respectively normal and
parallel to the local velocity vector as shown in Fig. 12.5. The total force normal to
the ship's longitudinal axis is

L cos ex + D sin ex kN

=L

kN

(12.34)

if ex is small

Hence if the induced angle of attack ex is small the total roll moment applied to the

ship by the appendage is

(12.35)

and the roll-damping coefficient attributable to the lift forces developed on the
appendages is

Total roll damping

Sec. 12.6]

b44AL

=i

pU

2: dd~L AA d.

kN metres/(radian/second)

233

(12.36)

where the summation is for all appendages. It should be noted that this damping
coefficient is independent of roll angle and no special linearisation techniques are
necessary.
12.6 TOTAL ROLL DAMPING

The total roll damping is obtained by adding the contributions from the individual
roll damping sources discussed above:
b44

b44w

+ b44E + b 44 p + b44Ao kN metres/( radian/second)

for U = 0
(12.37a)

b44w

+ b44E + b44F + b44AL kN metres/(radian/second)

for U > 0
(12.37b)

or
b44

13
Ship motions in regular waves

13.1 INTRODUCTION
The strip theory outlined in previous chapters may be used to estimate the motions a
ship would experience in regular sinusoidal waves of small amplitude. For conventional ships at moderate speeds these estimates are usually found to be of adequate
accuracy for everyday engineering purposes. As an example this chapter gives the
results of a specimen set of calculations of the motions of a frigate of length 125
metres and explains ,the physical reasons for their nature. Ship motions are of course
functions of hull shape and size, and the results given here should not be used to give
numerical estimates of the motions of other hull forms. Nevertheless, the same
general characteristics will be found to apply to the motions of all conventional

monohull ships.

13.2 TRANSFER FUNCTIONS


We define the wave depression at the moving origin 0 (equation (8.7)) as
1,; = 1,; 0

sin (ro et)

metres

(13.1)

and the resulting ship motions (equations (8.14)) are taken to be


X;=

xiO sin (roet +

o;)

metres or radians

(i = 1, 6)

(13.2)

The motion amplitudes x ;o and the phases oi are functions of the speed U, heading
p., and encounter frequency roe. The amplitudes are assumed to be proportional to the
wave amplitude 1,;0 and it is usual to express them in non-dimensional form: linear
motion amplitudes x 10 , x 20 and x 30 are non-dimensionalised by dividing by the wave

Vertical plane motions in regular head waves

Sec. 13.3]

235

amplitude ~0 ; angular motion amplitudes x 40 , x 50 and x 60 are divided by the wave


slope amplitude k~ 0
Graphs of the resulting non-dimensional amplitudes plotted as a function of
encounter frequency are called transfer functions: they give the proportion of wave
amplitude or wave slope amplitude 'transferred' by the ship 'system' into the ship
motions.
The phase angles O; give the phase relationship between the motion and the wave:
a positive value means that the maximum positive motion occurs o;lroe seconds
before the maximum wave depression is experienced at 0. Negative values imply
that the motion lags the wave depression.
13.3 VERTICAL PLANE MOTIONS IN REGULAR HEAD WAVES
We begin by considering the simplest case of ship motions in regular head waves
(f.L = 180). Symmetry ensures that roll, sway and yaw are absent and the motions are
confined to surge, heave and pitch.
The heave and pitch equations ((8.25) and (8.27)) are coupled so that heave
motions are influenced by pitch and vice versa. However, the coupling is usually
fairly weak and to a first approximation we may regard the equations as independent.
The heave and pitch motions then approximate to the motions of two independent
second-order spring-mass systems as described in Chapter 6. The analogy is not
rigorous because the coefficients in the equations are frequency-dependent, in
contrast to the constant coefficients assumed in the classical equations. Nevertheless,
we may define approximate natural frequencies for heave and pitch using equation
(6.8):

~(~::)
= ~(~::)

ro.3 =

radians/second

ffis

radians/second

II

(13.3)

(13.4)

where a33 and a:ss-!!r:_e to be evaluated at the natural frequency.

0
G(__ ____J/

Fig. 13.1 -Maximum heave and pitch excitations in very long waves.

236

Ship motions in regular waves

[Ch. 13

The surge equation is independent of all the other equations and has no stiffness
term c 11x 1 Surge motions would therefore be expected to be approximately
analogous to those of a damped system with no stiffness and there is no natural surge
frequency.
In very long waves the encounter frequency roe is very low and dynamic effects
associated with added mass and damping are virtually negligible. So the excitations
and reactions experienced by the ship are almost wholly attributable to the buoyancy
changes as the waves pass the hull. Maximum pitch moment occurs at the wave nodes
and maximum heave force occurs at the wave crests and troughs as shown in Fig.
13.1.
These large excitations in very long waves result in the large motion amplitudes
illustrated in Fig. 13.2. For moderate ship speeds the wave celerity is very much
greater than the ship speed and the vessel may be regarded as virtually stationary as

Fig. 13.2- Motions in very long head waves.

Sec. 13.3]

Vertical plane motions in regular head waves

237

the wave passes by. The ship will behave more or less like a particle of water at the
surface, following a circular orbit of radius ~ 0 given by equation (3.32). So maximum
heave (equal to ~ 0) will occur at the wave crests and troughs and maximum surge
(also equal to ~ 0 ) will be experienced at the wave nodes. The ship surges towards
the approaching crests and recedes after the crest has passed by.
Viewed from the vessel's deck the ship will appear to be crawling like a tiny ant
over a succession of very long shallow hills. The ship will always be aligned with the
wave surface so that maximum pitch (equal to the wave slope amplitude k~ 0) will
occur at the wave nodes.
In shorter waves the buoyancy forces alternate along the ship's hull as shown in
Fig. 13.3. This, together with the growing importance of dynamic effects at the
higher encounter frequencies, results in a general reduction in excitation in shorter
waves. Smith measured the total excitations experienced by a restrained model of the
Friesland Class destroyer in regular head waves in 1967. He used apparatus similar to
that shown in Fig. 8.5 and some of his results are shown in Figs. 13.4 and 13.5. These
show that the ship only experiences significant excitations when the waves are longer
than about three-quarters of the ship length.
Typical calculated heave and pitch transfer functions are shown for the 125 metre
frigate in head waves in Figs 13.6 and 13.7. As expected, all responses approach unity
at zero encounter frequency, corresponding to the long wave case discussed above.
The responses are generally reduced at higher frequencies because of the
substantial reductions in excitation experienced in these shorter waves. However, as
the speed is increased, the range of wave lengths having significant levels of

(a) Long waves i,>>Ls

(b) Medium waves i.=Ls

(c) Short waves .<<Ls

Fig. 13.3- Buoyancy forces on a restrained ship in regUlar waves of constant amplitude.

238

Ship motions in regular waves

[Ch. 13

Wave length/ship length /../L 5

'

\.

\
\

1.6
Encounter frequency we (radians/second)

1oor---------------------------------------------------,

0-=--------~
-"'

Q)

Q)

.;;:. -100-

0.2

0.4

0.6

0.8

1.0

1.2

1.4

1.6

Encounter frequency w. (radians/second)

Fig. 13.4- Heave exciting force; Friesland class destroyer in head seas. (After Smith (1966).)

excitation is encountered over a wider range of frequencies. This range of frequencies may eventually include the natural frequencies of heave and pitch given by
equations (13.3) and {13.4). The responses may then exhibit resonant peaks as
shown at 30 knots in Fig. 13.6. Pitch and heave motions are, however, invariably
heavily damped and the resonant peaks are never very pronounced.
Figs 13.6 and 13.7 also give the heave and pitch phases. In very long waves the
heave phase

Sec. 13.4]

Vertical plane motions in regular following waves

239

Wave length/ship length /../L 5

l.orx--~----T---T-T4~3r---~2~~1T.5~-----1T---~~----~r----0.,.5

'

'

Encounter frequency

we (radians/second)

(/)

Q)

- - - - - - ----~~

g' -90

"0

1.6

0
Encounter frequency

We

(radians/second)

Fig. 13.5- Pitch exciting moment; Friesland class destroyer in head seas. (After Smith (1966).)

indicates that the heave motion is synchronised with the wave motion and that
maximum heave (down) occurs in the wave troughs.
The pitch phase is

at zero encounter frequency .This indicates that maxi.mum positive (bow up) pitch
motion occurs one quarter of an encounter period after the wave trough has passed
the ship's centre of gravity (see Fig. 13.2).
-

240

Ship motions in regular waves


X

[Ch. 13
0.5

30 knots

2.5

Encounter frequency

O>e

(radians/second)

2oor-----------------------------~

-200

Encounter frequency w. (radians/second)

Fig. 13.6- Heave transfer functions for a frigate in head waves.

At higher frequencies these simple phase relationships are modified soQlewhat by


dynamic effects and coupling with the other motions. Nevertheless, they remain
largely true over much of the range of frequencies over which appreciable motions
are experienced.

13.4

VERTICAL PLANE MOTIONS IN REGULAR FOLLOWING WAVES

In following waves the motions are again confined to surge, heave and pitch. Figs
13.8 and 13.9 show calculated heave and pitch transfer functions in regular following
waves. These again approach unity when the waves are very long and the encounter
frequency approaches zero. Only a limited range of (positive) frequencies can be
encountered at any given speed for the reasons discussed in Chapter 7. The transfer
functions consequently adopt the shapes shown with two possible motion responses
(corresponding to different wave lengths) at any one positive encounter frequency.

Vertical plane motions in regular following waves

Sec. 13.4]

I'
I'

1.5

... I"'

. 0.

r'

0.5
30 knots

iJLs
k ....

knots

241

120 knots
1 0 knots

JJ

:": 1.0
:il

'<

0.5

2.5

0
Encounter frequency'''" (radians/second)

2.5

0
Encounter frequency

'''e (radians/second)r fl

Fig. 13.7- Pitch transfer functions for a frigate in head waves.

A third motion response occurs at the corresponding negative encounter frequency


(when the ship overtakes the waves). For moderate ship speeds this will only occur in
very short waves: the excitations and resulting responses are then usually very small.
The heave phase is close to

over most of the range of encounter frequencies for which the response is significant,
indicating that the heave motion is again nearly synchronised with the wave motion.
The pitch phase is now approximately
.

242

[Ch. 13

Ship motions in regular waves

.IJ

.;;

c
0

.E
c

.2

(/)

"'

(.J1

~
Q)

>

"'

Q)

0
-0.6

0.4
Encounter frequency

Ole

0.6

(radians/second)

0
Ui'
Q)
~

Ol
Q)

'"

/.:

Q)
(/)

"'c.

.!:
Q)

>

"'
Q)

-0.6

0
Encounter frequency

Ole

0.6
(radians/second)

Fig. 13.8- Heave transfer functions for a frigate in following waves.

over most of the significant range of encounter frequencies. Maximum positive (bow
up) pitch motion now leads the maximum wave depression at the centre of gravity by
approximately one-quarter of the encounter period.

Sec. 13.5]

Lateral plane motions in regular oblique waves

243

.0'

'if
c
0

'

-g
c

:::J

....

Q)

( /)

tn
~

,...

..c

.B

0::

Encounter frequency we (radians/second)

15
Ui
~

-100

Cl

Q)

~
Q)
(/)

-200

co
..c

c.

..c
(.)

J:

-300

-0.6

0.6

Encount frequency we (radians/second)

Fig. 13.9- Pitch transfer functions for a frigate in following waves.

13.5

VERTICAL PLANE MOTIONS IN REGULAR OBLIQUE WAVES

In oblique waves the ship motions are no longer confined to the vertical plane. Roll,
sway and yaw motions are also present. However, the vertical plane equations of
motion (8.23), (8.25) and (8.27)) for a symmetrical shiP are independent ofthose for
the lateral plane ((8.24), (8.26) and (8.28)). So the lateral plane motions in oblique

244

Ship motions in regular waves

[Ch. 13

waves of small amplitude will have no effect on the vertical plane motions and these
may therefore be considered in isolation.
In very long oblique waves the ship again appears to be crawling over a succession
of long shallow hills as shown in Fig. 13.10. At the crests and troughs the heave
motion will again equal the wave amplitude, exactly as in head and following waves.
The 'effective wave length' measured along the ship's track is
A sec IL metres
and the corresponding effective wave slope amplitude is therefore reduced to

21T/,;o
cxw = A

sec IL

= k/,;0 cos IL

radians

(13.5)

In these long oblique waves the ship will again always align itself with the wave
surface, and the maximum pitch, equal to k/,; 0 cos fL, will occur at the wave nodes.

~0

f
Fig. 13.10- Heave and pitch motions in very long oblique seas.

Figs 13.11 and 13.12 show typical oblique wave transfer functions for the frigate
at 20 knots. On headings forward of the beam (90 < IL < 180) the responses are
broadly similar in general form to the head wave responses already discussed. Only
one response is possible at any given encounter frequency and the motions generally
decrease with increasing encounter frequency. The heave responses increase as the
heading approaches 90 and the wave excitation becomes synchronised along the
entire length of the hull. The pitch response decreases as the heading approaches 90
and would be zero in beam waves if the hull had fore/aft symmetry.

Sec. 13.5]

Lateral plane motions in regular oblique waves

245

I~= :x

.:;

c
0

-~

....::::>
....Q;
Cll

~
Q)

>

"'
::r:
Q)

Encounter frequency

We

(radians/second)

(i)
Q)

Cl

Q)

"0

--;:;

.-c

Q)

Cll

"'c.

.<=
Q)

>

"'
Q)

::r:

-0.5

2.5

0
Encounter frequency

We

(radians/second)

Fig. 13.11 -Heave transfer functions for a frigate in oblique waves; speed 20 knots.

On headings abaft the beam (0 < p, < 90) the responses adopt the general form
of those already described fo:r following waves. The ,range of possible encounter
frequencies is reduced, depending on the heading, and more than one response is
possible at any given encounter frequency.
-

Ship motions in regular waves

246

[Ch. 13

0
-0.5

2.5
Encounter frequency we (radians/second)

(i)
Q)

C)
Q)

:9.

"'

tO

Q)

"'"'0.

120

..r::;

..r::;

....0

a:

-0.5

2.5

Encounter frequency we (radians/second)

Fig. 13.12- Pitch transfer functions for a frigate in oblique seas; speed 20 knots.

Heave phase is always


03=00

in very long waves, indicating that heave is synchronised with wave depression at all
headings. Pitch phase is

Lateral plane motions in regular beam waves

Sec. 13.6]

247

85 =

90 on headings forward of the beam (90 < .f.t < 180)

85 =

270 or 85 = + 90 on headings abaft the beam (0 < f.t < 90)

and

It is sometimes more convenient to present the motion responses as functions of

wave frequency or non-dimensional wave length. Fig. 13.13 shows the frigate's
oblique wave transfer functions plotted in these forms. They have the singular
advantage that the responses are now all single-valued and the complications of
multivalued responses at a given encounter frequency are avoided.
1.5

1.5

1.0
0

~
.;(

0.5

0
Wave frequency"' (radians/second)

2.0
Wave frequency w (radians/second)

Wave length/ship length

i.IL,

Wave length/ship length

'telL,

Fig. 13.13- Alternative presentations of oblique wave transfer functions for a 125 metre
frigate at 20 knots.

13.6 LATERAL PLANE MOTIONS IN REGULAR BEAM WAVES

In beam waves pitch motions are, as we have already seen, usually very small. Yaw is
usually also negligible and the ship motions are essentially confined to heave, sway
and roll. Fig. 13.14 illustrates these motions in veryJong waves. The ship again
follows the circular orbit of a particle of water at th~ surface. The heave and sway
motions are therefore equal to the wave amplitude 1;0 : maximum heave motion

248

Ship motions in regular waves

[Ch. 13

Fig. 13.14- Motions in very long beam waves; f.L = 90.

occurs at the wave crests and troughs (as already seen) and maximum sway occurs at
the wave nodes. The ship sways towards the approaching wave crest and recedes
after the crest has passed by.
If there are no internal free surface effects to reduce the effective metacentric
height (see Chapter 10) the ship's deck will always be aligned with the wave surface.
Maximum roll, equal to the wave slope amplitude k~ 0 , will occur at the wave
nodes. Fig. 13.15 shows the sway transfer function in beam waves. The sway
equation (8.25) has no stiffness term c22x2 so there is no sway resonance. Sway
amplitudes decrease with increasing encounter frequency and the phase remains
essentially constant with

Lateral plane motions in regular beam waves

Sec. 13.6]

249

2.0

w=w. (radians/second)
0

en
Q)

Cl
Q)

:s
.s
r

-150
0

2.0 I

w=w. (radians/second)
Fig. 13.15- Sway transfer function for a frigate at 20 knots in beam waves.

indicating that maximum positive sway (to starboard) occurs one-quarter of an


encounter period after the wave trough has passed by (see Fig. 13.14).
Fig. 13.16 shows roll transfer functions for the fiigate in beam waves. Roll motion
is affected by the sway and yaw motions (see equation (8.26)) but in beam waves the
yaw coupling is negligible because there are practically no yaw motions. The sway
coupling, though of significant proportions, does not alter the basic second-order
spring-mass system chracteristics of the roll motion. The roll motion is usually lightly
damped so that there is a pronounced-resonance close to the undamped natural roll
frequency, given approximately by

Ol- 4 =

~ ( ~:)

radians/second

(13.6)

250

Ship motions in regular waves

[Ch. 13

v=we (radians/seconds:

w=we (radians/second)

Fig. 13.16- Roll transfer functions for a frigate in beam waves; GMF = GM 5 .

The damping increases with forward speed. This gives a general reduction in the
peak roll response and a slight reduction in the frequency at which the peak roll
response occurs.
At zero frequency the roll phase is

Sec. 13.6]

Lateral plane motions in regular beam waves

251

indicating that positive roll (to starboard) leads the maximum wave depression by
one-quarter of a period as shown in Figs. 13.14 and 13.17. At the natural roll
frequency the roll phase is

and the maximum roll is then synchronised with wave crests and troughs as shown in
Fig. 13.17. At very high frequencies

and the ship then rolls in opposition to the wave slope. Roll motions are then,
however, quite small.

c<:J

c<:J

Fig. 13.17- Rolling motions in regular beam waves; !L = 90.

The responses shown in Fig. 13.16 were calculated with no allowance for free
surface effects on the metacenttic height. If these effect~ are significant the reduction
in the roll stiffness reduces the natural roll frequency and increases the roll response
at low frequencies as shown in Fig. 13.18.
-c

252

Ship motions in regular waves

[Ch. 13

<>=<>e (radians/second)

'''="' (radians/second)
Fig. 13.18- Roll transfer functions for a frigate in beam waves; effect of metacentric height;
IL =goo.

13.7 LATERAL PLANE MOTIONS IN REGULAR OBLIQUE WAVES


Fig. 13.19 shows roll transfer functions for the frigate in regular oblique waves. These
are plotted in the alternative wave frequency form and the fluid and solid metacentric
heights are assumed to be equal.
In very long waves (ro = 0) the roll motion amplitude approaches the effective
wave slope amplitude X 10 given by equation (13 .5). In bow waves 90 < p, < 180) the
forward speed of the ship increases the frequency of encounter and the roll resonance
is excited at lower wave frequencies. In quartering waves (0 < p, < 90) the range of
encounter frequencies is limited. At p, = 30 in the case shown the waves are never
encountered at the roll natural frequency and the roll resonance is never excited.
However, at 11=60o a very wide range of wave lengths (and wave frequencies) is

Sec. 13.8]

Absolute motions

253

,5\

..

~
)(

c
0

.;=
u

::J

(/)

~
0

0:::

Wave frequency w (radians/seconds)

Fig. 13.19- Roll transfer functions for a frigate at 20 knots in oblique waves.

encountered at frequencies close to the natural roll frequencYrK see Fig. 7.4) and the
roll response is significantly increased. For this reason roll motion is often a
maximum in quartering seas, particularly at high speed.
Figs 13.20 and 13.21 show the sway and yaw transfer functio'ns in oblique waves.
In very long waves (ro = 0) the sway amplitudes approach the athwartships component of the wave orbit radius ~0 sin p,. On headings forward of the beam (90 < JL <
180) both sway and yaw amplitudes decrease rapidly with wave frequency. Maximum responses occur in quartering seas and rise to very high values when the
encounter frequency approaches zero. Strip theory predictions are likely to be
inaccurate in these circumstances. In practice the ship would be steered by a
helmsman or an autopilot and this would effectively limit these large-motion
amplitudes.

13.8 ABSOLUTE MOTIONS

The six motions considered so far completely define the possible movements of a ship
in a seaway. However, seakeeping studies often call for assessments of the motions
experienced at some particular; point on the ship, such ~s the bridge or the flight deck
of a warship. These can be calculated from a knowledge of the six motions we have
already defined with respect to the centre of gravity~"

254

Ship motions in regular waves


[!=30
w.=O

[Ch. 13

~t=60

we=O

Wave frequency w (radians/second;


Fig. 13.20- Sway transfer functions for a frigate at 20 knots in oblique waves.

~
X

c
0

"
c

.2

;:

>-"'
0

Wave frequency w (radians/second)


Fig. 13.21- Yaw transfer functions for a frigate at 20 knots in oblique waves.

The angular motions are the same everywhere in the ship but the local linear
motions depend on the location within the ship. Let us consider a location defined by
the coordinates (x 81 , x 82 , x 83 ) with respect to the centre of gravity. The longitudinal
displacement of this point includes contributions from the surge of the whole ship as

Sec. 13.8]

255

Absolute motions

well as the products of the lever arms and the pitch and yaw motions. If the angular
motions are small the longitudinal displacement relative to the moving origin 0 is
positive forward

(13.7)

Similarly the lateral and vertical displacements of the point (xB 1, xB 2, xB 3 ) are
positive to starboard

(13.8)

positive down

(13.9)

Substituting equations (13.2) we find that each motion is sinusoidal with


S;

= S;o sin (ole(+ 6

5 ;)

metres

(i = 1,3)

(13.10)

where the amplitudes are

s10 = (Pj + Pl) 112

metres

(13.11a)

Szo

= (P + Pi0) 112

metres

(13.1lb)

S3o

= (Pi1 + Pi2) 112 metres

(13.11c)

and the phases are given by

Ps

tan 651 = p

(13.12a)

Pw
tan 652 = p

(13.12b)

(13.12c)
where
(13.13a)
(13.13b)

P9 = x 20 cos 62 P10 = x 20 sin 62 -

xB~ 40
xB~ 40

cos 64 + xB 1 x 60 cos 6 6

metres

(13.13c)

sin ~ 4 + xB 1 x 60 sin 6 6 metres

(13.13d)
(13.13e)

P12 =

x 30

sin o3 + xB 2x 40 sin 64- xB 1x 50 sin~ 5 metres

(13.13f)

256

Ship motions in regular waves

[Ch. 13

The form of the absolute motion transfer function depends of course on the
position considered in the ship. Fig. 13.22 shows some typical head wave absolute
vertical motion transfer functions for a point on the bridge of the frigate. In very long
waves (ro = 0) the transfer functions approach unity as the ship contours the waves.
At very high frequencies the motions become negligible; but at intermediate
frequencies the motion phases are such that the contributions from pitch and heave
are synchronised and large absolute motions, considerably greater than the wave
amplitude, are the result.
0

2.5

"'c0

.E
c

.2

~Ul
c

~
c
0

g
E

.g"'
Q;
>
Ql

:;
0

Ul
.Q

<1:

Wave frequency oJ (radians/second)

Fig. 13.22- Absolute motion transfer functions at the bridge of a frigate in head waves.

Wave frequency w (radians/second)

Fig. 13.23- Relative motion transfer functions for the bow of a frigate in head waves.

Sec. 13.10]

Velocities and accelerations

257

13.9 RELATIVE MOTIONS


Slamming and deck wetness are of considerable importance in assessing the seakeeping performance of a ship. These qualities are largely determined by the magnitude
of the relative motion between the hull and the adjacent sea surface.
The relative vertical motion at a point (x 81 , X 82 , x 8 3) is given by
positive for increasing immersion

(13.14)

where s 3 is the absolute vertical motion given by equation (13.9) and~ is the local
wave depression given by equation (9.32). Substituting equations (9.32) and (13.9)
we find that the relative motion is sinusoidal with
(13.15)
where the relative motion amplitude is
(13.16)
and the phase is given by
(13.17)
In practice the presence of the hull causes a considerable distortion of the waves
close to the ship and equation (13.14) is only likely to be reliable at the forward
perpendicular. Further aft the equation may underestimate thd felative motion by as
much as 50%. Techniques for estimating this distortion or 'swell-up' are still the
subject of research and no method has yet won universal agreement. Nevertheless,
equation (13.14) is still used to estimate relative motion, sometimes with empirical
corrections for the swell-up as described in Chapter 20. Fig. 13.23 shows some typical
calculations for a point on the forefoot of the frigate in head waves. In very long
waves the relative motions are zero because the ship contours the waves. In short
waves the ship is essentially stationary so that the wave motion is the only sizable
contribution to the relative motion and the transfer function approaches unity.
At some intermediate frequency the absolute, motion phase is such that the
upward absolute motion is synchronised with the wave depression at the particular
location chosen for the calculation. The relative motion is then a maximum and
sharply peaked resonances can occur at high speed.

13.10 VELOCITIES AND ACCELERATIONS


Since all the motion displaceme.nts are of the form

[Ch. 13

Ship motions in regular waves

258

the motion velocities and accelerations are given by

which may be written

where the velocity and acceleration amplitudes are

Xo

= Xo (J)e,

..

Xo

= Xo

2
(J)e

So the velocity and acceleration transfer functions for any motion can simply be
obtained by multiplying the displacement amplitude responses by the encounter
frequency and the square of the encounter frequency. Fig. 13.24 shows examples of
this procedure for heave in head and following waves.

2.0
Wave frequency oJ (radians/second)

Wave frequency oJ (radians/second)

Fig. 13.24- Heave displacement, velocity and acceleration transfer functions for a frigate at 20 knots.

Sec. 13.11]
13.11

Lateral force estimator

259

LATERAL FORCE ESTIMATOR

Many seakeeping studies call for estimates of the motions experienced by passengers
and crew and inanimate objects within the ship. These motions are important in
determining seasickness and the ability of the crew to work effectively as well as
estimating the likelihood of unsecured objects sliding across the deck or toppling
over. These problems are discussed in more detail in Chapter 21, but for the time
being we may observe that they are essentially functions of the apparent accelerations experienced by the people and objects within the ship.
Consider, for example, an object of mass m tonnes on the deck of a ship as shown
in Fig. 13.25. If the object is at (xB 1, xB 2,xB3) the absolute lateral and vertical motion

I //1
I

s3.

Ss3;

I
I
Fig. 13.25- Apparent forces on an object on the deck.

displacements relative to the mean track of the ship are given by equations (13.10).
Following the procedure outlined in Section 13.10 the lateral and vertical accelerations, relative to the Earth are given ~y
positive to starboard
(13.18)
positive downwards
(13.19)

Ship motions in regular waves

260

[Ch. 13

Resolving these accelerations in directions normal and parallel to the deck we


obtain
sB 2

= Sz cos x 4 + s 3 sin x 4

sB 3

= s 3 cos x4 -

s2

metres/second2
in the plane of the deck to starboard

sin x4 metres/second2
downwards normal to the deck

These accelerations of the ship will tend to lift the object off the deck and slide it
towards the port side. The object is also subject to the components of gravity
resolved in directions normal and parallel to the deck. The total apparent force
experienced by the object in the plane of the deck is
msB 2 - mg sin x4 kN

to port

When the ship is upright the gravity force on the object is mg kN downwards, but this
is reduced to mg cos x 4 kN normal to the deck when the ship is rolled. So the total
apparent force normal to the deck is
upwards

The apparent accelerations perceived by the object are then

s AZ

sA3

= sB 2 - g sin x 4
= s2 cos x4 + s3 sin x4 =

sB 3 -

g (1- cos x 4) =

s3

g sin x 4

cos x4 -

metres/second 2
s2

to starboard

sin x4 - g (1- cos x4)'


metres/second2
downwards

For small-amplitude motions these reduce to


s AZ

= Sz -

gx4 metres/second 2

sA3

=s3

metres/second2

to starboard

downwards normal to the deck

(13.20)

So the apparent acceleration normal to the deck is the same as the absolute vertical
acceleration. The apparent acceleration in the plane of the deck is often called the

Lateral force estimator

Sec. 13.11]

261

lateral force estimator (LFE). Substituting equations (13.2) and (13.18) in equation
(13.20) we find that the LFE in regular waves is sinusoidal with

sA2 = sAzo sin (roe t + BAz)

metres/second 2

positive to starboard

The LFE amptltude is

and the phase is given by


p14

tan OAz =-p


13

~
Q)

E
Q;
Cl.

'bc:
0

()

Q)

"'

0;
~

Ql

.S
0

J.J>

6N
..:

'"'
Q)

"0

::J

.'!:

c.
E

"'
Q)

>

"'

~
_;
"0

Wave frequency w (radians/second)

1.0

c.

E
w

"'

LL

...J

2.0
Wav.e frequency w

(radians/sec~nd)

Fig. 13.26- LFE transfer functions for a frigate at 20 knots;xB 1 = 40 m, xB2 = 7 m, xB3= 10m.

262

Ship motions in regular waves

[Ch. 13

where

P13 = - s20
P14 =

ro~

- Szo ro~

cos lls2 - gx40 cos b 4

metres/second2

(13.21a)

sin lis2 - gx40 sin b 4

metres/second2

(13.21b)

The LFE varies throughout the ship and the LFE transfer function will depend on
the location chosen for the calculation. Fig. 13.26 shows an example for a particular
location on the frigate at 20 knots.

13.12 NON-LINEARITIES

The foregoing discussion of the characteristics of motion transfer functions presupposes that the motion responses are linear: that is, the motion amplitude at any
particular speed, heading and wave frequency is directly proportional to the wave
amplitude. Indeed this assumption is implicit in the definition of the transfer
function.
The equations of motion (8.23)-(8.28) will only yield linear motion responses if
all the coefficients aii' bii and cii are independent of the motion amplitude. In
principle this ~equires negligible changes of underwater hull form as the ship
proceeds through the waves: in other words the relative motion amplitudes must be
small.
If the motion amplitudes are large the underwater shape may change considerably. Consider, for example, the relative motion at the bow of a ship in head waves
(Fig. 13.23). If the motion exceeds the local draught of the ship the keel will emerge
from the water during the upward part of the motion cycle. The local excitation and
all the hydrodynamic reactions will temporarily disappear until the keel re-enters
and they resume finite values. This might be expected to have a dramatic effect on the
motion responses in these conditions but it is found that the non-linearities associated with this kind of phenomenon are surprisingly weak. Perhaps this is because
both sides of the equations of motion are affected and the effects are only ~ssociated
with the extremes of the motion cycles. Extremely large motions are in any case only
of limited interest because they cannot be readily tolerated by the crew.
When the roll damping is very small, large roll motions may develop in quite
moderate waves and non-linearities in the righting lever curve may become important. In other words the effective roll stiffness c44 will depend on the roll angle. The
roll damping itself will usually be dependent to some extent on the roll amplitude and
this will be a further source of non-linear behaviour. These effects are not usually

very in:~portant for ships with adequate levels of roll damping.

14
Ship motions in irregular waves
14.1

THE ELECTRONIC FILTER ANALOGY

For many years the assessment of seakeeping performance at the design stage
progressed no further than comparisons of ship motions in regular waves. The
shortcomings of this approach were widely recognised but further progress had to
await the development of new techniques first proposed by St Denis and Pierson
(1953). These methods were based on ideas developed in the electronics and
communications fields and it is no exaggeration to suggest that their introduction,
together with the development of strip theory, form the two main foundations of the
modern theory of seakeeping.
St Denis and Pierson suggested that the ship could be treated in much the same
way as the 'black box' electronic filter shown in Fig. 14.1. The input signal received

!d

Input

Waves

--

Filter

Output _

Ship

Motion _

Fig. 1~.1- Electronic filter analogy.

264

Ship motions in irregular waves

[Ch. 14

by the filter contains a number of different frequency components and these are
amplified or attenuated to produce some modified output signal according to the
characteristics of the filter. For example, a so-called 'low-pass' filter will attenuate
the high-frequency components of the input signal and allow the low-frequency
components to pass more or less unscathed.
The analogy suggests that the ship can also be regarded as a filter, not of electrical
signals, but of the waves. In other words we can think of the ship as a black box which
receives the waves as input and generates ship motions as output. Of course there are
a number of different ship motion outputs so we should really regard the ship as a
collection of filters, each with its own individual characteristics.
Let us consider the case of heave motion in head waves. Figure 13.6 shows typical
heave transfer functions for p, = 180 and we may regard these as defining the
characteristics of the 'heave filter' of the black box ship. We can see that this is
essentially a low-pass filter: at low frequencies the wave motions are translated into
corresponding heave motions without attenuation or phase shift. As the frequency
rises, the heave motions are reduced and at very high frequencies the input is
completely attenuated so that there are no resulting heave motions.
These ideas can be formalised and quantified by means of the so called 'spectral
calculation'. This is mathematically valid and rigorous provided that the ship motions
are directly proportional to the wave amplitude at any given speed, heading and
frequency. This is nearly always true, as we have seen, and the spectral calculation is
widely used in seakeeping calculations.

14.2 THE ENCOUNTERED WAVE SPECTRUM

The first step is to determine the wave energy spectrum as described in Chapter 4.
For the time being we shall assume that the waves are long crested. The spectrum
may be measured but it is more usual to employ one of the idealised wave energy
spectrum formulae (equations (4.33) or (4.49)). These formulae give the wave
energy spectrum for a fixed point in the ocean: we require to transform this to the
reference frame of an observer on the moving ship.
We have already seen that waves are encountered by the ship at the encounter
frequency defined in equation (7.3). So the frequencies with which the waves are
encountered are increased in head waves and decreased in following waves. It
follows that the wave energy spectrum must be shifted along the frequency axis to
cover a different range of frequencies when observed from a moving ship.
Fig. 14.2 illustrates the result obtained in head waves: every wave frequency is
transformed into a corresponding encounter frequency according to equation (7 .3).
The frequency interval oro centred on the wave frequency ro transforms into a
corresponding encounter frequency interval oroe. The relationship between the two
intervals is obtained by differentiating equation (7.3):
droe = _ 2roU cos p,
1
dro
g
or

Sec. 14.2]

The encountered wave spectrum


-

265

bw

Wave energy spectrum

Areas equal

Corresponding

Encounter spectrum

Fig. 14.2- Transforming the wave energy spectrum into the encounter spectrum.

<>roe = ( 1-

2
: U cos IL )oro

radians/second

(14.1)

Now we have seen in Chapter 4 that the area under the wave energy spectrum
bounded by the frequency interval oro is proportional to the energy contained within
that band of frequencies. Transforming the spectrum to the moving frame of
reference of the ship does not change this energy and it follows that the area within
the wave frequency range oro must be exactly reproduced as an equal area within the
corresponding encounter frequency-range <>roe. Hence the ordinates of the wave
spectrum and its counterpart in the encounter frequency domain must be related by
(14.2)

266

Ship motions in irregular waves

[Ch. 14

or, if oro and oroe are allowed to become infinitesimal,

dro
Sr,(ro)-d
roe

S (ro)
g
metres 2/(radian/second)
r,
g- 2roU cos t-t

(14.3)

A typical transformation of a Bretschneider spectrum into an encountered


spectrum for a ship at 20 knots in head waves is shown in Figs 14.3(a) and 14.3(b).
The effect is to shift the spectrum to a higher, wider range of frequencies and to
reduce its height. The areas under the two spectra are of course, identical since the
total wave energy and the significant wave height are unchanged by the
transformation.

14.3

THE MOTION ENERGY SPECTRUM

The next step is to generate a motion energy spectrum by filtering the encountered
wave energy spectrum with the appropriate motion transfer function. This is
achieved by multiplying each spectral ordinate of the encountered spectrum by the
square of the motion transfer function at the corresponding encounter frequency.
This approach is valid and appropriate for any ship motion when the transfer
function is normalised by dividing by the wave amplitude (e.g. surge, heave, absolute
motion, relative motion, etc.). A typical calculation for heave motion at 20 knots in
head waves is shown on the right-hand side of Fig. 14.3. The heave motion energy
spectrum ordinates (Fig. 14.3(f)) are given by

(14.4)

and similar expressions apply for other motions.


The effect of the motion transfer function filter can clearly be seen. At low
encounter frequencies the transfer function is nearly unity and the encountered
spectrum is reproduced almost exactly as the heave motion spectrum. At higher
frequencies the filter attenuates the waves and the heave energy spectrum ordinates
are virtually zero even though some wave energy is present at these frequencies.
Angular ship motions are usually normalised by dividing by the wave slope
amplitude and this calls for a slightly different procedure, illustrated for pitch motion
in Fig. 14.4. In this case it is first necessary to calculate the wave slope spectrum (Fig.
14.4(a)) as described in Chapter 4. The rest of the calculation follows the procedure

The motion energy spectrum

Sec. 14.3]

Wave frequency
domain

Encounter Frequency
domain

6
(a)

267

(b)

()
Q)

!!'_

""0

1
2

J;
0

0.5

1.0

1.5

1.5

1.5
(c)

0.5

1.0

1.5

(d)

1.0
,'j.

'a

.;;:

.;;:

0.5

0.5

1.0

1.5

1.5

6
(f)

(e)
()
Q)

!!'_
""0

J
0
<J (radians/second)

'" (radians/second)

Fig. 14.3- Calculation of heave motion at 20 kn9ts in irregular head waves.

described above. For pitch motion the pitch energy spectrum (Fig. 14.4(f)) is given
by
(14.5)
and similar expressions apply for other angular motions.
Chapter 4 derived various formulae (equations (4.3), (4.17) and (4.22)-(4.28))
for the statistical quantities associated with wave energy spectra and these may be

Ship motions in irregular waves

268

Wave frequency
domain

[Ch. 14

Encounter frequency
domain

o.ooa----r---........--"""
(a) Wave slope spectrum

::::0.006

T0 =12.4 Sec
Hy =5.5m
3

(.)
Q)

~
'0

0.004

1.5

1.5

1.5
(c)

(d)

1.0
0

JJ

JJ

""'-,

><

)(

0.5

1.5

(e)

2.0

(f)

Q)

0.004

i;

2.0

''' (radians/second)

<>~e

(radians. second)

Fig. 14.4- Calculation of pitch motion at 20 knots in irregular head waves.

applied with equal validity to ship motion energy spectra in the encounter frequency
domain. In particular the rms motion displacement is given by
cr0

= \/m0

metres or radians

(14.6)

Alternative method of calculating motion statistics

Sec. 14.4]

269

where m 0 is the motion variance given by the area under the motion energy
spectrum:

(14.7)
Similarly the rms velocities and accelerations are given by

<Jz

cr4 =

vmz metres/second or radians/second

ym4

(14.8)

metres/second2 or radians/second2

(14.9)

where the variances of velocity and acceleration are

(14.10)

(14.11)

and in general the nth moment of area of the motion energy spectrum is
.

The associated average periods of the motions are theri given by equations
( 4.26)-(4.28).
14.4 ALTERNATIVE METHOD OF CALCULATING MOTION STATISTICS
The motion energy spectrum in the encounter frequency domain may be transformed back into the wave frequency domain as shown in Figs 14.3(e) and 14.4(e).
The ordinates in the wave frequency domain are calculated by rearranging the
transformation formula given by equation (14.3). With appropriate change of
notation this becomes

(14.13)
or radians2 /( radian/second)

Ship motions in irregular waves

270

[Ch. 14

The area under the curve in the wave frequency domain is of course the same as the
area under the motion energy spectrum in the encounter frequency domain. So the
motion variance may also be obtained by integrating in the wave frequency domain:

m0

J:

Sx (ro) dro metres 2 or radians 2

(14.14)

The ordinates of this transformed motion energy spectrum could have been
obtained much more simply from a direct calculation in the wave frequency domain
with none of the complication of transforming from one frequency domain to
another.
For motion transfer functions normalised by dividing by wave amplitude:

Sx(ro) =

S~(ro) ( ~:r metres 7'(radianlsecond)

(14.15)

and for motion transfer functions normalised by dividing by wave slope amplitude:

Sx(ro)

S"'(ro)

(:;J

radians2/(radianlsecond)

(14.16)

It is important to understand that the curve defined by the transformed spectral


ordinates in the wave frequency domain has no physical meaning and is not a true
motion energy spectrum: the motions are experienced at the encounter frequency,
not the wave frequency.
The rms velocity and acceleration and the motion periods may also be calculated
in the wave frequency domain. By analogy with equation (14.2) the motion energy
spectrum and its transformation are related by

Sx(ro) oro = Sx(roe) oroe

metres 2/(radianlsecond)
or radians 2/(radianlsecond)

and using equations (14.12) and (7.3) the spectral moments are

oo (

ro -

g U cos p, )nSx(ro) dro


(!)2

metres2/second n
or radians 2/secondn

(14.17)

This may then be used with equations (14.8) and (14.9) to calculate the required
motion statistics.

Sec. 14.5] Effect of matching the wave spectrum and the transfer function

271

This alternative procedure has particular advantages in following and quartering


seas since it avoids the complexities of multivalued spectral ordinates at the same
encounter frequency. If the true motion spectral ordinates are required they may be
obtained by rearranging equation (14.13):

S (ro)
x

g
g - 2ro U cos fJ-

metres 2/(radian/second)
or radians 2 /(radian/second)

14.5

(14.18)

EFFECT OF MATCHING THE WAVE SPECTRUM AND THE


TRANSFER FUNCTION

High transfer function ordinates occurring at frequencies with a good deal of wave
energy will give large contributions to the motion energy spectrum. It follows that the
rms motion depends on the extent to which the motion transfer function 'matches'
the wave spectrum. Fig. 14.5 shows as an example the effect of varying the modal

10

10

Q)

:a

T0 =9.7 sec

18

'J1

0.5

1.5

"0
~

Wave frequency"' (radians/second)

Fig. 14.5- Effect of matching.the wave energy ~ectrum with the transfer function, Relative
motions at 20 knots, H 113 = 5.5-fnetres.

Ship motions in irregular waves

272

[Ch. 14

period of the wave energy spectrum on the rms relative motion for the forefoot of a
frigate in head waves.
Fig. 14.6 shows these and other motions plotted as a function of modal period.

en

E:

Q)

Head waves

E
.....
c

QJ

E
QJ
(,)

"'
c.
.~
'0

Heave

(/)

Relative motion
at bow

25

0
Modal period T0 (seconds)
4

en
QJ

E:

Cl
QJ

.....
<.:

QJ

E
QJ

(,)

"'c.

.~

'0

head waves

(/)

25

0
Modal period T0 (seconds)

Fig. 14.6- Effect of modal period on ship motions of a frigate at 20 knots; H113 = 5.5 metres.

The motions which are most sensitive to modal period are those, like relative motion
and roll, which have distinct transfer function peaks.

14.6

MOTIONS IN SHORT CRESTED WAVES

The procedures outlined above may be used to calculate motions in irregular long
crested waves. These are rare, as we have seen, and it is sometimes necessary to
extend these techniques to cope with more realistic short crested waves.
When the waves are short crested the 'total' wave energy spectrum S~;(ro) is

Sec. 14.6]

Motions in short crested waves

273

obtained by integrating the spread wave spectrum S~(ro,v) over the range of
directions from - Vmax to + Vmax (see equation (4.58)). Now it was shown in Chapter
4 that the continuous short crested wave spectrum S~(ro,v) could be represented by a
finite number of reduced long crested spectra distributed around the predominant
wave direction (see Fig. 4.15). Each long crested spectrum is given by

W S~(ro)
where W is a weighting factor depending on the secondary wave direction v - p, given
in Table 4.1.
Table 14.1- Specimen calculation of roll motion in short crested waves
v- /J(deg)

(predominant)

-90
-75
-60
-45
-30
-15
0
15
30
45
60
75
90

mo

(deg)
-45
-30
-15
0
15
30
45
60
75
90
105
120
135

0.000
0.011
0.042
0.083
0.125
0.156
0)67
0.156
0.125
0.083
0.042
0.011
0.000

rl

omo

(deg2)

(deg 2)

36.24
12.11
2.92
0.00
2.92
12.11
36.24
73.27
18.23
8.64
4.34
2.72
1.42

0.00
0.13
0.12
0.00
0.37
1.89
6.05
11.43
2.28
0.72
0.20
0.03
0.00

Total variance =22.57 deg2


rms roll is short crested waves = \122.57 = 4. 75 deg
rms roll in long crested waves = 6.03 deg
Cosine squared spreading: Vmax = 90.
Heading relative to predominant wave direction: f.L
Significant wave height 5.5 metres.
Modal wave period 12.4 seconds.
Ship speed 20 knots.

= 45.

It follows that the contribution to the motion variance from each secondary wave
direction is
...

and the total motion variance in short crested wave2 is obtained by summing the
contributions from all the reduced long crested wave spectra.

Ship motions in irregular waves

274

[Ch. 14

Table 14.1 gives the results of a specimen calculation of the rolling motion of a
frigate at 20 knots in quartering waves (JL = 45). In long crested waves the rms roll at
this heading is 6.03 for the particular wave spectrum used in the calculation. Cosine
squared wave spreading reduces this to only 4.75. Fig. 14.7 shows the effect of wave

...c

Q)

E_

--

Q)rn
t.lQ)
_..,.
CO'Q.Q)

E
-c-

.!!!

Heave

rn

E
...
180

0
Heading 1-1 (degrees)
10

(i)
Q)
Q)

,.
I \

e
Q)

...

::!:!.
cQ)
E
Q)

Roll

I
I
\

t.l
co

c.
rn
'0
rn

E
0
Heading 1-1 (degrees)

...c

Q)

EQ)Vl
t.l Q)
coQ)

-C..Cl
...

'

VlQ)

':

rn

E
...

II

\
\

"C
-c_

180
Heading 1-1 (degrees)

Fig. 14.7- Effect of wave spreading on ship motions; ship speed 20 knots, H113 = 5.5 metres,
T0 = 12.4 seconds.

spreading on heave and pitch as well as on roll. In general, wave spreading smooths
out the more extreme variations of the motion. The effects are small for heave but
quite dramatic for roll. Wave spreading results in significant roll motions in following

Sec. 14.7]

275

Spectral calculations for non-linear motion responses

waves and, to a lesser extent, in head waves. It also reduces the roll motions at the
worst heading by a considerable amount. In the same way spreading increases
the pitch motions in beam waves, but the effects at other headings are less
pronounced.
14.7

SPECTRAL CALCULATIONS FOR NON-LINEAR MOTION


RESPONSES

The procedures outlined above rely on the assumption that the motion responses are
linear with respect to the wave amplitude. We have seen that this is usually the case,
but roll motions may be an exception to this general rule. In this case a slightly more
involved procedure, illustrated in Fig. 14.8, is required.

Input speed, heading, significant wave height, modal period

I
I

I
Initial estimate of average roll
amplitude and period
I
Calculate equivalent roll damping ...---- Calculate average roll amplitude

I
Regular wave calculation

I
irregular wave calculation

I
RMS roll amplitude
average zero-crossing period

Difference not acceptable

~ Compare with previous values I

'

Difference acceptable

!
r

Final result:
select new calcualtion
conditions

r---

Fig. 14.8- Motion calculation with non-linear damping.

It is first necessary to estimate or guess the rms roll expected in the particular
combination of speed, heading, significant wave height and modal period for which
the calculation is being performed. The averag~" roll amplitude may then be

276

Ship motions in irregular waves

[Ch. 14

estimated using equation (17.32). This done, the total equivalent linearised rolldamping coefficient may be calculated for the chosen roll amplitude using the
methods outlined in Chapter 12. The average frequency may be taken as the natural
roll frequency of the ship. The calculation then proceeds through the usual stages of
determining the motion transfer functions and combining these with the appropriate
wave spectra to obtain rms values. Wave spreading should be taken into account.
The rms roll results are compared with the initial guess: if the differences are large
(as will usually be the case) the calculation is repeated using the new rms value and an
improved estimate of the roll damping.
This procedure is repeated until the rms roll angle reaches an asymptotic value
and the calculation is terminated. The whole calculation must then be repeated for
every speed, heading, significant wave height and modal period required.

15
Seakeeping trials

15.1

FULL-SCALE TRIALS

Full-scale seakeeping trials, in which the motions, deck wetness and other seakeeping phenomena of interest are monitored in a measured wave environment, may
seem to be an attractive method of assessing and comparing the performance of ships
in rough weather. The waves would, of course, be irregular and it would be necessary
to record them and the motions simultaneously and tg analyse the results using the
spectral analysis techniques discussed in Chapter 4. The rms motions (or other
seakeeping responses such as deck wetness frequency) could then be plotted as a
function of significant wave height and compared with those obtained from trials in
'
i II
other ships.
However, we have seen that the rough weather behaviour of a ship is a function
not only of the significant wave height, but also of the modal wave period, the shape
of the wave spectrum and the degree of wave spreading. So results obtained and
compared in this simple way are likely to be scattered and possibly misleading if the
wave conditions at the time of the trials were in any way dissimilar.
The only way that these problems can be overcome is by running two or more
ships side by side in simultaneous trials in nominally identical wave conditions. Trials
of this nature have been reported by Bledsoe, Bussemaker and Cummins (1960) and
by Andrew and Lloyd (1981). Fig. 15.1 shows the rms pitch motions and the mean
zero-crossing periods measured on two frigates in the latter trials in severe head
waves (significant wave heights 6-7 metres). These results form an objective
comparison between the motions experienced by two particular ships in a particular
rough weather environment: any pecUliarities or changes in the wave conditions
during the time of the trial must have been experienced in equal measure by both
ships (since they were only about 400 metres apart) so any differences in the motions
may be attributed to differences in the design of the two ships.
Trials with a single ship are infrequent _because of the expense involved, and
comparative trials involving two ships are even more_,9f a rarity. Such trials certainly

278

[Ch. 15

Seakeeping trials
3
U)
Q)
Q)

...

C)
Q)

~
.s::::

...
0

c.
rn

E
22

U)

-o
c
0

Q)

!!!.

h'
-o
0

;::
Q)

0.
C)

c
rn

Trial

u;

eu
6

Q;

Leander

6.

Tribal

Theory

----

c
co
Q)
~

22
Speed (knots)

Fig. 15.1- Comparative seakeeping trial results. (After Andrew and Lloyd (1981)).

cannot be regarded as a routine way of ascertaining the performance of a new design.


In any case, seakeeping trials, by their very nature, cannot be used as part of the
design process since they require the design to be finalised and the ship to be built
before- they can be conducted. Model experiments and theoretical studies provide
the only practical method of estimating the seakeeping qualities of a new ship at the
design stage.
Trials do, however, offer the definitive method of verifying theoretical calculations or predictions based on model experiments. If the wave spectrum is measured it
may be used with estimated motion transfer functions to predict the rms motions and
periods experienced during the trial as shown in Fig. 15.1.
Alternatively the motion transfer functions may be estimated from the ratio of
the measured motion energy spectrum to the measured wave energy spectrum.

Equation (14.4) may be rearranged to give

Sec. 15.3]

279

Ship motion measurements

(15.1)
with similar expressions for other linear motion transfer functions. In the same way
equation (14.5) may be rearranged to give

(15.2)
with similar expressions for other angular motion transfer functions.

15.2 WAVE MEASUREMENTS


The heart of any seakeeping trial is the measurement of the waves. Without a proper
technique for recording and analysing the waves exp~rienced by the ship, the results
obtained can only be related to potentially unreliable visual estimates of the sea state
at the time of the trial. For this reason the use of relatively simple wave buoys such as
the Waveridert (Fig. 15.2) has become de rigeur in all serious seakeeping trials.
The Waverider consists of a stainless steel sphere of 700 mm diameter weighing
106 kg. An accelerometer to monitor vertical accelerations is mounted on a heavily
damped pendulum within the sphere. The pendulum keeps the accelerometer
aligned with the true vertical and the accelerometer's output is integrated twice by
electronic circuits to provide an analogue record of the vertical displ<tcement of the
buoy. The resulting signal is transmitted to a receiver ashore or on board the trials
ship. The buoy may be moored to the sea bottomor allowed to float freely. The
system will give satisfactory measurements of waves covering :tftequency range from
0.22 to 4.1 radians/second, corresponding to a wave length range from about 30 to
1300 m. This is sufficient for most practical purposes except whe:t;e measurements of
very short waves are required (for example, where trials are being conducted on a
small boat).
Simple buoys of this kind can give no information about the directional spreading
of the waves (or even the predominant wave direction). More sophisticated buoys
which can give information on wave spreading are now becoming available and
_techniques of hindcasting the waves from measured wind data have also been used.
These have not yet become routine, however, and most trials still use simple
measurements of point wave spectra, relying on visual observations to determine the
predominant wave direction.

15.3 SHIP MOTION MEASUREMEN~S


Most seakeeping trials are concerned with the measurement of the ship's displacements in the six degrees of freedom. Angular motions are generally measured using

t Registered Trade Mark. Manufactured by Datawell bv, Zomerlhstraat 4, 2012 LM Haarlem, The
Netherlands.

280

Seakeeping trials

[Ch. 15

Fig. 15.2- Waverider buoy. Reproduced by permission of Datawell bv.

gyros of the type used in aircraft navigation systems. In warships it is often possible to
use the ship's own weapon system gyros but on other ships the trials team will usually
have to supply its own transducers.

Measurements of other seakeeping responses

Sec. 15.4]

281

Direct measurements of the linear motion displacements are impossible because


no suitable fixed datum levels are available. Instead, the usual practice is to measure
the surge, sway and heave accelerations using accelerometers mounted on a small
platform stabilised by gyros to remain in a horizontal plane (the gyros may
conveniently be used to measure the roll and pitch). Stabilisation is necessary
because a 'strapdown' accelerometer fixed to the ship's deck would measure the
lateral force estimator rather than the true horizontal acceleration. Surge acceleration would be affected by pitch in the same way. Measurements of heave
acceleration are reasonably immune from these effects provided that the pitch and
roll angles are not too large.
The stabilised accelerometers should ideally be located at the centre of gravity so
that true measurements of the surge, sway and heave accelerations are obtained. In
practice this is often impossible to achieve. Even if the location of the centre of
gravity is known when the transducers are installed, it may well turn out to be in some
inaccessible or inconvenient location. It is, in any case, much more likely that the
exact location of the centre of gravity will not be known at the planning stage and
precise determination of its position will only be possible at the time of the trial.
So in practice it is likely that the transducers will be located at some arbitrary
position relative to the centre of gravity and it will be necessary to correct their
measurements to allow for this error. If the transducers are located at (xB 1, xB 2 , xB 3)
they will measure local absolute accelerations St. s2 and 53 given by differentiating
equations (13. 7)-(13. 9). The true accelerations of the centre of gravity are then given
by

il

St

+ XBzX6- xB3is

metres/second 2

(15.3)
1

+ XB3i4- XBt i6

Xz

i3

= 53 -

Sz

xB 2i 4

+ xB 1i 5

11

(15.4)

metres/second 2

(15.5)

metres/second

15.4 MEASUREMENTS OF OTHER SEAKEEPING RESPONSES


Seakeeping trials are often concerned with the measurement of responses other than
motions in waves. Typical examples are deck wetness and slamming. Very simple
instrumentation will often suffice for-measurements of the frequency of occurrence
of these events. Indeed a seasoned observer with a watch, a pencil and a log book is
really all that is required. More permanent records of deck wetness, which can be
analysed at leisure in the less distracting environment df the shore-based laboratory,
can be obtained by a video recording of the forecastle. This technique was used

Seakeeping trials

282

[Ch. 15

successfully in the comparative seakeeping trials reported by Andrew and Lloyd


(1981) and their results are shown in Fig. 15.3.
Measurements of deck wetness severity can be obtained by mounting pressure
transducers at suitable locations on the ship's upper works.
Slamming frequency measurements may be obtained by using strain gauges to
monitor the bending moment experienced by the main hull girder. Slams will then be
readily detected as short periods of high-frequency (typically 1.0-2.0 Hz) oscillation
in these records. These oscillations are caused by the hull whipping after each slam
and are quite distinct from the longer period oscillations in the bending moment
experienced at the wave encounter frequencies. Alternatively the high frequency
vibrations may be detected using an accelerometer fixed to the ship's structure.
Slamming severity may also be monitored by analysing these records.
Pressure transducers are sometimes let into the hull surface to measure local
slamming impact pressures. It will of course be necessary to dock the ship if the
transducers are to be fitted below the waterline. This technique requires transducers
and a recording system capable of responding to the very short rise times (of the
order of milliseconds) typical of hydrodynamic impact phenomena. This will often be
incompatible with the requirements for the recording system for ship motions and a
separate system may be necessary.

15.5

RUN LENGTHS AND SHIP COURSES

We have seen in Chapter 4 that at least 100pairs of peaks and troughs are required in
an irregular time history in order to ensure a reasonably reliable estimate of the rms
motion. Each trial run must be of sufficient duration to achieve this minimum
standard. The actual length required may be estimated from strip theory calculations
of the motions in the wave spectrum expected during the trial. The mean period of
the peaks for each motion may then be calculated from equation (4.27). Fig. 15.4(a)
shows the results obtained for a trial planned in a frigate.
As expected, the mean periods are longest in following waves where the
encounter frequencies are low. The run time required to achieve 100 motion cycles is
given by

100TP
60

minutes

(15.6)

where TP is the mean period of the peaks for the chosen motion.
The required run time is given by the maximum value of T H for all the motions
and this is shown in Fig. 15 .4(b).
In practice longer runs than this absolute minimum are advisable. A certain
amount of additional time should be allowed for the ship to settle onto its new course
and speed at the beginning of each run: more importantly it should be realised that
every precaution should be taken to ensure that data of adequate quality are
collected. The opportunity of conducting a seakeeping trial occurs so rarely that it

283

Run lengths and ship courses

Sec. 15.5]

Ill 111!1111

!I!

ill

JW+--Leander
, I,

10

15

25

20

30

Elapsed time (minutes)

100
en
-o
c

90

(.)
Q)

en

80

co

70

Q)

+'

c
en
en
Q)
c

+'

Q)

$:
.::.:.
(.)
Q)

-o
Q)

Cl

co
.._
Q)

>

60
50

"

40

30
20
10

<(

12

14

16

18

20

Speed (knots)

Fig. 15.3- Measurements of deck wetness. (After Andrew and Lloyd (1981)). Reproduced by
permission of the Royal Institution of Na':~l Architects.

284

Seakeeping trials

en
"0

[Ch. 15

25
(a)

r:::

(.)

Q)

.!E.

"'

-"

Ctl
Q)

c.
r:::
0

'_;j

E
Q)

0
"0

.Q

Qj

c.

r:::

Ctl
Q)

:;E

180

0
Heading (degrees)
60
(b)

en
Q)

"5
r:::

.E
Q)

'_;j

r:::
::J

a:

180

0
Heading (degrees)

Fig. 15.4- Minimum run times for 100 motion peaks; frigate at 20 knots, T0

= 12.4 seconds.

would be false economy to shorten the runs because of economic or operational


pressures. If time is short it is better to reduce the number of runs rather than their
lengths.
Very long runs are undesirable because they may take the ship too far from the
wave buoy. The wave measurements will then be unrepresentative of the conditions
experienced by the ship during at least part of the run. Long runs also increase the
risk of the wave conditions changing during the run.
A good rule of thumb is to add a contingency of 10 minutes to the minimum
calculated run time for each course. Fig. 15.5 shows a sequence of courses for a trial

Run lengths and ship courses

Sec. 15.5]

fl

285

Predominant

V wave direction

Ol

c:

.2

"C

::c

LL

co
Q)

B Wave buoy

20 km

10 miles

Fig. 15.5- Typical sequence of runs for a 20 knot seakeeping trial.

planned on this basis in the frigate at 20 knots. The wave buoy is launched at the
beginning of the head sea run and the course sequence is chosen to minimise the
distance from the ship to the buoy.
c

16
Model testing

16.1 REASONS FOR MODEL SEAKEEPING EXPERIMENTS


Preparing a ship for a seakeeping trial is an expensive and time-consuming business.
It is usually necessary to select the trials period some time in advance and there can
be no guarantee that suitable weather conditions will occur. Many a trial has been
postponed or cancelled because there were no appreciable waves on the days
allocated for it!
Even if waves of suitable severity are experienced, uncertainties about the degree
of wave spreading may still limit the utility and general applicability of the results
obtained. However, the main disadvantage of full-scale trials is that they require the
ship to be built before they can be run. As such they are virtually useless as a method
of assessing the seakeeping qualities of a ship at the design stage.
Model testing provides an attractive alternative. Models are much less expensive
than ships and can often be entirely dedicated to the required experiments.
Moreover the model can be built in advance of the prototype ship and a number of
alternative designs can be tested. Indeed, before the advent of strip theory, model
testing provided the only method of assessing the seakeeping qualities of the ship at
the design stage.
If the model is to be tested in a towing tank or a seakeeping basin the waves can be
produced (and reproduced) to order. Suitable measurements are generally easier to
accomplish than at full scale. However, scaling problems can never be completely
overcome and it must be admitted that model tests in the controlled artificial
environment of the laboratory always lack something of the uncertain harsh reality of
the real world experienced by the ship.

16.2 MODEL EXPERIMENT SCALING


16.2.1 Dimensional analysis
Consider a model ship in a system of regular long crested waves. How should the test
conditions be scaled to ensure that the model's motions are an accurate reproduction

Sec. 16.2]

Model experiment scaling

287

of the motions which would have been experienced by the ship at full scale? To
answer this question we employ the techniques of dimensional analysis. These are
discussed in detail by Massey (1986) and we shall not give a general treatment here.
Suffice it to say that the technique allows the proper identification of the correct
model test conditions in terms of non dimensional groups of the quantities which are
relevant.
Let us consider as an example the heave motions of a ship and its model. For the
time being we may assume that we have no detailed knowledge of the physical
processes involved: even so we might surmise that the heave displacement amplitude
will be a function of the wave amplitude and frequency, the speed and heading and
the size, shape and inertias of the hull. In addition, the heave amplitude would be
expected to depend on the physical properties of the water (density and viscosity)
and the acceleration due to gravity. We might therefore write a general mathematical
expression relating these eleven quantities as:
(16.1)
where f 1 is some as yet undetermined function. [xB] represents a sufficient number of
coordinates to define the shape of the hull and [I] represents the moments of inertia
of the hull.
Massey (1986) shows how an expression of this form can be rearranged and
written in terms of a lesser number of non-dimensional parameters. Many different
formulations are possible and equally valid but it is COJ;lvenient to consider the form:

X3o

fz

{So

L'

00

/(L) U
[xB] [I] pUL}
\j g 'y(gL)' p., L' pL 5 ' P.w

(16.2)

The heave amplitude is now expressed in non-dimensional form by dividing by the


wave amplitude. This non-dimensional heave amplitude is given in terms of seven
non-dimensional 'groups' of parameters composed of the quantities which we have
previously decided are relevant. For convenience, the dimensions of these quantities
are listed in Table 16.1.
Equation (16.2) tells us that the non-dimensional heave amplitude will be the
same at both model and full scale provided that all the parameter groups on the righthand side of the equation have the same numerical values at model and full scale.
This requirement dictates the conditions required for the model experiment.
16.2.2 Fronde and Reynolds numbers
Before considering the consequences~ of this result we need to pay particular
attention to two of the non-dimensional groups in equation (16.2). The Froude
number is

(16.3)

[Ch. 16

Model testing

288

Table 16.1- Dimensions of quantities appearing in equation (16.2)


Quantity

Mass
(tonnes)

Length
(metres)

Time
(seconds

0
0
0
0
0
0
0
1
1
1
0

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

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

X3o

~0
(0

u
fJ-

[xB]
[I]
p
f.J-w

and the Reynolds number is

_ pUL
RN -

(16.4)

f.J-w

The significance of these two groups can be explained by considering a small


particle of fluid somewhere in the vicinity of the ship. Suppose that there is a
corresponding model-scale fluid particle at the corresponding location near to the
model and that each particle has dimensions which are some small fixed fraction of
the ship's or model's length. Then the mass of the particle in each case will be
proportional to pL 3 The particle's velocity v will be proportional to the velocity of
the ship or model and the distance ds it moves in a given time dt will be proportional
to the ship's length. The particle's acceleration can be written

dv

dv ds

dt

ds dt

dv

=vds

which is proportional to U 2/L. So the inertia force ( mass x acceleration) experienced by the particle will be proportional to pL 2 U 2 The gravity force on the particle
is simply (mass x g) which is proportional to pgL 3 So the ratio of the inertia and
gravity forces on the particle is

Referring to equation (16.3) we see that the Froude number is therefore

Model experiment scaling

Sec. 16.2]

p,

289

/( iner~ia force)
\j gravity force

The surface area of the fluid particle is proportional to L 2 and the viscous shear
stress is proportional to #1-w times the velocity gradient. The velocity gradient is
proportional to U/ L and the viscous force on the particle is therefore proportional to
11-wUL. So the ratio ofthe inertia and viscous forces on the particle is
pU2 L

11-wUL

pUL
11-w

Referring to equation (16.4) we see that the Reynolds number is


inertia force
viscous force
So the Froude and Reynolds numbers are measures of the relative importance of
the inertia, gravity and viscous forces acting on the flow around the ship and its
model.
16.2.3 Model scaling laws
We now examine the consequences of insisting that the seven non-dimensional
groups in the functionf2 in equation (16.2) are identical for model and ship. Let us
suppose that the model dimension ratio R is defined such that
I rJ

R=

(16.5)

where Lm and Ls are the lengths of the model and the ship.
Then for the dimensionless coordinates [x8 ]/L to be identical for model and ship
we require each corresponding dimension to be related by
Xsm

Xss

Lm

Ls

so that
~

Xss
Xsm

metres

(16.6)

In other words the model must be geometrically similar to the ship in all respects. The
underwater hull shape should be accurately reproduced and it is convenient to model

290

Model testing

[Ch. 16

the hull up to the weather deck. It is not usually necessary to represent the
superstructure as this has little effect on ship motions except possibly in very severe
conditions.
The requirement to represent the underwater hull form accurately demands that
the model's waterline be correctly located. This requires that the model's mass and
trim be correctly scaled. The model's mass is

(16.7)
where the integration is performed over the length of the hull. Now all the model's
dimensions must be scaled according to equation (16.6). So the model mass may be
written as

Pm ms
tonnes
Ps R3

(16.8a)

ms
R 3 tonnes

(16.8b)

if Pm = Ps So if the water densities are identical the model mass is reduced in


proportion to the cube of the dimension ratio. In practice it is usual to test models of
ocean-going ships in fresh water so that the densities are not identical and this should
be taken into account by using equation (16.8a).
The model's trim is determined by the longitudinal location of the centre of
gravity. This may be regarded as a hull dimension and must be located according to
equation (16.6).
The model's moments of inertia are scaled by ensuring that

and the model's moments of inertia are then given by

Im

= Pm Is

Ps Rs
I

tonne metres 2
.

Rs5 tonne metres2

(16.9a)
(16.9b)

if Pm = Ps So the model's moments of inertia must be reduced in proportion to the

Sec. 16.2]

Model experiment scaling

291

fifth power of the dimension ratio if the water densities are the same at model and full
scale. Again it is necessary to allow for differing water densities by using equation
(16.9a).
Taking the pitch moment of inertia as an example we may write
lssm
lsss

= mm k'lm tonne metres2

msk'ts tonne metres 2

where k 5 m and k 58 are the pitch radii of gyration at model and ship scale respectively.
Substituting equations (16.8a) and (16.9a) we find that
kss

metres

(16.10)

and similar expressions may be derived for the radii of gyration appropriate to the
other angular motions. So the radii of gyration are, like other model dimensions,
reduced in proportion to the dimension ratio.
16.2.4 Wave scaling laws
In order to make the non-dimensional wave amplitude '(, 0 / L the same for both model
and ship we require

= Sos
R

metres

(16.11)
1

rl
so that the wave amplitude, like other model distances, must be reduced in
proportion to the dimension ratio.
The model wave frequency is determined by the identity

and

rom

00 8

y'R radians/secondt

(16.12)

Model wave frequencies must therefore be increased in proportion to the square root
of the dimension ratio and are higher than the corresponding frequencies in ship
scale.
Since the wave length is given by
t It is assumed that model and full-scale values of g are the same.

292

Model testing
211'g
roz

metres

= As
R

metres

[Ch. 16

it follows that

(16.13)

and we see, as expected, that the model wave lengths are also reduced in proportion
to the the dimension ratio.
The wave period is
T

= 211' seconds
(f)

and it follows from equation (16.12) that model wave periods are reduced in
proportion to the square root of the dimension ratio:

Ts

yR

seconds

(16.14)

The wave number is

k =

211'
A

and the model and full-scale wave numbers are therefore related by
(16.15)
The model heading must, of course, be the same for both model and ship:
IJ-m

P-s

radians or degrees

16.2.5 Speed scaling laws


For model and full scale Froude numbers to be identical we require

or

(16.16)

Model experiment scaling

Sec. 16.2]

us
y'R

metres/second

293

(16.17)

so that the model speed must be reduced in proportion to the square root of the
dimension ratio.
For Reynolds number identity we require
Ps Us Ls
f.Lws

so that the model speed must be

Ps f.Lwm Us R
Pm f.Lws

UsR

metres/second

(16.18a)
(16.18b)

metres/second

if Pm = Ps and f.Lwm = f.Lws


So for identical water properties the model speed must be increased in proportion
to the dimension ratio. Likely differences in density and viscosity do not substantially
change this result.
Evidently equations (16.17) and (16.18) cannot simultaneously be satisfied
(except by testing at full scale so that R = 1). In other words it is impossible to satisfy
the requirement for Froude and Reynolds number identities at the same time. Model
tests in which the proper relationship between inertia, gra'fifr and viscous forces is
maintained are therefore impossible. t

Fortunately viscous forces do not play a very important role in ship motion
dynamics (except perhaps in the determination of roll damping). If this is the case,
the requirement to scale viscous forces in the correct proportion to inertia forces may
be waived and it is no longer essential to match model and full scale Reynolds
numbers. In any case this matching is not usually a practical proposition since it
would demand impossibly high model test speeds: for example, a ship speed of 30
knots at a model dimension ratio of 30 would demand a model test speed (equation
(16.18b)) of 900 knots! In contrast, Froude number identity requires reduced model
test speeds and this example would yield a much more practical model speed of about
5.5 knots (2.8 metres/second).
Neglect of Reynolds number can yield misleading results in certain specific
circumstances. If the Reynolds number is too iow (as is usual in ship model
experiments) the transition point will be too far aft and too much of the boundary
layer will be laminar. Flow separation is then more likely and the skin friction will be
too low. This may have some effect on the behaviour of the model. It is usual practice
to stimulate turbulence by roughening the model surface at the estimated transition
"

.,

t In principle it would be possible to achieve proper Reynolds number scaling if the model could be tested
in a fluid with decreased viscosity and/or increased density. No suitable fluid has yet been proposed.

294

Model testing

[Ch. 16

point to compensate for the neglect of Reynolds number scaling. This precaution
can, however, only be regarded as a palliative measure and it is always advisable to
adopt the largest practical model scale to reduce Reynolds number scaling problems
to a minimum.
16.2.6 Functional form of motion responses in regular waves
For a correctly scaled model equation (16.2) reduces to

(16.19)
where f 3 is an unknown function of the four listed non-dimensional quantities.
In practice the non-dimensional motion amplitude is often found to be essentially
independent of the non-dimensional wave amplitude '(, 0 / L provided that the wave
amplitude is moderate. In other words the motion amplitude is linearly dependent on
the wave amplitude, as assumed in strip theory. If this is the case we may write
equation (16.19) in the simple.r form:

(16.20a)
In passing, it should be noted that the non-dimensional frequency could be
expressed as a function of fhe non-dimensional wave length or the non-dimensional
encounter frequency. The functional relationships could then be written as

(16.20b)
or

(16.20c)

16.2. 7 Application of dimensional analysis to other seakeeping responses


The approach described above can be applied to all ship motions (linear and angular
displacements, velocities and accelerations) and equations of the same general form
as equations (16.19)-(16.20) will always be obtained. These other seakeeping
responses must be non-dimensionalised by dividing by appropriate combinations of
the relevant variables. Since the response amplitudes are generally proportional to
the wave amplitude it is convenient to include the wave amplitude in the denominator. For linear displacements like surge, sway, heave, relative and absolute motions
this is sufficient to give a non-dimensional quantity:

Sec. 16.2]

Model experiment scaling

295

For other seakeeping responses it will generally be necessary to include additional quantities to give the denominator the same dimensions as the numerator. For
example, angular displacement amplitudes are usually non-dimensionalised by
dividing by the wave slope amplitude to give quantities like

Linear acceleration amplitudes may be non-dimensionalised in the form

Scaling laws for seakeeping responses may easily be derived from these relationships. For example, if the non-dimensional heave velocity is expressed as .X30 /ro1;; 0 we
may infer that in a properly scaled model experiment

Using equations(16.11) and (16.12) we find that

showing that the heave velocity, like the forward speed, is reduced in proportion to
the square root of the dimension ratio. Similarly we may show that accelerations at
model and ship scale are identical:

and that model encounter frequencies are, like wave frequencies, increased in
proportion to the square root of the dim~nsion ratio. One important consequence of
this result is that the model's response to the waves appears, to the untutored eye, to
be too lively. A more realistic appearance can be obtained by recording the model's
motions on film or video tape and playing back at a reduced speed. The playback
speed should be reduced in prDportion to the square root of the dimension ratio.
Table 16.2 gives a comprehensive list of scaling f~ctors for model test conditions
and responses.

Model testing

296

[Ch. 16

Table 16.2 - Model scaling laws


Quantity

Examples

Mass
Length

Ship mass
Ship length; all dimensions; surge, sway, heave, absolute and
relative displacements; wave amplitudes and wave lengths

Time
Velocity

Wave and motion periods, run time; intervals between events

lt\/R

Ship speed; surge, sway, heave, absolute and relative motion


velocities; wave celerity and group velocity

1/"yR

Acceleration

Surge, sway, heave, absolute and relative accelerations; acceleration due to gravity

Angle

Roll, pitch, and yaw angles; heading, stabiliser and rudder


angles; phases

Angular
velocity

Roll, pitch and yaw velocities; heading, stabiliser and rudder


rates
Roll, pitch and yaw accelerations; heading, stabiliser and
rudder angle accelerations

Angular
acceleration
Pressure and
stress
Frequency

Multiply ship scale


value by

Slamming and wetness impact pressures; hydrostatic pressure;


dynamic pressure; stress
Wave and encounter frequencies; frequency of intermittent
events; propeller rpm

Force

Exciting force; shear force, tension; weight; thrust

Moment

Exciting moment; bending moment; torsional moment; torque

lm/(I,R3)
lm/(I,R4)

16.2.8 Tests in irregular waves


All of the foregoing analysis applies to seakeeping experiments in regular waves. An
analogous approach can be adopted for irregular waves but we must now deal in
statistical rather than in deterministic quantities.
Let us consider again the heave motion of a model and its full scale prototype.
Drawing a parallel with the regular wave equation (16.1) we might suppose that the
rms heave motion for a given wave spectrum formulation would have some
functional relationship witb a number of relevant quantities as follows:
(16.21)
Applying the same general approach we find that the non-dimensional rms heave for
a geometrically scaled model is conveniently expressed as

(16.22)
In other words the non-dimensional rms heave is a function of the nondimensional significant wave height and modal period, the Froude number and the
heading. It is easily seen that the model significant wave height must therefore be

Sec. 16.4]

Laboratory test facilities

297

scaled in proportion to the dimension ratio, and the modal period must be reduced in
proportion to the square root of the dimension ratio. For moderate significant wave
heights the functional dependence on the non-dimensional significant wave height is
often weak and the rms heave motion for a given modal wave period, Froude number
and heading is then directly proportional to the significant wave height.

16.3 OPEN WATER MODEL EXPERIMENTS


The simplest kind of model experiment involves testing an instrumented remote
controlled model in the open sea in what amounts to a miniature seakeeping trial. A
dimension ratio of the order of 10 is typical and the model may be powered by a small
marine diesel engine or an electric motor. Ship motion instrumentation similar to
that employed in full-scale trials is used and motor speed setting and steering are
achieved by radio control.
A wave buoy is required to measure the waves. This must be designed to respond
to the short waves which will be of interest in the model experiment and a standard
full-scale wave buoy will not usually be adequate in this respect. Wave and motion
spectra and statistics are obtained in exactly the same way as at full scale.
The required full-scale wave spectrum must be reproduced at model scale
according to the appropriate scaling laws. This generally dictates a test area in
sheltered water with a limited fetch. Unfortunately this often results in multiple wave
reflections from the nearby coast and a high degree of directional spreading is often
present during these experiments. This makes the results difficult to interpret and
misleading conclusions can easily be drawn. Nevertheless, such experiments do have
an intrinsic appeal for their apparent realism and they are certainly much cheaper
than full-scale trials.

16.4 LABORATORY TEST FACILITIES


Tests in the controlled environment of the indoor laboratory are much preferred to
open water experiments. The traditional type of long narrow towing tank is
illustrated in Fig. 16.1. Such tanks are usually of the order of 100 metres long and
5-10 metres wide. The depth should be at least half the longest wave length
envisaged to avoid unwanted shallow water effects on the waves. A towing carriage
runs on rails and is powered either by onboard electric motors or hauled by a winch at
the end of the tank. Waves are generated at one end of the tank by a wave maker and
absorbed by a beach at the other. A section of the beach can usually be lowered to
allow the end of the tank to be used as a docking area for ballasting and trimming
mod~ls. Models may be tested at forward speed in head or following waves. In
addition, tests at zero speed in beam wayes (with the model moored across the tank)
are possible.
The model to be tested is mounted under the carriage using one of the
arrangements shown in Fig. 16.2; The rigidly restrained arrangement shown in Fig.
16.2(a) is used only in specialised experiments to measure wave loads: Fig. 16.2(b)
shows the rig commonly used with an unpowered model. The model is free to pitch
around a hinge pin at its centre of gravity. The pitch pivot is mounted at the end of a

298

Model testing

[Ch. 16

Fig. 16.1- Typical towing tank.

vertical rod which slides in linear bearings, allowing the model freedom to heave. No
surge motion is allowed.
A limited freedom to surge can be obtained using the sprung arrangement shown
in Fig. 16.2(c). True surge motions can only be obtained ifthe model is self-propelled
and the arrangements shown in Figs 16.2(d) and 16.2(e) are possible.
It is of course necessary to use very high quality bearings in these rigs to reduce
frictional effects on the measured motions to an absolute minimum. The weight of
the rig is supported by the model and due allowance for this must be made when
ballasting and trimming. Frictional effects may be virtually eliminated by dispensing
with any guidance arrangements as shown in Fig. 16.2(e).
In recent years the seakeeping basin, specifically designed for seakeeping model
tests, has been introduced. These tanks are usually of the order of 50 metres square
and are fitted with wave makers and a beach at opposite ends. An ideal arrangement
is shown in Fig. 16.3. A main carriage spans the tank and runs on rails in much the
same way as on the traditional towing tank. A subcarriage is mounted on the main
carriage so that it may be positioned at any point over the water surface. During a
self-propelled experiment the subcarriage's position may be maintained over the
model by an automatic control system. Alternatively for towed experiments the
subcarriage may be driven across the tank at some predetermined heading to the
waves.
For self-propelled experiments the model is connected to the subcarriage by an
umbilical cable as shown in Fig. 16.2(e). This cable supplies the model with electrical
power for its propulsion and instrumentation and also serves to feed the model's
response signals back to the carriage for recording and analysis. The cable is

299

Laboratory test facilities

Sec. 16.4]

t
----~r

II

II

--C:_________J/---(a) Rigidly restrained

h..._____

--C::---~----J/-(b) Free to heave and pitch


(unpowered model)

-~
(c) Free to surge, heave and
pitch (unpowered model
restrained by springs)

--~~------k====/-~--(e) Model restrained only


by umbilical cable
(powered model)

(d) Free to surge, heave and


pitch (powered model)

-~~~------~1--(f) Self-contained model


(no carriage)

Fig. 16.2- Model restraint systems.

supposed to be sufficiently light and flexible to preclude any interference with the
model's motions.
This ideal arrangement is of course very expensive and many tanks have no
carriage, relying instead on free running self-contained models as shown in Fig.
16.2(f). This means that the model mus~ carry its own batteries for power supply and
that its responses must be recorded on board or telemetred ashore. In either case the
additional weight makes the achievement of proper mass and inertia scaling more
difficult and large models may be necessary to allow adequate freedom of ballast
adjustment.

300

[Ch. 16

Model testing
Beach

M'
/
am carnage

",~)

,~" 4

Subcaiiage

Wave maker

Fig. 16.3- Typical seakeeping basin.

16.5 WAVE MAKERS AND BEACHES


16.5.1 Wavemakers
Fig. 16.4 shows several different designs for laboratory wave makers in current use.
In modern installations the wave maker is usually driven by a servo-contolled
hydraulic ram which will follow an electrical input drive signal. So both regular and
irregular waves can be reproduced provided that appropriate drive signals are
available.
Fig. 16.5 shows a block diagram representing the wave maker servo and the wave
maker itself. The wave maker transfer function relating its motion amplitude to the
wave amplitude it produces may in principle be calculated and Crapper (1984) gives a
theory for a simple piston wave machine. However, most laboratory test tanks rely
on an experimentally determined transfer function for the wave maker and its servo.
Fig. 16.6 shows a typical calibration for a wedge-type wave maker. The diagram
allows the voltage amplitude required to achieve any desired regular wave amplitude
to be determined.
16.5.2 Beaches
The beach performs the important function of absorbing the waves after they have
travelled the length of the tank. In practice all beaches allow a certain amount of
wave energy to be reflected and the results of tests by Hsiung et at. (1983) on beaches
of various slopes are shown in Fig. 16.7. The graphs show the reflection coefficient
defined as the ratio

301

Wave makers and beaches

Sec. 16.5]

-(a) Simple flap

(b) Simple wedge

(c) Double-angle
wedge

,,

I)

II
II

If

ff

(e) Double flap- flaps out of phase

(d) Double flap- flaps in phase

(f) Pneumatic

Fig. 16.4- Wave ma~ers.

Drive signal

Wave maker
servo

Wave maker

Wave maker
motion

'

Fig. 16.5- Wave maker block diagram.

Waves

302

Model testing

[Ch. 16

Wave length A (metres)

Wave frequency w (radians/second)

Fig. 16.6- Typical wave maker transfer function.

'Aid

Beach slope

c
Q)
c::;

i:Q)
0

"c0

u
Q)

;;::
Q)

cr

1.8

2.0

"'\/(dig)

Fig. 16.7- Typical beach reflection characteristics. (After Hsiung et al. (1983).)

amplitude of reflected wave


amplitide of incident wave

--~-----------------

plotted as a function of the non-dimensional wave frequency ro y(d/g) (where dis the
depth of the tank). The results show that the beach is most effective at high

Sec. 16.6]

Instrumentation

303

frequencies (i.e. in short waves) and the best results are obtained when the beach
slope is very small. Such a shallow beach may well occupy a significant proportion of
the length of the tank and practical considerations may place a limit on the beach
slope which can actually be used.
Over the range of wavelengths of interest to most model experiments the
reflection coefficient is usually of the order of 0.05-0.10. These reflections will mix
with the incident waves and spoil their characteristics. It is therefore important to
ensure that measurements are taken before these unwanted reflections reach the
model. This is discussed in more detail in Section 16.9.

16.6 INSTRUMENTATION
Where a carriage is available this provides a convenient datum for the measurement
of ship motions. Some examples of commonly used techniques for pitch and heave
are shown in Fig. 16.8 and adaptations of these systems are used for the other
2x,

-b:::::..____----J/=(a) String and potentiometer to


measure absolute motions

(c) Sonic transducer to measure heave;


gyro to measure pitch

(b) St,ring and pote~ti~meter to measure heave;


potentiometer to measure pitch

(d) Accelerometer to ~easure heave acceleration;


Gyro to measure p1tch; results
telemetered ashore or recorded on board

Fig. 16.8- Instrumentation for pitch and heave.

motions. A simple arrangement with strings and potentiometers is shown in Fig.


16.8(a). The potentiometers give signals which are proportional to the absolute
motions at the attachment points and these may be combined to give estimates of the
pitch and heave:

Model testing

304

x3

Xs

St+Sa

2
Sa -sf

[Ch. 16

metres

(16.23)

radians

(16.24)

Xr

where St and sa are the absolute motions measured forward and aft and 2xr is the
longitudinal separation of the two measurement locations.
Another technique is illustrated in Fig. 16.8(b). Here the heave is measured
directly by monitoring the motion of the heave post and the pitch is obtained by
coupling a potentiometer to the pitch pivot pin in the model.
These arrangements are suitable for the towing and guidance arrangements
shown in Figs, 16.2(b)-16.2(d). Where there is no physical connection {apart from
the umbilical cable) between the model and the carriage, systems like that shown in
Fig. 16.8(c) have found favour. Here the pitch is measured by a gyro of the type used
in aircraft navigation systems and the heave is monitored using a sonic transducer.
The transducer emits a short-duration pulse of high-frequency sound (above the limit
of human perception) and this is reflected from a horizontal board mounted under
the carriage. The time required for the sound to travel from the transducer to the
board and back again is monitored and is proportional to the distance from the
transducer to the board.
Where there is no carriage, heave and other linear ship motions cannot be
measured directly because no convenient datum level is available for use as a
reference. It is then necessary to resort to accelerometers as in full scale ship trials
and this is illustrated in Fig. 16.8(d).
Two kinds of wave transducers are shown in Fig. 16.9. Both involve a pair of
metallic surface piercing elements. Electronic circuits are used to monitor the
resistance of the water between the two elements and this is a function of the depth of
immersion. Alternatively the elements may be regarded as the plates of a capacitor
using the water as the dielectric medium. The capacitance is then monitored to
provide an analogue of the depth of immersion.
These wave probes have some disadvantages. Surface tension effects may cause
the water level experienced on the probe surface to be slightly different to the true
level away from the immediate vicinity of the probe. The errors due to this effect are
not, however, very serious unless the probes are used to measure very small waves.
Much more significant effects are experienced if these surface-piercing probes are
used on a moving carriage to measure the waves encountered by the model. The
probe inevitably causes some surfaye disturbance and this is likely to introduce errors
in the measured wave profile dueio the probe's own 'bow' wave. Speeds in excess of
1-2 metres/second may introduce noticeable errors.
More sophisticated wave probes which avoid contact with the water surface have
been developed. One type uses transducers to detect an ultrasonic pulse reflected
from the water surface. Optical systems using lasers are also being considered and
the 'servo needle' is being developed in Japan. This uses a servo-controlled probe
which is continually adjusted so that it is just in contact with the water surface.

Instrumentation

Sec. 16.6]

(a) Foils

305

(b) Wires flush with


probe surface

Fig. 16.9- Wave probes.

It is advisable to position the wave probe perhaps qne metre ahead and one metre
to one side of the model to avoid measuring any surface disturbance caused by the
model. Care should be taken to ensure that any surface disturbance due to the probe
does not interfere with the model.
Fig. 16.10 shows some of the types of instrumentation ti~d to measure relative
bow motion. Most of the transducers are developments of those used to monitor
waves. The simplest form, using a pair of foils ahead of the m<;>del, is shown in Fig.
16.10(a). This gives a general indication of the relative motion ahead of the bow but
is positioned so that the disturbance due to the proximity of the hull may not be
measured.
This problem is overcome with the arrangement of flush mounted tapes shown in
Fig. 16.10(b). Here an aluminium foil tape is fixed directly to the surface ofthe model
using double-sided adhesive tape. Tl;J.e aluminium is insulated from the water by a
layer of 'Teflon' tape. The aluminium and the w~ter form the plates of a capacitor,
the Teflon tape being the dielectric. The electrical circuit of the capacitor is
completed through an uninsulated aluminium tape on the surface of the hull, and this
allows the capacitance, and hence the relative motion, to be monitored by suitable
circuits. Insulated tapes may be located at a number of stations, allowing the
measurement of the longitudinal variation of relative motion. A single return tape
will suffice for several measuremennocations.
Measurements with this arrangement will certainly include the effects of the
disturbance due to the hull Q,ut may suff<:<r from the surface tension effects experienced by wave probes. The rig shown in Fig. 16.10(c) has been proposed as an

Model testing

306

[Ch. 16
Aluminium foil tape

Teflon tape

(a) Capacitance strips


ahead of model

(b) Flush capacitance


strips

(c) Resistance wires

Fig. 16.10- Relative motion instrumentation.

alternative. Here resistance wires are stretched taut from keel to deck. The wires do
not touch the hull surface so they should not be subject to the unwelcome effects of
surface tension: at the same time they should be sufficiently close to the hull to give
measurements which include the local wave disturbance effects due to the hull.
Experiments to investigate deck wetness and slamming are also of some importance and suitable transducers to monitor deck wetness are shown in Fig. 16.11. In

(a) Pressure cells

(b) Resistance probes

Fig. 16.11- Deck wetness instrumentation.

Sec. 16.7]

Model materials

307

Fig. 16.11(a) a vertical plate incorporating pressure sensitive cells is mounted on the
forecastle and may be used to measure both the impact pressures and the frequency
of deck wetness. The cells are covered with a thin flexible diaphragm and are
connected to pressure transducers mounted on the carriage by lengths of flexible
tubing. An alternative technique using resistance wire probes is shown in Fig.
16.1l(b). Here the probes measure the depth of water on the deck but can again bt;
used to monitor the frequency of deck wetness.
In slamming experiments pressure transducers are usually mounted flush with the
model keel and under the bow flare to measure impact pressures directly.

16.7 MODEL MATERIALS

William Froude developed the technique of building models from paraffin wax and
this is used in some establishments to this day. This somewhat unlikely material has
many advantages. It is easily worked and models can be modified at any time using
simple hand tools. After a model's useful life has expired it can be melted down and
the wax reused. However, wax models are not very robust: indeed they will gradually
distort if left unsupported over a prolonged period. This can be avoided by keeping
them submerged if they are required for future experiments. This is, however, hardly
convenient since all the internal equipment must be removed. A harder wearing
material is usually chosen for seakeeping model experiments.
Wood is a favourite material, being easily worked and durable. The usual
technique is to cut out a series of boards to the shapes of the hull waterlines and to
assemble these in the so called 'bread and butter' construction. The excess material is
removed by hand using templates for guidance in the final stages. The inside of the
model is carved away to allow room for the required equipment and to reduce
' il
weight.
It is always necessary to allow a considerable wall thickness to ensure adequate
strength in a wooden model and the usable internal space is often restricted. Wooden
models are often quite heavy and only a small amount of additional ballast is needed
to bring them up to the required mass. There is therefore little freedom to position
the ballast to obtain the proper inertias and centre of gravity. Wood also has the
disadvantage that it always seems to absorb moisture from the water, no matter how
well it is waterproofed. So dimensional stability is difficult to maintain as the wood
swells and subsequently contracts as>it dries out.
For this reason alternative modern materials are often used, particularly for free
running models which must carry their own batteries for power supply. A favourite
material is glass reinforced plastic (GRP). The shell of the hull need then be only a
few millimetres thick and a very light yet strong and stiff model can be produced. It is
first neccessary to build a plug (male) mould to the exact finished shape of the hull
and this is conveniently done in paraffin wax. The plug is coated with a release agent
and the female mould is built up on the plug with layers of glass cloth impregnated
with resin. When the resin has cured, the two moulds are separated and the plug is
discarded. The inside of the female mould is then coated with the release agent and
layers of resin-impregnated cloth are built up to form the finished GRP hull. After
curing, the hull and the mould are separated and the hull is ready for fitting out. If

308

Model testing

[Ch. 16

required the mould can be used again to reproduce any number of exactly identical
models.
Expanded polystyrene foam has also been used in some laboratories. This is very
light and stiff and has many of the advantages of glass reinforced plastic.
16.8

TRIMMING AND BALLASTING

The first step in trimming and ballasting a model is to weigh the hull, complete with
all internal fittings such as instrumentation, batteries, propulsion motors, etc. A
weight to represent the weight of any towing or restraint apparatus to be supported
by the model should be included. The additional ballast required to bring the model's
mass up to the required value may then be calculated and the necessary ballast
weights stowed in the hull. The model may then be placed in the water and the
positions of the ballast weights adjusted until the required trim is obtained. The
model is usually required to have no heel angle and this can be checked by a spirit
level sited on some suitable datum surface. The longitudinal trim is best determined
by simple adjustable trim gauges of the type illstrated in Fig. 16.12. These allow the

Fig. 16.12- Trim gauge.

freeboard to be determined at specified locations forward and aft. If the model's trim
is correct it follows that the longitudinal position of the centre of gravity must be
correctly located. It remains to determine the vertical location of the centre of gravity
and to adjust it if necessary. This is done by means of the inclining experiment. A
small measured heeling moment is applied to the model and the resulting heel angle
is measured. The moment is most easily applied by moving a known weight a
measured lateral distance. This enables the solid metacentric height GM5 to be
determined (see equation (10.8)). The position of the metacentre M will be known
from the ship's hydrostatic diagrams and this allows an estimate of the vertical
location of the centre of gravity. The VCG can be adjusted by appropriate vertical
movement of ballast weights.

Sec. 16.8]

Trimming and ballasting

309

The moments of inertia are measured by various adaptations of the compound


pendulum illustrated in Figs 16.13 and 16.14. Figure 16.13(a) shows a simple

(a) Pitch

(b) Roll

Fig .. 16.13- Compound pendulum rigs to determine pitch and roll radii of gyration.

lrJ
technique for measuring the pitch moment of inertia Iss The model is suspended in a
light frame so that the centre of gravity ish metres below a pivot point. The entire rig
is then oscillated by hand and the natural period of oscillation T. determined by
measuring the time required for, say, ten complete oscillations. The total moment of
inertia of the complete rig is, by the parallel axis theorem,
tonne metres 2

(16.25)

and the stiffness of the compound pendulum is


c = ( mm h + mp hp)

g kN~metres/radian

where
mp is the mass of the suppot.ting frame in tonnes
lp is the mass moment of inertia of the supportingJrame in tonne metres2

(16.26)

Model testing

310

[Ch. 16

Fig. 16.14- Bifilar suspension rig for estimating yaw radius of gyration.

hp is the distance from the centre of gravity of the supporting frame to the pivot
point in metres.
'
Now, from equations (6.8) and (6.21), the oscillation frequency is approximately

ro. =

~ = ~ ( ~)

radians/second

(16.27)

and the model's radius of gyration is given by

(16.28)
An exactly similar procedure is used for finding the roll radius of gyration as
shown in Fig. 16.13(b). In both cases it is desirable to keep the mass and inertia of the
supporting frame as small as possible to minimise errors in the estimation of the hull's
characteristics. The inertias of the frame may be found by measuring its natural
period of oscillation without the hull attached.
A somewhat simpler procedure for finding the yaw radius of gyration without
using a supporting frame is shown in Fig. 16.14. The model is suspended on two wires

Trimming and ballasting

Sec. 16.8]

311

from a suitable overhead beam. Typically the wires will be five or six metres in
length. The model is oscillated in yaw, taking care to avoid roll or sway motions. The
natural period of oscillation is recorded as before. The stiffness of the system is
calculated as follows.
Suppose that the model is yawed through a small angle x 6 radians as shown. Then
the wires will swing through a small angle

radians

Now the tension in each of the two wires must be half the model weight:

mmg kN
2
and the horizontal component of these forces tending to swing the model back to its
equilibrium position is approximately

kN

So the restoring moment on the model is

kN metres

and the stiffness of the system is then

mmgx;
h

(16.29)

kN metres/radian

The moment of inertia of the wires supporting the model is negligible and tpe natural
frequency of the system is, from equations (6.8) a_nd (6.21),

ro. =

~ = ~ ( m: k~) = ~ ( ~) Z:

radians/ second

(16.30)

and the yaw radius of gyration is given by

(16.31)

Model testing

312

[Ch. 16

The bifilar suspension method is widely used to estimate the yaw radius of
gyration because of its simplicity and convenience, requiring no more apparatus than
a stopwatch and a pair of wires suspended from hooks in a suitable overhead beam.
In principle the same method could also be used to estimate the pitch radius of
gyration with the model turned on its side. This is, not usually practical, however,
because much of the internal equipment is not sufficiently well secured. Instead it is
often assumed that the pitch radius of gyration is the same as the yaw radius of
gyration.
The model's radii of gyration may be adjusted to the required values by moving
the internal ballast weights. The radii of gyration may be reduced by moving ballast
towards the middle of the model and vice versa. Care should be taken to ensure that
any adjustment to ballast on one side (or end) of the model is exactly balanced by a
corresponding adjustment at the other. Otherwise the location of the model's centre
of gravity will be changed. It is good practice to check the centre of gravity position
after swinging the model to ensure that all is well in this respect.

16.9

TESTING IN REGULAR WAVES

16.9.1 Measurement of motion transfer functions


Most tests in regular waves are concerned with the experimental determination of
the motion transfer functions. It is therefore necessary to record the sinusoidal
motions of the model and to determine the motion amplitudes experienced for a
variety of different wave frequencies. It is usual to keep the wave slope constant
while varying the wave length, but experiments in waves of constant amplitude are
also used. Care should be taken to ensure that the wave steepness is always small
(unless the tests are specifically intended to investigate non-linear effects) and that
there is no risk of the waves breaking.
If phase information is required it is necessary to measure the incident waves
using a wave probe mounted on the towing carriage. It is usual to position the probe
ahead and to one side of the model, as shown in Fig. 16.15, to avoid measuring the
wave distortion caused by the presence of the model. This introduces a phase shift in
the recorded motions and it is necessary to correct for this effect in the analysis.
Suppose that tile wave probe is positioned x lp metres forward of the mean position of
the modej's centre of gravity and x 2 P metres to starboard. Then the probe will record
the waves. xP metres after they have passed the centre of gravity. The distance xP is
given by
xP

x 1P cos IL- x 2 P sin IL

metres

Now the waves are overtaking the model with a relative velocity

c - U cos

IL

metres/second

and a wave trough recorded at the probe would have been alongside the model's
centre of gravity at a time

Testing in regular waves

Sec. 16.9]

313

Fig. 16.15- Calculation of phase shift due to probe location.

c- U cos

p,

seconds

earlier. So the phase lead measured with reference to the waves recorded at the wave
probe should be reduced by an amount

op

(l)e

tp

roe (x 1P cos p,.-x2 P sin p,)


c- U cos

p,

radians

(16.32a)

[Ch. 16

Model testing

314

21T (
)
T
x 1P cos 11-- x 2 P sm 11-

k (x 1P cos 11-- Xzp sin !J-)


0)2

(x 1P cos 11-- x 2 P sin ~J-)

radians

(16.32b)
(16.32c)

radians

(16.32d)

radians

16.9.2 Effects of wave reftections


It was explained in Section 16.5 that all beaches reflect a certain amount of wave

energy and that these reflections will eventually spoil the characteristics of the waves
generated by the wave maker. Model experiments run in these contaminated waves
will give misleading results. This problem can be avoided by careful attention to the
timing of the experiment run in relation to the time at which the wave maker is
started.
Consider the model in the experiment tank shown in Fig. 16.16. Now it was

Beach
1--

------------r---------Reflected wave

disturbance

UG

Wave front

tyu
/7

XT

(/

-L.....-

Wave maker
Fig. 16.16- Wave fronts and model location.

Testing in regular waves

Sec. 16.9)

315

shown in Chapter 3 that the main body of waves having the proper wave amplitude
propagates down the tank at the group velocity. This is preceded by a secondary
wave front of reduced amplitude travelling at the wave celerity. The experiment run
(i.e. that part of the model's run in which measurements of its behaviour are
recorded) must obviously be confined to the part of the tank which contains waves of
the proper amplitude. In other words the model's responses should only be observed
and recorded while the model is behind the advancing wave frontWW. At the same
time it is necessary to avoid taking measurements of the model's responses after it has
encountered the initial wave disturbance RR reflected from the beach.
This may be analysed with the aid of the distance/time diagrams shown in Fig.
16.17. In these diagrams xT is the distance in metres from the wave maker and tis the

(/)

Q)

t (seconds)

t (seconds)

Fig. 16.17- Optimum model experiment runs.

time in seconds. The wave maker is started at t = 0 seconds and the model progresses
'down' the tank at a component velocity U cos IL metres/second.
The path of the initial wave .disturbance is represented by the line 0 A with slope c
metres/second. This initial disturbance will be reflected from the beach at time

Model testing

316

tL

= -LT
c

seconds

[Ch. 16

(16.33)

(where LT is the effective length of the tank) and will arrive back at the wave maker
at time

2tL

seconds

(16.34)

The path of the 'proper' wave front is represented by the line OB with slope ua
metres/second. This reflects from the beach at time
t

2tL

seconds

(16.35)

(since the group velocity is half the wave celerity in deep water).
When the model is at a location represented by a point above the line OB it will
not be experiencing the proper wave amplitudes of the main body of waves. Similarly
a model located at a position represented by a point above the line AC will be
experiencing the unwanted reflected wave disturbance. It follows that the conditions
outlined above for valid model test observations in regular waves are only experienced when the model's location is defined by points within the triangle ODC.
Let us first suppose that the model is to be tested in following or quartering waves
so that
U cos f.t

> 0 metres/second

Ignoring the short distances needed to accelerate the model up to the required test
speed and to bring it to rest at the end of the run, the model's progress down the tank
can be represented as a line with a positive slope equal to U cos f.t metres/second. The
line may be positioned anywhere in the diagram depending on the time chosen to
start the experiment run, but the optimum locations, depending on whether U cos f.t
is greater or less than the group velocity, are as defined by the broken lines in Fig.
16.17(a).
If U cos f.t > ua the optimum track (giving the longest possible run) is shown by
the line ED and the run 'length' is given by

(16.36)

This is achieved by starting the experiment observations at time

t = -2LT (2
-3
c

u cos f.t

seconds

(16.37)

Testing in regular waves

Sec. 16.9]

317

and the run time is given by

u cos f.L

3U cos f.L

seconds

(16.38)

If U cos f.L < u 0 the optimum track is shown by the line OF and the run 'length' is
given by

2U cos f.L
u cos f.L + c

(16.39)

and this is achieved by starting the model as soon as the wave maker is started at time
t = 0 seconds. The run time is then given by

TH

2LT
u cos f.L + c seconds

(16.40)

The case of head and bow wave experiments is shown in Fig. 16.17(b). U cos f.L is
now negative and the model's track is represented by lines with a negative slope: If
the model speed and heading are such that U cos f.L < .- c, the optimum model track
is shown by the line DG and the run 'length' is given by
2
3

(16.41)

This is achieved by starting the model run at a point LT/3 metres from the beach at
time

4tL

(16.42)

seconds

giving a run time

3U cos f.L

seconds

(16.43)

For the case 0 > U cos f.L > - c the optimum m.odel track is shown by the line HC
and the run 'length' is now giwen by

2U cos f.L

(16.44)

318

[Ch. 16

Model testing

This is achieved by starting the model run at a point


cLT
c- 2U cos 1.L

metres

from the beach at a time


t

4LTU cos /.L


3 (2U cos /.L- c)

seconds

(16.45)

The run time is then given by


TH

2LT
= ---:-::-:--=--c- 2U cos /.L

seconds

(16.46)

These results are summarised in Fig. 16.18. They show that even in the most
1.0

Beach

End of run

Start of run
f-

::::'f- 0.5
)(

Start of run
Wave maker

-3

-2

-1

(U cos !l)lc

Head and .,.


.....__ bow waves

Follqwing and
quartering waves

Fig. 16.18- Optimum model experiment runs.

favourable conditions no more than two thirds of the tank length is usable. In
practice slightly greater lengths of the tank may be used if the beach is very effective.
16.10

TESTING IN IRREGULAR WAVES

Tests in irregular waves call for a slightly more involved procedure. It is first
necessary to scale the required wave energy spectral ordinates and frequencies using
the scaling laws listed in Table 16.2. A typical result for a Bretschneider spectrum

Testing in irregular waves

Sec. 16.10]

319

with significant wave height of 5.5 metres and modal period of 12.4 seconds for a
dimension ratio of 36 is shown in Fig. 16.19(c). We require to drive the wave maker
150
(a)

u
en
=o

Wave maker drive


signal spectrum

QJ

::'

~
0

2.

50

>
C/)
8

0.030
(b)

~
~

QJ

E
Wave maker
transfer function

1~1

0.03
(c)

u
en
=o

QJ

Wave energy
spectrum

0.02

::'

0.01

JJ
0
Model wave frequency w (radians/second)

Fig. 16.19- Generation of irregular wave spectrum.

320

Model testing

[Ch. 16

with an irregular electrical signal which will produce a train of irregular waves having
this wave spectrum. The spectrum required for the input signal may be deduced from
a knowledge of the wave maker transfer function. Using the spectral calculation
procedures described in Chapter 14 we may write the wave energy spectrum as

Sr, (ro)

Su(ro)

(~:r

metred(radian/second)

(16.47)

from which we may derive the drive signal spectrum

Sr,(ro)

volts 2/(radian/second)

(16.48)

1;,0 )
( Vo

A suitable time history having the spectrum given by equation (16.48) must then
be constructed using the wave synthesis techniques described in Chapter 4. Driving
the wave maker with this signal would then be expected to produce a wave time
history with the desired wave energy spectrum. In practice this simple technique may
not give results of adequate accuracy. This is believed to be because the wave maker
response suffers from poorly understood interactions between the many frequencies
present in the irregular waves being generated. These interactions are absent when
the wave maker is used to generate a single-frequency regular wave. So the regular
wave transfer function can only be regarded as a first approximation to that required
to quantify the response in irregular waves. Moreover, the required transfer function
apparently depends on the particular time history being generated and not just on the
spectrum characteristics. So a different transfer function is required for every new
time history.
These difficulties can be overcome by empirical adjustments t<;> the wave maker
drive signal spectrum. Where the measured wave spectral ordinates are too low the
drive signal spectral ordinates should be increased and vice versa. It is usually
possible to achieve a good match to the desired wave spectrum with two or three
adjustments of this kind.
Each component of the system of irregular waves will propagate down the tank at
its own group velocity, preceded by an advance party of reduced amplitude waves of
that frequency moving at the appropriate wave celerity. So the lowest frequency
component, which has the highest group velocity and wave celerity, will overtake the
rest of the waves and arrive at the end of the tank well before the other frequency
components. Clearly the 'complete' wave spectrum will not be experienced at a given
location in the tank until the highest frequency component has arrived at that point.
By this time the lowest frequency waves may well have been reflected from the beach
and already be spoiling the waves propagating down the tank in the proper direction.
This can be avoided by introducing the frequency components to the wave maker

321

Testing in irregular waves

Sec. 16.10]

drive signal (and hence to the generated waves) in descending order. The highest
frequency is introduced first and subsequent components are included at times
specified to ensure that all frequency components arrive simultaneously at some
specified point in the tank.
Suppose that the required wave spectrum contains N frequency components

ro1 - ron - roN radians/second

The wave component with the highest frequency roN will propagate down the tank
with the group velocity

uaN

2roN

(16.49)

metres/second

and arrive at a point xT metres from the wave maker at time

(16.50)

seconds

arriv~~~t

The nth wave component will take xT!uan seconds to


this point and it is
therefore necessary to delay its introduction into the wave tim~ history until time

(16.51)

Obviously the pointxT at which the waves are required to coalesce, should be chosen
to maximise the length of the tank available for the experiment. Reference to Fig.
16.18 shows that for a given speed and heading the most critical conditions occur
when the wave celerity is highest. ln other words the available test length is small
when the wave frequency is lo.w. So the lowest frequency in the wave spectrum will
dictate the location of the coalescence point and this will also determine the
permissible run time. This is illustrated in the following worked example.

Model testing

322

[Ch. 16

Consider a model to be tested in head waves in a ship tank with a usable length
LT = 100 metres. The model's dimension ratio is R = 36 and the model test speed is to
represent a ship speed of 20 knots. The wave spectrum (at full scale) includes
frequencies in the range 0.3-1.6 radians/second and this is to be represented by
discrete frequency components at intervals of 0.1 radians/second. t
It is required to find the optimum run time and start position for the test and the
frequency component time lags for the wave maker drive signal.
It is convenient to work in ship scale: the tank length becomes

LT

3600 metres

and the ship component velocity is

U cos 11-

= -

10.3 metres/second

The lowest frequency component (ro 1 = 0.3 radians/second) will have the greatest
celerity:

32.7 metres/second

Now 0 > U cos 11- > - c 1 and equations (16.44)-(16.46) apply. So the start
position is given by

Xr

2LTU cos Jl2U cos 11-- c1


1391

metres

or 38.7 metres

at ship scale
at model scale

The run time is


t In practice a much smaller frequency interval would be used to produce a better approximation to the
continuous wave energy spectrum.

Tank wall interference

Sec. 16.11]

323

c 1 - 2U cos J.t
135

seconds

or 22.5

at ship scale

seconds

at model scalet

The wave maker drive signal lags are, from equation (16.51):

Frequency
(radians/second)
(ship scale)

Lag
(seconds)
(model scale)

0.3
0.4
0.5
0.6
0.7

10.3

9.5
8.7

7.9
7.1
6.3
5.5
4.7
3.9
3.2
2.4
1.6

0.8
0.9
1.0
1.1
1.2
1.3
1.4

o.8

1.5
1.6

16.11

rl

0.0

TANK WALL INTERFERENCE

An oscillating model acts like a wave maker and radiates waves on either side.
Indeed this is the mechanism responsible for dissipating energy and providing
motion damping. The model will of course oscill;te at the encounter frequency and
the generated waves will radiate away from the model at the celerity appropriate to
this frequency. The waves will eventually reach the tank walls and be reflected back
towards the model, as shown in Fig. 16.20.
~

t Such a short run time will be insuffiCient to ensure stable motion statistics. A run time sufficient to
encounter at least one hundred wave crests is usually regarded as necessary to obtain consistent estimates
of rms motions in irregular waves. Even longer runs will be necessary to obtain consistent results for rarely
occurring events like deck wetness an9 slamming. The required run ~me can be achieved by scarfing short
runs together. The 'mean' rms niotions can be calculated by averaging the motion variances obtained in
each short run.

Model testing

324

[Ch. 16

If the model speed is very low these reflected waves will return to the centre of the
tank before the model has moved away, as shown in Fig. 16.20(a). In this case the
model's motions will be influenced by these reflected waves and misleading results
will be obtained. If the model speed is high enough the reflections will reach the
centre of the tank after the model has passsed by and no interference will occur.
The critical velocity at which tank wall interference occurs may be calculated as
follows.
Suppose that the model is running in head or following waves down the centre of a
long narrow tank. The encounter frequency is then given by

ro 2

ro -

u cos J-t
g

where cos J-t is - 1 or


radiated waves is

radians/second

(16.52)

+ 1 depending on the model's heading. The celerity of the

metres/second

(16.53)

If BT is the width of the tank an individual wave crest takes

(16.54)

to travel from the model to the tank wall and back to the tank centreline. Tank wall
interference will occur if the model moves less than its own length in this time. The
critical model speed at which interference begins is

Lm metres/second

(16.55)

Combining equations (16.52)-(16.55) we find that the critical speed is

Ucrit

00

~os J-t [ 1 ~ (1- 4 cos J-t ~~)]

metres/second (16.56)

or, in non-dimensional form,

(head waves) (16.57a)

Sec. 16.11]

Tank wall interference

325

(a) U<Ucrit: interference

(b) U>Ucrit: no interference

Fig. 16.20- Tank wall interference: ---radiated waves; - -1 ~ - reflected waves.

(following waves)
(16.57b)
Figure 16.21(a) illustrates the relationship giv~n by equation (16.57a) for head
waves. In this case the Froude number must be greater than FN crit to avoid tank wall
interference: the critical Froude number increases with the length of the model and is
very large for low frequencies (long waves).
Fig. 16.21(b) shows the critical Froude number for model tests in following waves
obtained from equation (16.57b). In this case the Froude number must lie within a
finite range to avoid interference and the range decreases as the model length
increases. When the model length is one-quarter of the width of the tank there is only
one Froude number for each w.ave frequency that will give results which do not suffer
from interference. Models of greater length will_,_always experience tank wall

326

[Ch. 16

Model testing

(a) Head waves


No tank wall
interference

0.3
FN crit

10
,,,V(Lmlg)
i.!Lm

0.5

(b) Following waves

0.4
Tank wall
interference

0.3

FN crit

Fig. 16.21- Tank wall interference: valid speeds.

interference whatever the Froude number or wave frequency. It follows that the
maximum permissible model length for tests in following waves is one-quarter of the
width of the tank.

17
Probability formulae

17.1

INTRODUCTION

In Chapters 4 and 14 it was shown that the irregular time histories of both waves and
ship motions could be characterised in terms of energy spectra and various statistical
quantities like mean values, periods, rms values and so on. Seakeeping studies,
however, often demand a more intimate knowledge of the characteristics of waves
and motions. In particular the likelihood of a particular event occurring (such as a
particular motion level being exceeded) is often of interest. Wave and ship motion
information. The discussion
time histories can be analysed to provide this sort
which follows is written in terms of wave analysis but applies equally to all ship
motions.

of

17.2

PROBABILITY ANALYSIS

Two methods of analysis of irregular time histories are comm0nly used. In the first
the time history is analysed by reading discrete values of the record at fixed intervals
of time (say every second) as shown in Fig. 4.3. This method can be used to find the
probability or the proportion of time that the wave depression exceeds a particular
level.
Some of the measurements obtained in this way may by chance be peaks, troughs
or zero crossings but they are not given any special significance. An alternative
method of analysis is concerned only with these ,salient points in the record and is
commonly used in many aspects of seakeeping work. Typically the analysis consists
of measuring successive wave amplitudes and periods as defined in Fig. 4.2. This
technique is used to extract information on the probability of peaks and troughs
exceeding a given level.
17.3 HISTOGRAMS

Whichever kind of analysis is. used the results will consist of an apparently random
sequence of measurements: these can be sorted _according to their values into

[Ch. 17

Probability formulae

328

discrete ranges or histogram 'bins'. Tables 17.1 and 17.2 show typical examples for a
30 minute wave time history using 0.5 metre bins. Table 17.1 gives results based on an
analysis of the record at discrete time intervals of 1.0 second: Table 17.2 gives the
corresponding analysis based on wave amplitudes measured from the mean surface
depression. No distinction is made between peaks and troughs.
In Table 17.1 it is shown, for example, that 37 measurements of the wave
depression were found in the range 4.5-5.0 metres below the arbitrary datum level
chosen for the analysis. In Table 17 .2, 23 measurements of wave amplitude (peaks
and troughs) were found in the range 3.5-4.0 metres. The corresponding histograms
are shown in Figs 17.1 and 17.2.

Table 17.1- Analysis of a wave time history at discrete time intervals

Histogram
bins

Number of
observations
in each bin

Probability
of
occurrence

(metres)

Nh

Probability
density
function
ordinate
f (metres- 1 )

0
1
3
11
23
36

0.000
0.001
0.002
0.006
0.013
0.020
0.039
0.068
0.100
0.117
0.133
0.139
0.117
0.096
0.067
0.042
0.021
0.011
0.005
0.003
0.002
0.001
0.000

0.000
0.001
0.003
0.012
0.026
0.040
0.079
0.136
0.200
0.233
0.267
0.278
0.233
0.191
0.133
0.083
0.041
0.022
0.010
0.006
0.004
0.001
0.000

-3.5 to
-3.0to
-2.5 to
-2.0 to
-1.5 to
-1.0 to
-0.5 to
O.Oto
0.5 to
1.0 to
1.5 to
2.0 to
2.5 to
3.0 to
3.5 to
4.0 to
4.5 to
5.0 to
5.5 to
6.0to
6.5 to
7.0to
7.5 to

-3.0
-2.5
-2.0
-1.5
-1.0
-0.5
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
5.5
6.0
6.5
7.0
7.5
8.0

71

122
180
210
240
250
210
172
120
75
37
20
9
5
4
1
0

Mean surface depression~ = 2.01 metres


Length of record: 1800 seconds.
Time interval: 1.0 second.

329

Histograms

Sec. 17.3]

Table 17.2- Amplitude anaiysis of wave time history


Histogram
bin

(metres)

Number of
measurements
of wave
amplitude
Nh

Probability
of
occurrence

169
171
234
265
204
145
101
23
12
0
Total: 1324

0.128
0.129
0.177
0.200
0.154
0.110
0.076
0.017
0.009
0.000

0.0 to 0.5
0.5 to 1.0
1.0 to 1.5
1.5 to 2.0
2.0 to 2.5
2.5 to 3.0
3.0 to 3.5
3.5 to 4.0
4.0 to 4.5
4.5 to 5.0
Mean wave amplitude ~a
TH

Probability
density
function

(metres -l)
0.255
0.258
0.353
0.400
0.308
0.219
0.153
0.035
0.018
0.000

= 1. 7 metres

1800 seconds.

300
Mean surface depresion
Arbitary
datum

;i

250

r--

200

Q)

r-

c
<J)

Q)

E 150

::J

<J)

Q)

Q)

_Q

r-

50

1-

r-

-4

co

....E 100

,w,

--1
1-

r-

E
::J
z

r-

~ rJ

r-

_Q

.c
(J
co

r-

rrf

-2

1~

h
0

.Surface depressions (metres)

Fig. 17.1 -Typical histogram of measurements at regut'ar time intervals (Table 17.1).

330

Probability formulae

300

Mean amplitude

).._
~
..c

()

co

Significant single
amplitude

ii
Q)

[Ch. 17

200

1--

c"'
Q)

::J

"'co
Q)

0 100

1-

!-

Q)

..0

E
::J
z

G:-,

Wave amplitude~. (metres)

Fig. 17.2- Typical histogram of amplitudes (Table 17.2).

The total number of measurements (whether they be amplitudes or measurements obtained at fixed time intervals) is obtained by summing the observations in all
the histogram bins:

(17.1)

The probability of an individual measurement lying in the range

s to s is
1

(17.2)

and the probability of an individual measurement exceeding a given value


t It is assumed that /;; 1 and 1;;2 corresponds to histogram bin boundaries.

s is
1

The probability density function

Sec. 17.4]

331

(17.3)

The mean value of all the observations is given by

-1;,

IS =

=-

(17.4)

where Sh corresponds to the centre of each histogram bin of width wh.


Tables 17.1 and 17.2 and Figs 17.1 and 17.2 show mean values of surface
depression and wave amplitude obtained from equation (17.4).
It is almost invariably found in seakeeping analysis that the histograms have
peaks close to the calculated mean value. In other words observations close to the
mean value are very common. The histogram of Fig. 17.1 also shows the characteristic bell shaped symmetry about the mean value obtained from measurements at
regular time intervals: large deviations from the mean are rare. The typical
amplitude histogram (Fig. 17 .2) is not symmetrical about the mean: small amplitudes
are more common than large ones.

17.4 THE PROBABILITY DENSITY FUNCTION

1
rl
The histogram has a major disadvantage: the ordinates (the numbers of observations in each bin) depend on the record length and the width,ofthe histogram bins
and this complicates comparisons between different results. The probability density
function (PDF) is a form of histogram which eliminates this dependency on record
length and bin width. The PDF is defined such that the area enclosed by the PDF
curve over a bin is equivalent to the probabilty of the measurement falling within that
bin.
Iff is the probability density function ordinate and wh is the width of the bin

p = Nh
N

(17.5)

and the PDF ordinate is

(17.6)

Probability formulae

332

[Ch. 17

The probabilities defined in equations (17.2) and (17.3) can now be written

(17.7)

(17.8)

The total area under the PDF is equivalent to the probability that an individual
measurement will lie within the range of all the measurements:

P(- oo < ~ < oo)

wh

Lf = 1.0

(17.9)

The mean surface depression is now given approximately by

(17.10)

Probability density function ordinates corresponding to the histograms already


discussed are given in Tables 17.1 and 17 .2. They are illustrated in Figs 17.3 and 17 .4.
The results are exactly the same shape as the histograms: all the ordinates have
simply been reduced in the ratio 1/(Nwh).
The probability density function defined in this way is discontinuous with a finite
number of bins and ordinates. If the bin width is reduced the number of ordinates is
increased and if the record is of sufficient length the PDF will gradually tend towards
a continuous smoothly varying curve. The probabilities defined earlier now become

(17.11)

(17.12)
and the total area under the continuous PDF curve is unity:

(17.13)

Area = Probability
of depression
lying in range
3.0-3.5 metres

~"'
Q)

E
~

333

The Gaussian probability density function

Sec. 17.5]

0.2

"'B
c

.....::J

u;
c

Q)

"'0

.
:c

0.1

co

.0
0

a:

o~--~~~~~~~._~

-4

-2

0
Surface

2
depression~

Fig. 17.3- Typical probability density function for

(metres)

measu~ements

at regular time intervals.

17.5 THE GAUSSIAN PROBABILITY DENSITY FUNCTION


It is usually found that the probability density funCtion for rrt~~surements of the wave
depression ~ sampled at regular time intervals is closely approximated by the
Gaussian or normal distribution formula:

(17.14)
where m 0 and ~ are the variance and mean value of the time history as defined in
equations (4.1) and (4.2). This convenient result,means that the PDF for a regularly
sampled wave record can be estimated if the mean and variance are known.
As an example, Fig. 17.3 shows the Gaussian PDF obtained for the wave time
history analysed in Table 17 .1.
In practice it is usual to arrange fOJ the mean value ~to be zero by making the
arbitrary datum for the measurements the same as the mean value and equation
(17.14) becomes
-

(17.15)

[Ch. 17

Probability formulae

334

Corresponding
Rayleigh PDF
c
0

.;::;
u

Area = probability
of amplitude
lying in range
3.0-3.5 metres

.2

.
(f)

Q)

"0

.
.a
co
.a
0

a:
2
Wave amplitude

sa (metres)

Fig. 17.4- Typical probability density function of amplitudes.

The probability that an individual measurement will lie within the range /,; 1 to /,;2 is
now given by

(17.16)
where the error function is defined as

erf(x)

= y'(121T) Jx exp (- -z2) dz


2
0

and is tabulated in Table 17.3. The error function has the properties

(17.17)

Sec. 17.5]

The Gaussian probability density function

335

Table 17.3- Values of the error function erf(x)

erf(x)

x erf(x)

x erf(x)

0.00
0.05
0.10
0.15
0.20
0.25
0.30
0.35
0.40
0.45
0.50
0.55
0.60
0.65
0.70
0.75
0.80
0.85
0.90
0.95

0.000
0.020
0.040
0.060
0.079
0.099
0.118
0.137
0.155
0.174
0.192
0.209
0.226
0.242
0.258
0.273
0.288
0.302
0.316
0.329

erf(- x)
erf(- oo)

= y'(1211')

1.00
1.05
1.10
1.15
1.20
1.25
1.30
1.35
1.40
1.45
1.50
1.55
1.60
1.65
1.70
1.75
1.80
1.85
1.90
1.95

0.341
0.353
0.364
0.375
0.385
0.394
0.403
0.412
0.419
0.427
0.433
0.439
0.445
0.451
0.455
0.460
0.464
0.468
0.471
0.474

Jx exp (-- 2- 2) dz
2

x erf(x)

2.00
2.05
2.10
2.15
2.20
2.25
2.30
2.35
2.40
2.45
2.50
2.55
2.60
2.65
2.70
2.75
2.80
2.85
2.90
2.95

0.477
0.480
0.482
0.484
0.486
0.488
0.489
0.491
0.492
0.493
0.494
0.495
0.495
0.496
0.497
0.497
0.497
0.498
0.498
0.498

erf(x)

erf( oo)

3.00
3.05
3.10
3.15
3.20
3.25
3.30
3.35
3.40
3.45
3.50
3.55
3.60
3.65
3.70
3.75
3.80
3.85
3.90
3.95

0.499
0.499
0.499
0.499
0.499
0.499
0.500
0.500
0.500
0.500
0.500
0.500
0.500
0.500
0.500
0.500
0.500
0.500
0.500
0.500

(17.18)

0.5

so that the probability of an individual measurement


level 1 is

x erf(x)

(17.19)

sexceeding a given positive


(17.20)

Note that equation (17 .20) refers to positive values and relates only to one side of the
Gaussian probability density function. The probability of an individual measurement
lying outside the range 1 is given by

Probability formulae

336

1-2

[Ch. 17

erf(~J

(17.21)

The probability that an individual measurement will lie within the total range of all
measurements is

P(- oo

< s< oo) = erf( oo) + erf(- oo) = 1.0

(17.22)

These properties are illustrated in Fig. 17.5. Table 17.4 lists some of the salient

erf

~1
no

Probability of
0

-'2
JJ

t less than zero


-

0.5

Probability of
between
~1 and ~2

~lying

.....
~
'Ui

QJ

"'0

Probability of

.0

~exceeding

C1l

~2

.0

[l_

Fig. 17.5- Properties of Gaussian probability density function

(-1?)

\1(2rrm 0 ) exp 2mo

properties ofthe Gaussian PDF based on equations (17.20) and (17.21). From these
results it can be seen that about 95% of all measurements obtained from a regularly
sampled wave record will lie within twice the rms value.

The Rayleigh probability density function

Sec. 17.6]
Table 17.4 -

337

Gaussian probability formula: probabilities of exceeding various


levels

t;lcr0

Probability of
exceeding t;

Probability of
exceeding t;

0.0
0.5
1.0
1.5
2.0
2.5
3.0

0.500
0.308
0.159
0.067
0.023
0.006
0.001

1.000
0.616
0.318
0.134
0.046
0.012
0.002

The mean value of all the upward or downward observations in the record can be
obtained from

{"" YI'AY

1~1

00

= 2j (jdt; = 0.798Vm0
0

oo

(17.23)

J fdt;
0

17.6 THE RAYLEIGH PROBABILITY DENSITY FUNCTION


It is usually found and invariably assumed that the probability density function for
wave amplitudes is closely approximated by the Rayleigh distribution formula:

_k_ exp ( m0

t;~)

2m0

(17.24)

where m0 is the variance of the time history defined in equation (4.2). This
convenient result allows the amplitude PDF to be estimated if the variance of the
wave depression is known. The Rayleigh PDF for the wave time history analysed in
Table 17.2 is shown in Fig. 17 .4.
The probability that an individual measurement of amplitude t;a will lie within the
range t;a 1 to t;a2 is given by
-

Probability formulae

338

[Ch. 17

-1;;~2)
-1;;~1)
exp - -exp 2m0
2m0

(17.25)

and the probability that the amplitude will exceed a given levell;;a 1 is

(17.26)

This function is plotted in Fig. 17.6. Again, the probability that an individual
amplitude will lie within the total range of all the measurements is unity:

P( -

oo

< Sa <

oo )

_..!__

mo

exp (

oo

oo

-I;;~) dl;;a

2mo

1. 0

(17.27)

Equation (17 .26) can be rewritten to give the amplitude which has a given
probability of being exceeded:

or

(17.28)

Table 17.5 gives some sample results from equations (17.26) and (17.28). From
these results it can be seen that the probability of an individual amplitude exceeding

Sec. 17.7]

Significant wave height and related statistics

339

0.

E
ro

Cll

"0
Q)
Q)

(.)

X
Q)

.
:0
ro

.0

Cl...

rl

Fig. 17.6 - Probability of exceeding amplitudes (Rayleigh formula).

about three times the rms value is very small: only about one peak (or trough) in
every hundred would be expected to exceed this level.
17.7

SIGNIFICANT WAVE HEIGHT AND RELATED STATISTICS

It was shown in Chapter 5 that the significant wave height, defined as the mean of the
highest third of the heights recorded in a wave tim~ history, was closely related to the
average wave height estimated visually by an experienced observer. In the same way
it might be expected that the experienced sailor's estimates of 'average' ship motions
might be similar to their significant amplitudes. Interest is therefore often centred on
these quantities.
In more general terms an expressionis required for the mean value of the highest
(1/n)th of all observations of amplitudes (where n = 3 for significant values). If the
probability density function is known the required amplitude ~ltn is given by the
moment of area of the shaded portion shown in Fig, 17.7. The shaded area is of
course equal to lin.

340

[Ch. 17

Probability formulae

Table 17.5 -

0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0

Rayleigh probability formula: probability of exceeding various


amplitudes Sa
Probability of
exceedance
p

Probability of
exceedance
p

1.000
0.882
0.606
0.325
0.135
0.044
0.011
0.002
0.0003

1.000
0.500
0.333
0.100
0.010
0.001
0.0001

0.00
1.18
1.48
2.15
3.03
3.72
4.29

Sln

.....
c

.E
c

.....::J

.
(/)

Q)

""0

.
:0

Area=
probability of
amplitude
exceeding ~Y"

co

..0

e
a...

Amplitude::,

Fig. 17.7- Calculation of mean of highest lin amplitudes.

If the PDF is given by the Rayleigh formula (equation (17 .23)) the amplitude 1; 11n
(which is exceeded with a probability 1/n or once in n amplitudes) is given by
equation (17.28):

Sec. 17. 7]

Significant wave height and related statistics

341

(17.29)

The mean value of the highest lin amplitudes is

_
r

J"" safdsa

::.._...::;ll.;..:.n_ _

':>lin -

J fdsa
" safdsa
J
P(sa >
co

lin

lin

(from equation (17.12))

Slln)

nf"" I,;Jdsa

(17.30)

lin

If the PDF is given by the Rayleigh formula (equation 17.24) equation (17.30)
becomes

~lin

co

1,;2

( _ 1,;2)

_!!_exp - a dl,;a
2mo
lin mo

rl

so
(17.31)

Selected results are given in Table 17 .6. Of particular interest are the results for
1 andn = 3:
Putting n = 1 gives the mean value of all amplitudes

~a

Putting n
~113

1.25cro

(17.32)

= 3 gives the significant ~mplitude

2.00cr0

and the significant height is

(17.33)

Probability formulae

342

[Ch. 17

Table 17.6 - Mean of the highest 1/n


amplitudes (Rayleigh formula)

n
1
2

1.25
1.77

2.00

10
100

2.54
3.34
3.72
4.29

1000
10000

2~113

= 4.00cr0

(17.34)

These results are widely assumed to apply to all ship motions and wave records. It
should be remembered however that they are strictly only true if the Rayleigh
formula (equation (17.24)) holds.
17.8 JOINT PROBABILITIES

In some seakeeping studies interest is centred on the probability of two events


occurring simultaneously. If the two events are independent the probabilty of them
both occurring is simply the product of the probabilities of each individual event
occurring in isolation.
If the probabilities are given by the Rayleigh formula (equation (17.26)) the
probability of motion X; exceeding some level xil at the same time as motion xi
exceeds some level xi 1 is

exp (

-xA) exp (-xA)

2m0 ;

2m 0i

x1)
= exp ( - -xA- - ..:::1.}_
2m0 ; 2m 0i
where m 0 ; and moi are the variances of the respective motions.

(17.35)

18
Roll stabilisation

18.1

MOTION REDUCTION

If motions are an undesirable feature of the behaviour of a ship in rough weather, it is


natural to consider ways of reducing them. Methods of motion reduction are often
known by the generic name of motion 'stabilisation', although it should be realised
that this is usually an incorrect use of the word. The oscillatory motions of all
practical conventional ship designs are already 'stable' in that they can generally be
expected to return to an equilibrium datum level after some small disturbance: this is
ensured by the stiffness terms in the equations of motion. The term 'stabilisation'
implies an increase in the stiffness coefficients like c44 , but almost every practical
motion stabilisation device derives most of its .effect by r}ncreasing the motion
damping (coefficients like b 44 in equation (8.27)). They should therefore more
correctly be called motion dampers. However, the term stabilisation is so widely
used that the adoption of more pedantic terminology for this book would be
confusing and will be avoided.
In principle, stabilisation is possible for any motion. It is simply necessary to
provide some means (active or passive) of ar~ificially increasing the damping terms in
the appropriate equation of motion. We have already seen that certain ship motion
responses can be approximately represented by the simple second-order spring-mass
system. The possibility of motion reduction may be nicely illustrated by considering
the response shown for such a system in Fig. 6.-l. If the damping is very small the
system's amplitude response is very high at frequencies close to the natural frequency. With random excitation (as in the case of a ship at sea) most of the resulting
motions will be experienced at frequencies close to this frequency.
Increasing the decay coefficient 11 retluces these motions and equation 6.13 shows
that doubling the damping will halve the amplitude at the natural frequency.
However, this is only effective if the inherent damping is small. For very high initial
values of the decay coefficient the maximum motion amplitudes occur at zero
frequency rather than at the natural frequency, and increasing the damping then has

Roll stabilisation

344

[Ch. 18

little effect. So motion stabilisation is only likely to be effective if the inherent


damping of the unstabilised system is small. Practical considerations also demand
that the damping force or moment required of the stabilisation system must be
relatively small so that an effective degree of stabilisation can be achieved without
the need for massive engineering.
Roll is the only motion which meets these two requirements of low inherent
damping and relatively small stabilisation moment demands. Roll stabilisation has
therefore received considerable attention and many successful systems have been
installed in ships in service.
Pitch and heave stabilisation have received some attention and some success has
been claimed for small craft. However, the inherent damping is usually already so
high and the required forces and moments so large that practical systems for ships
remain an elusive goal.
18.2 BILGE KEELS

Bilge keels are the simplest form of roll stabilisation device. These are long narrow
keels mounted at the turn of the bilge as shown in Fig. 18.1. If active roll stabiliserfins

~c::::~

_____

c ______

bf-----~-BK_._P~-----J~
~I

Fig. 18.1- Bilge keel notation.

Sec. 18.2]

Bilge keels

345

are also required the bilge keels may be segmented to accommodate them as shown
in Fig. 18.2.

Fig. 18.2- Bilge keels and roll stabiliser fins on a destroyer. (MoD Photo.)

Bilge keels are very effective roll stabilisation devices which work well at all
speeds. They have the significant advantage that they have no moving parts and
require no maintenance beyond that normally given to the hull surface. Their only
disadvantage is that they increase the resistance of the ship, but the effects can be
minimised by carefully aligning the keels with the :flow streamlines around the bilges.
This is usually done using some kind of flow visualisation technique on a model
during the design stage. Correct alignment can only be achieved at one speed (the
cruising speed is usually chosen) but the resistance penalty at other speeds is usually
small.
Bilge keels work by generating drag forces which oppose the rolling motion of the
ship. The mechanism is similar to that shown for appendages at zero speed in Fig.
12.5. The roll damping moment for a single bilge keel is given by equation (12.28)
and the equivalent linearised toll damping coefficient is given by equation (12.31). It
remains to determine suitable values of the drag coefficient for bilge keels, C0 . Cox
and Lloyd (1977) cited experimental data published by Martin (1958) and by

[Ch. 18

Roll stabilisation

346

Ridjanovic (1962).This is presented (in slightly different form) in Fig. 18.3. The
normal force coefficient is given as a function of the equivalent aspect ratio
(18.1)
and the non-dimensional bilge keel radius parameter
(18.2)
The results may be expressed in the form

C0

= 0.849

(f,

[1- exp(- Kr')] +

J)

(18.3)

0.025
0.05

...c

0.10

Q)

'()

0.15

Q)

(.)

0.20

Ol

0.30
0.40

0.14

0
Equivalent aspect ration a 8 K

Fig. 18.3- Bilge keel drag coefficients.

Sec. 18.2]

Bilge keels

347

where

/=14.66-J
K

1(_2_)

'V

(18.4)

aBK

J = 2.37- 5.33aBK + 10.35a~K

K=

(18.5)

y'(2aBK)
(14.66- J)(0.109- 0.208aBK)

(18.6)

Fig. 18.3 shows the benefits of increasing the bilge keel aspect ratio for a given bilge
keel area. In other words a short wide bilge keel is much more effective than a long
narrow bilge keel.
Figs 18.4 and 18.5 show examples of the calculated effects of bilge keels on roll
motion for a frigate in a moderately severe seaway. The heaviest rolling occurs in
quartering seas and a pair of 30 metre 2 bilge keels offer a substantial reduction in
motion. Fig. 18.4 demonstrates the superior performance of bilge keels of high
aspect ratio and Fig. 18.5 shows that the bilge keels are most effective at low speeds.

Bilge keel 30 x 1.0m

e
en

E
FT, 3 =5.5 m,

T0 = 12.4 sec

'

Heading (degrees)

Fig. 18.4- Effect of bilge keel aspect ratio on roll motion of a frigate at 20 knots.

348

Roll stabilisation

[Ch. 18

H113 =5.5 m, T0 =12.4 sec


15
(ij'
Q)

Cl
Q)

::9.

e
"'
E
....

20

25

30

Speed (knots)
Fig. 18.5- Effect of bilge keels on rolling at worst heading.

Fig. 18.6.- Low aspect ratio non-retractable fins fitted to a warship. (MoD Photo.)

Sec. 18.3]

Active roll stabiliser fins

349

18.3 ACTIVE ROLL STABILISER FINS


18.3.1 General features
Active roll stabiliser fins are usually mounted on rotatable stocks at the turn of the
bilge near the middle of the ship as shown in Fig. 18.6. The angle of incidence of the
fins is continually adjusted by a control system which is sensitive to the rolling motion
of the ship. The fins develop lift forces which exert roll moments about the centre of
gravity of the ship. These roll moments are arranged to oppose the moment applied
by the waves and the roll motion is reduced.
At speeds above 10-15 knots active fins are probably the most effective method of
stabilising a ship. Reductions in rms roll motion of at least 50% are usually possible in
moderate waves with a well designed system. However, the fins become progressively less effective as the speed is reduced and they are not usually specified for ships
which habitually operate at low speed. It should also be understood that fins have a
limited capacity and their ability to reduce roll motion decreases in very severe sea
states. They are relatively sophisticated and expensive pieces of equipment and
require considerable maintenance. Nevertheless, their ability to work well over a
wide range of conditions has earned them almost universal acceptance and they are
now fitted to many ships.
Retractable fins are often specified for merchant ships as illustrated in Fig. 18.7.
The fins can be withdrawn into the hull when the ship is operating in calm weather to
eliminate their small resistance penalty. This feature is also used to eliminate the risk

Fig. 18.7 -

High aspect ratio retractable fin fitted tQ,.a merchant ship. (Reproduced by
permission of Sperry Marine Inc.)

[Ch. 18

Roll stabilisation

350

of grounding when the ship is operating in shallow water or coming alongside.


Retractable fins are usually of high aspect ratio and are hydrodynamically very
efficient, giving a relatively large lift for a given fin area (see Figs 2.21 and 2.22).
In warships it is usual to fit non-retractable fins as these have a greater immunity
to damage from shock and explosion. It is then necessary to confine the fins to the
enclosing rectangle defined by the ship's maximum beam and draught (see Fig. 18.8).
This places an effective limit on the area and aspect ratio which can be adopted and
these fins are usually rather less efficient than their retractable counterparts.

[J

Fixed fins must


lie within the
rectangle

Fig. 18.8- Fin notation.

18.3.2 Constraints on stabiliser fin design


The requirement to keep the tips of non-retractable fins inside the enclosing
rectangle limits the fin outreach. It follows that the only way to increase the lift
available from the fin is to increase the chord. Unfortunately this reduces the aspect
ratio and the lift curve slope (see Figs 2.21 and 2.22).
Consider a rectangular fin of unit aspect ratio and outreach limited by this
constraint. Let the lift developed at some angle of incidence be (L)a.= 1 kN. Let the
lift developed by another fin of the same limited outreach but a different aspect ratio,
chord and area beL kN. Rearranging equations (2.61), (2.63) and (2.64) it can be
shown that the ratio of the lifts developed by the two fins is

Sec. 18.3]

Active roll stabiliser fins

351

4.04
(L)aF=l -1.8 + y(4 +a~)
L

(18.7)

Equation (18.7) is plotted in Fig. 18.9 and it can be seen that the gains associated
with increasing the chord if the outreach is limited are minimal if the aspect ratio is
reduced below about 1.0. Thus practical non-retractible fins are limited to an aspect
ratio of about 1.0. More lift can therefore be achieved only by installing more pairs of
fins.

Limited
outreach

o __o__

Increasing
area

3.0
Aspect ration aF

Fig. 18.9- Effect of aspect ratio at constant outreach.

18.3.3 Equations of motion


Fig. 18.10 shows the forces and moments applied to the ship by a pair of fins at an
angle of incidence to the flow. Each fin develops a lift force

(18.8)
exerting a roll moment
2L

rF

kN metres

to port about the -centre of gravity

352

[Ch. 18

Roll stabilisation

View looking
forward

-.Lsmf-1
L cos f-l
L

2L Sin

f-l

Fig. 18.10- Sway force and yaw moment caused by stabiliser fins.

The two lift forces have vertical components which cancel so that there is no resultant
vertical force on the ship. However, their horizontal components add and yield a
sway force
2 L sin

kN

to starboard

If the fins are mounted xBlF metres forward of the centre of gravity, this horizontal
force will exert a yaw moment
2L

xBlF

sin

kN metres

to starboard

Sec. 18.3]

Active roll stabiliser fins

353

The effects of the stabiliser fins on the motions of the ship in waves may be
computed by including these additional terms in the equations of motion
(8.24)-(8.29). Only the lateral plane equations are affected and these become

+ ~ EF L sin ~ = Fwzo sin (ole{+ y2 )

kN

(18.9)

(18.10)

yaw:

a62 Xz + b62 Xz + a64 X4 + b64 X4 + (/66 + a66)i6 + b66 i 6 + c66 x 6

+ ~ EF L

xBlF sin~=

Fw 6 o sin (wet+ y6 )

kN metres

(18.11)

In these equations the summations refer to the number of fins fitted to the ship and
is an effectiveness factor defined as

EF

E = effective lift of fin


F
nominal lift of fin

(18.12)

EF is generally less than 1.0 because of various hydrodynamic effects which are
discussed in section 18.3.4.

18.3.4 Hydrodynamic losses

18.3.4.1 General
Lloyd (1975, 1977) investigated the effectiveness of roll stabiliser fins by measuring
the lift developed by model stabiliser fins and bilge keels in a variety of configurations
on a ground board. His experiments were conducted in the Circulating Water
Channel at the Admiralty Research Laboratory at Haslar in the United Kingdom.
His apparatus is sketched in Fig. 18.11. He identified three major causes ofloss in fin
performance:
(a) hull boundary layer
(b) fin-fin interference
(c) fin-bilge keel interference.

Roll stabilisation

354

[Ch. 18

0 0.

Fig. 18.11- Lloyd's (1975, 1977) fin stabiliser experiments.

18.3.4.2 Hull boundary layer losses


The fins are partially immersed in the slow moving boucdary layer on the hull
surface. The flow velocity near the root of the fin and the lift developed in this region
are reduced. Lloyd (1975) measured the boundary layer thickness (defined as the
point at which the velocity is 99% of the freestream velocity) and the lift developed
by an isolated fin at various locations on his ground board. Fig. 18.12 shows the
effectiveness of the fin as a function of the boundary layer thickness.

1.0

Fig. 18.12- Effect of boundary layer on fin lift.

Sec. 18.3]

Active roll stabiliser fins

355

The results may be approximated by the empirical equation

E
BL

lift developed in boundary layer


nominal lift (no boundary layer)

=1 0_ 0

21

bF

(18.13)

The boundary layer thickness on the hull may be estimated using the equation
8 = 0.377

Xpp (RN)

-o.z

(18.14)

where RN is the local Reynolds number defined as

_ p U Xpp
R N-

(18.15)

J.Lw

18.3.4.3

Fin-lin interference

Roll stabiliser fins, like all lifting surfaces, work by developing a pressure difference
between their upper and lower surfaces. The water is tempted to roll round the tip of
the fin from the high-pressure to the low-pressure surface and a vortex is formed. Fig.
18.13 shows the vortex generated by a fin at a fixed (nose up) angle of incidence ex.
This vortex is shed from close to the tip of the fin and trails away along the side of the
hull imparting a swirling motion to the water close to the hull. This causes a
'downwash' in the region between the vortex and the hull surface and an 'upwash' in
the region outboard of the vortex. Clearly the sense of the ~lfirling motion and the

Lift

Fig. 18.13- Trailing vortex generated by a lifting surface.

356

Roll stabilisation

[Ch. 18

resultant flow directions depend on the direction of the lift developed by the fin. In
Fig. 18.13 an upward lift on a port side fin produces a clockwise motion (looking
forward). A nose-down fin deflection would produce an anticlockwise vortex
rotation. These directions are reversed for fins on the starboard side of the ship.
An oscillating fin produces a vortex of continually varying strength and direction
which is convected away along the side of the ship. In fact the vortex is a record or
'memory' of the lift developed by the fin. Fig. 18.14 illustrates the flow behind an
oscillating fin and it can be seen that there are alternate regions of downwash and
upwash in the wake ofthe fin, depending on the lift developed in the immediate past.
A second stabiliser fin mounted immediately behind the first fin will experience a
downwash over most of its outreach. This will generally decrease its angle of
incidence and reduce the lift developed. If the fin is mounted further aft in a region of
upwash, the lift developed will be increased.
Fig. 18.15 shows the fin-fin interference factor E 1F measured by Lloyd (1977) for
a pair of oscillating fins. The results are plotted as a function of a non-dimensional
frequency parameter and the longitudinal separation of the fins. At zero frequency

(a) Unfavourable

Increased
lift

(b) Favourable

Fig. 18.14- Fin-fin interference for oscillating fins.

Sec. 18.3]

Active roll stabiliser fins

357

(I)~F =0.20

0.16

0.12

0.08
0.04
0

30

Fig. 18.15 -Fin-fin interference factors. I rJ

the interference is quite dramatic even for well spaced fins. For example, a fin spaced
20 outreaches behind the first fin will develop only about 50% of the nominal lift. The
interference effects become less important as the frequency and the separation are
increased,' until at very high frequencies and separations the second fin is in a region
of upwash and the interference becomes beneficial.
18.3.4.4 Fin-bilge keel interference
A bilge keel mounted abaft a stabiliser fin will also experience downwash and will
develop a lift which opposes the fin lift. Lloyd (1975) measured this opposing lift for
the case of zero frequency and his results are presented in the form of an effectiveness
factor
E IBK

in Fig. 18.16.

= 1 0 _ bilge keel lift = 0 84


.

fin lift

(18.16)

[Ch. 18

Roll stabilisation

358

20

Fig. 18.16- Fin-bilge keel interference factors.

The detrimental effects of an aft mounted bilge keel are mitigated if the bilge keel
is followed by a second stabiliser fin. The bilge keel has a straightening effect on the
flow and removes some of the down wash due to the trailing vortex from the upstream
fin. This enhances the lift generated by the second fin as shown in Fig. 18.17. This
effect can be approximated by setting
EIBK

= 1.0

(18.17)

for the case of a bilge keel between two fins.

18.3.4.5 Overall effectiveness


The hydrodynamic losses described above are cumulative and the overall effectiveness of each stabiliser fin-bilge keel combination is given by multiplying the
individual effectivenesses:
(18.18)

18.3.5 Design recommendations


Clearly a single pair of fins with no aft mounted bilge keel is the most effective
stabiliser fin configuration. This suffers from no interference effects and

Sec. 18.3]

Active roll stabiliser fins

20

30

359

40

50

Fig. 18.17- Fin-fin interference factors. Effect of bilge keel at zero frequency.
1

rl

Such a configuration will only suffer from relatively insignificant hull boundary layer
losses. It is probably not worthwhile taking the trouble to locate the fin as far forward
as possibl~ to minimise these losses as this will make it difficult to accomodate a bilge
keel of adequate size forward of the fin.
If sufficient stabilisation cannot be obtained from a single pair of fins (bearing in
mind the limitations which may be imposed on the fin outreach) it will be necessary to
adopt a multiple fin configuration. In this case, the separation should ideally be
chosen to take advantage ofthe favourable interference effects shown in Fig. 18.15.
The fin configuration should be optimised to achieve the best performance at the
natural roll frequency (where most of the roll motion occurs) and at the ship's
cruising speed. In principle we require ..the second fin to be at a distance xFF metres
abaft the first fin such that the tim~ required for the vortex to convect from the
upstream fin is equal to half a roll period:
Xpp

T.4

1T

-=--

seconds

360

Roll stabilisation

[Ch. 18

so that

7TU

xFF=- metres
(J).4

(18.19)

Unfortunately this gives fin separations which are often impractical. For example,
taking a cruising speed of 10.3 metres/second (20 knots) and a natural roll frequency
of 0.5 radians/second (natural period 12.6 seconds) we find an optimum separation

xFF = 63 metres
which may be difficult to achieve on all but the largest ships.
Bilge keels should not be located abaft the after fin.
In order to obtain the maximum possible roll moment the stabiliser fins should be
mounted at the turn of the bilge so that the roll lever arm rp is maximised (see Fig.
18.8). It is also advantageous to mount the fin stock normal to the hull surface. This
simplifies the mechanical arrangements and minimises the gap between the hull
surface and the fin root when the fin is at an angle of incidence. Such gaps are a
potential source of leakage between the high- and low-pressure sides of the fin and
will result in a considerable loss of effectiveness. Fixed fins should also be aligned to
maximise the fin outreach within the enclosing rectangle of the ship.
These requirements often result in large values of the fin depression angle ~ (45
degrees or more). Reference to the modified equations of motion (18.9)-(18.11)
shows that this will result in sway and yaw motions whenever the fins move in their
efforts to control the roll motion.
Suppose that the fins are set to some fixed angle of incidence to give steady lift
forces L kN to generate a stabilising roll moment to port as shown in Fig. 18.18. We
have already seen that this will result in a sway force and a yaw moment to starboard.
The ship will respond in exactly the same way as it responds to motions of the rudders
and will begin to turn to starboard. A centrifugal force to port then acts through the
centre of gravity, opposed by inboard hydrodynamic forces acting to starboard below
the waterline. These forces form a couple which tends to roll the ship to port,
enhancing the port roll moment directly generated by the fins. The total roll moment
generated by forward mounted fins is therefore increased by this sway-yaw effect.
If the fins are mounted abaft the centre of gravity x 81p is negative. The ship then
turns in the opposite direction in response to the fins (to port for the example given
above). The roll moment caused by the turning motion of the ship then opposes the
roll moment directly generated by the fins and their effectiveness is reduced. In
extreme cases with near vertical fins mounted well aft the roll moment due to the
turning motion may actually exceed the roll moment directly generated by the fins
and the total moment will then be in the 'wrong' direction.
These effects are reduced when the fins are oscillating at higher frequencies and
they are not usually very significant at the natural roll frequency where most of the
rolling motion occurs. However, we shall see in section 18.3.6 that extreme aft fin

Sec. 18.3]

Active roll stabiliser fins

361

Outboard .,.,..,__ _
.-..,
1- - - - - - "
centrifugal
force

Inboard
hydrodynamic
force

(a)

Couple enhances
roll moment
generated
by fins

Fins mounted
forward

IJo
View looking forward

Net force to starboard causes


ship to turn to starboard

Outboard
centrifugal _ _,.. - - - - - force

Couple opposes
roll moment
generated by
fins

(b) Fins mounted

aft

Inboard
hydrodynamic 01111114t--- \ - - - - - - - - ' - force
View looking forward

Net force to starbo'ard causes


ship to turn to port

Fig. 18.18- Beneficial and detrimental sway and yaw effects.

locations with large angles of depression may result in motion amplification at very
low frequencies. For this reason these locations should be avoided if possible.
Beneficial sway and yaw effects can be maximised by mounting the fins well
forward with a large angle of depression. However, this makes it difficult to
accomodate a bilge keel forward of the fins and is rarely attempted.
In practice fins are usually mounted somewhere near the middle of the ship and
the sway-yaw effects may thep degrade their performance at low frequencies. The
degradation can be minimised by keeping the depression angle ~ as small as possible
but this will also tend to reduce the roll lever arm r.p':

362

Roll stabilisation

[Ch. 18

18.3.6 Active fin control systems

18.3.6.1 Introduction
Fig. 18.19 shows a block diagram representation of a ship stabilised with active fins.
Each component or block in the diagram may considered as a 'black box' having an
input and an output which are related by the block's transfer function. For example,
the ship block accepts an input in the form of a roll moment Fw4 from the waves and
generates a roll motion output x 4 Similarly the stabiliser fin controller generates a
demanded fin angle <Xn in response to the roll motion of the ship. The fin servo
mechanism responds and drives the fins to an achieved fin angle ex and the fins convert
this into a stabilising roll moment Fp4 This is subtracted from the roll moment
generated by the waves, thus reducing the roll motion of the ship.

Fw4

Ship

x4

Roll
F

FF4

Fins

Gyro

Measured
roll
E

a
Actual
fin
angle

Fin
servo

uo

D
Controller

Demanded
fin an g le

Fig. 18.19- Block diagram for a ship with roll stabiliser fins.

The transfer functions of the ship, the fin servo and the fin are essentially fixed for
a given design. The fin controller transfer function is, however, adjustable and must
be set up in such a way as to ensure that the fins develop roll moments which generally

oppose the moments provided by the waves.

18.3.6.2 The lin controller transfer function


Fin controllers generally have transfer functions of the form
(18.20)

Sec. 18.3]

Active roll stabiliser fins

363

where
CXo
X4m

Ko
Ku

K1
Kz
K3
b 1 , b2 and b3

is the demanded fin angle


is the measure roll angle
is the overall gain setting
is the speed-dependent gain setting
is the roll angle sensitivity
is the roll velocity sensitivity
is the roll acceleration sensitivity
are fixed controller coefficients
is the Laplace transform operator (d/dt).

The coefficients K 0 , Ku, K 1 , K 2 , and K3 may be adjusted to achieve the desired


roll stabiliser performance within the capacity of the stabiliser fins. Table 18.1 shows
typical values which might be found on a modern controller. K 1 , K 2 and K3 should be
adjusted to match the rolling characteristics of the ship and will remain fixed for a
given loading condition, metacentric height and natural roll frequency. K 0 and Ku
govern the overall level of activity of the roll stabiliser system and are adjusted to
achieve a given roll stabilisation performance.
Table 18.1- Typical roll stabiliser controller coefficient values

Ko

Kl

Kz

K3

0.00
0.25
0.50
0.75
1.00
1.25
1.50
1.75
2.00
2.25
2.50

0
1
2
3
4
5
6
7
8

0
5
10
15
20
25
30
35
40
45
50

0
1
2
3
4
5
6
7
8

9
10

rl

9
10

bl = 1.00
b 2 =0.5
b3 =0.05
Ku = 0.0 for U < 5 knots
= 1.0 for 5 < U < 15 knots-= 225/ U 2 for U > 15 knots

The speed dependent gain Ku is introduced to compensate for the inherent


reduction in stabiliser performance as the ship speed is reduced and to avoid

Roll stabilisation

364

[Ch. 18

overloading the fin stocks at very high speed. In modern systems Ku is varied
automatically with

cc-

uz

as shown in Fig. 18.20. At very low speeds this would give very large gains, resulting
in excessive stabiliser fin activity and frequent demands for fin angles greater than the
maximum available. The fin servo mechanisms would be continually driving the fins
up against the mechanical stops, which are usually set to limit their travel to 25 or 30
degrees, leading to rapid wear and possible damage to the mechanical components.
So the speed-dependent gain is usually limited to some finite value at speeds less
than, say, half the cruising speed. When the ship is hove to the fins are completely

Ku= Constant

---,
I

I
I

c
co
O'l

....c

Kuu U2

Q)

"0

Manually
~switched
I

Q)

c.
Q)

"0

-o
Q)
Q)

c.

(Jl

Ku=O

u,

Speed

Fig. 18.20- Speed-dependent gain.

ineffective and the speed-dependent gain is then set to zero to avoid needless wear on
the system.
In less sophisticated systems the ship's crew are required to switch manually from
low to high-speed settings depending on the ship's speed.

Sec. 18.3]

Active roll stabiliser fins

365

18.3.6.3 Choosing settings for the sensitivities K 1 , K 2 and K 3


The object of the stabiliser control system is to ensure that the roll moment generated
by the fins opposes the roll moment generated by the waves. Consider the behaviour
of the system illustrated in Fig. 18.19. Let us suppose that the ship is in regular waves
of small amplitude. Then the sinusoidal disturbance initiated by the waves will
propagate around the system and each component will generate a sinusoidal output
depending on its individual sinusoidal input. Each component will introduce a phase
shift and we require the total phase around the loop from A to F to be zero. This will
ensure that the stabilising roll moment generated by the fins exactly opposes the roll
moment disturbance from the waves. This can only be achieved at one frequency and
it is customary to choose the natural roll frequency since this is usually the dominant
frequency in the roll motion.
In practice it is assumed that the only components which have significant phase
responses are the ship and the fin servo. So the controller must provide a phase
advance which exactly compensates, at the natural roll frequency, for the phase lags
introduced by these two components. The controller phase is then required to be
(18.21)
where e8 and Eps are the phases of the ship and the fin servo at the natural roll
frequency.
By making the substitution

and setting

in equation (18.20) it can be shown that the controller phase is given by

Ec -

tan

_ 1 (

K2 ro.4
K _K
1

)
2

3 ro.4

tan

_ 1 (

b2 ro.4
b _b 2
1

3 ro.4

rad1ans

(18.22)

and hence the controll7r sensitivities must satisfy the relationship

(18.23)
where the required phase angle is
(18.24)

Roll stabilisation

366

[Ch. 18

Equation (18.23) may be written


(18.25)

from which it can be seen that the required phase must be achieved by choosing
appropriate ratios between the sensitivities.
Practical application of this technique requires the phases E8 and Eps to be
determined. These may be found in a forced rolling trial at sea as shown in Fig. 18.21.
The ship is run in calm water so that the roll moment from the waves is zero or at least
negligible. If any waves are present their effects should be minimised by running in
head or following seas. Rudder motions will influence the rolling motions of the ship
so the helmsman should keep the wheel amidships and the autopilot, if the ship has
one, should be switched off.

Fig. 18.21- Forced rolling trial in a frigate. (Photo: Author.)

Sec. 18.3]

367

Active roll stabiliser fins

The stabiliser controller is isolated from the system by breaking the circuit at the
point Din Fig. 18.19. The fin servos are instead driven by a sinusoidal demand signal
equivalent to, say, 15 degrees fin amplitude at some selected frequency. The ship
will then roll at the same frequency and the roll response is measured at the point C in
Fig. 18.19. The actual fin angle is monitored at point E. Es is the phase between the
signals monitored at C and E.
Fig. 18.22 shows the expected form of the results. These depend on the location
and depression of the fins. For fins mounted somewhere near the middle of the ship
with a moderate angle of depression the phase at zero frequency is zero, showing that
the ship rolls in the expected sense: a fin incidence giving a steady roll moment to port
results in a steady port heel.
0.8

.01 -0
~

)(

Q)

"'0

-~
0.

E 0.4

"'c

!;

=e1

0
a:

"'I
~I

z
0

I
1.0

1.5

: il

200

(j)

7
7

Q)

~
O'l
Q)

2.

Q)
(/)

"'0.

..c

-100

Frequency (radians/second)
-

Fig. 18.22- Typical forced roll responses.

Roll stabilisation

368

[Ch. 18

If the fins are mounted well aft with a large depression angle the phase at zero
frequency becomes 180 for the reasons already explained in section 18.3.5. The roll
response to the fins can be quite large but is then in the opposite sense to that
expected.
In either case the ship phase Es can be measured at the natural roll frequency as
shown.
The fin servo phase Eps is measured between the signals monitored at points D
and E in Fig. 18.19. Typical results are given in Fig. 18.23.

1.5

1.0

"

t$

I=>
0 {)

0.5

I Cii~

I :::J~

r-

....

Ql

:::J

0"

Ql

"' ~
JZ"""
II
0

0.5

1.0

1.5

0.5

1.0

1.5

0
Ui
Ql

~
Ol
Ql

~-100
Ql
(/)

"'

..c

a..

-200

Frequency (radians/second)

Fig. 18.23- Typical fin servo response.

An alternative approach is to simulate a forced rolling trial using the equations of


motion (18. 9)-(18.11). The excitations from the waves on the right-hand sides are set
to zero and the lift force L is made to vary sinusoidally by putting

<X= <Xo

sin (ole()

radians

(18.26)

in equation (18.8). The equations will then give sinusoidal motion responses(in sway
and yaw as well as roll) and the phase E5 may be determined.

Sec. 18.3]

Active roll stabiliser fins

369

Worked example
Find appropriate controller coefficients to match the responses given for forward
mounted fins in Figs 18.22 and 18.23 using the controller coefficients listed in Table
18.1.
Natural roll frequency:

co. 4 = 0.45 radians/second

Phases at natural roll frequency:

~>s

= - 70,

~>Fs

= - 10

From equation 18.24 the required phase angle is

~:- 70 + 10 +tan

-l(

0.5X0.45
)- 0
l.O _ O.OS x 0.4S 2 - 93

hence from equation (18.25)

(Kl-

Kz = - 42.4
K3
K3

0.2)

This relationship can be achieved more or less exactly by a large number of


combinations of the available control sensitivities. For example, the following
settings all satisfy equation (18.25) and give the required phase advance within 1
I fi
degree:

K1

Kz

K3

0
0
0
0
0
0
1
1
1
1
1

10
15
25
35
40
50
40
30
25
15
10

1
2
3
4
5
6
10
9
8
7
6

All of these settings will give a satisfactory perfq;rmance since they all ensure that
the moment applied by the stabilisers exactly opposes the wave moment at the

370

Roll stabilisation

[Ch. 18

natural roll frequency. In practice it may be found that some of the settings are
marginally preferable to others but the benefits to be gained are usually small. For
example, setting K 1 to zero will ensure that the fins do not waste energy by
attempting to correct a steady list or try to hold the ship upright in a turn. The steady
roll moment available at zero frequency is so small (see Fig. 18.22) that efforts in this
area are probably doomed to failure. The only appreciable effect will be an increase
in the ship's fuel consumption due to the small increase in resistance. Similarly, high
values of K 3 may lead to excessive stabiliser fin activity at high frequencies and
increased wear, noise and vibration with no noticeable reduction in roll motion.

18.3.6.4 Fin servo transfer function


The fin servo transfer function may be expressed in the form

ex

al

cx0

a1 + a2 s + a3 s2

(18.27)

Putting

we obtain for the amplitude response


!Xo

al

CXno

V[(al- a2 ro~f + ai ro~]

(18.28)

and the phase response is given by


tan(phase) =

-a2 ro

e2

al- a3 roe

(18.29)

The coefficients a1 , a2 and a3 are chosen to match these equations to the measured fin
servo responses.

18.3.6.5 Choosing the overall gain K 0


Having chosen the sensitivities K 1 , K 2 and K 3 to match the control system to the
rolling characteristics of the ship, the next step is to determine the overall gain setting
K 0 . This governs the magnitude of the roll reduction achieved by the stabilisers and
should be chosen so that the ship meets some agreed roll specification in moderately
severe weather conditions. An appropriate specification would be written in the
form:
'The rms roll motion at the worst heading at 20 knots in sea state 7 must not exceed 4
degrees. Fin motions in these conditions should not exceed 25 on more than one

Sec. 18.3]

Active roll stabiliser fins

371

oscillation in ten. Sea state 7 is to be interpreted using the WMO sea state code and
the most probable modal wave period for annual conditions in the North Atlantic.
Cosine squared wave spreading is to be assumed.'
The stabilised roll motion is computed using equations (18. 9)-(18.11) with the fin
incidence a: now given by equations (18.20) and (18.27). Typical results of such a
calculation, taking account of the speed dependent gain, are shown in Fig. 18.24. As

Overall gain KG

Fig. 18.24- Effect of gain on stabiliser performance.

expected the roll motion decreases with increasing overall gain at the expense of
increased fin motion. Using the gains available in the control system specified in
Table 18.1 we find that the roll target is achieved with

K0

=1.25

The requirement to limit the fin activity is introduced to avoid excessive


mechanical wear and possible damage to the fin m_~chanism which will occur if fin
angles greater than the maximum available are continually demanded. Excessive fin

Roll stabilisation

372

[Ch. 18

demands will also lead to cavitation which, in extreme cases, may damage the fins
and will certainly generate noise. The latter may be of particular importance in
warships. It should also be noted that the assumptions of linearity inherent in
equation (18.8) describing the fin lift characteristics will lead to an overestimate of
the stabiliser performance if the fin motion (and hence the lift) is actually limited by
mechanical constraints. In any case the equation will overestimate the lift at large
angles of incidence (see Fig. 2.21).
The specification given above requires the probability of the fin motion amplitude
exceeding CXmax ( = 25) to be no more than 0.1. Equation (17 .28) may be used
to calculate the corresponding maximum allowable rms fin motion. Using Table 17.5
we find that
CXmax
O'o

= 2.15

and the maximum permissible rms fin motion is

The rms fin motion required to meet the roll target in Fig. 18.24 is 12.4. In this
case the stabiliser capacity will need to be enhanced by increasing the fin area, using a
more effective aspect ratio, improving the fin/bilge keel layout to avoid interference
or increasing the number of fins.

18.3.6.6 System stability


Roll stabiliser systems have the potential, like all automatic control systems, of
becoming unstable at certain frequencies. Clearly this possibility must be considered
at the design stage and steps must be taken to prevent it occurring.
Fig. 18.25(a) shows a simplified block diagram of the roll stabiliser system. The
ship is represented by a block with a transfer function G and the entire 'feedback'
network consisting of the gyro, the fin controller, the fin servo and the fins themselves
is represented by the single block with a transfer function H. G and Hare, of course,
complex quantities of the form

G = IGI (cos EG + i sin EG)

(18.30)

where IGI is the amplitude response or gain and EGis the phase response.
Suppose that the system is excited with an input vin and responds with an output
Vout Then the output of the feedback block will be Hv out and the total input to the
ship will be vin - Hvout. So the input and output of the complete closed loop system
are related by
(18.31)

Active roll stabiliser fins

Sec. 18.3]

373

G
Vaut

H Vout

(a) Stabilised system

...

...

G
G

V in

(b) Unstabilised system

Fig. 18.25 -Block diagrams for stabilised and unstabilised systems.

and the transfer function or 'closed loop gain' of the complt:1tff system is given by
Vout
Vin

G
1+G H

=---

(18.32)

GHis the transfer function the system would have if the feedback loop were left open
and is termed the open loop gain. The stability of the closed loop system may be
examined using the Nyquist diagram illustrated in Fig. 18.26. In this diagram the
open loop gain is plotted as a vector of length IGHI and argument BaH where

The location of the end of the vector varies with frequency, moving around the
diagram as the frequency increases. At :lero frequency, in conventional systems with
fins in the middle part of the ship, the phase is zero and the open loop gain vector lies
along the positive real axis. As the frequency increases some phase advance is
introduced by the ship phase r~sponse (see Fig. 18.22) and by the controller. In a well
designed system the phase is, as we have already seen, arranged to be zero at the
natural roll frequency co. 4 and the gain vector again lies along the positive real axis.

374

Roll stabilisation

[Ch. 18

c:co
c

Ol

co

Fig. 18.26- Open loop gain Nyquist diagram.

The gain should be a maximum at this point. At higher frequencies the phase
becomes negative and, in a stable system, the gain steadily diminishes until it
becomes zero at infinite frequency and the locus curve approaches the origin.
If the ship were unstabilised the response to the excitation vin would be G vin as
shown in Fig. 18.25(b). So the roll stabilisers will only reduce the rolling motion at
some particular frequency if

This occurs if the quantity

I1+GHI > 1
Now 11 + G HI is the distance from the point (- 1, 0) to the appropriate point on the
open loop gain locus (see Fig. 18.26). So we can see that the stabilisers will only
reduce the rolling motion at frequencies for which the open loop gain locus lies
outside a circle of unit radius centred at the point (- 1, 0). The roll motion will be
amplified for all frequencies lying within this unit circle. Fig. 18.27(a) shows that this
will always occur at high frequencies for conventional fin locations. For aft mounted
fins with large angles of depression the phase reversal at zero frequency results in the
locus of the open loop gain vector beginning somewhere on the negative real axis as
shown in Fig. 18.27(b). So these installations will always give motion amplification at
low frequencies unless the gain is made zero by setting

Active roll stabiliser fins

Sec. 18.3]

R
(a) Amplification at
high frequencies

375

c::.......___,;

(b) Amplification at high


and low frequencies

.,

'!
(c) Unstable system

Fig. 18.27- Amplification and instability.

Equation (18.32) shows that the system will become unstable (i.e. the amplitude
response will become infinite) when the open loop gain
GH= -1

(18.33)

and this occurs if the gain yector locus passes thro,ugh the point (- 1, 0) on the
negative real axis as shown in Fig. 18.27(c). The phase is then - 180 and the roll
moment due to the fins then enhances the roll moment due to the waves at some

Roll stabilisation

376

[Ch. 18

particular frequency. Any excitation at this frequency will then cause very large fin
motions which will enhance the initial excitation and increase the fin motions still
further. In practice, of course, the reponse will be limited by the mechanical stops,
fin rate limits on the fin servos and fin stall but large undesirable fin oscillations may
still occur.
Clearly we must ensure that the open loop gain locus never passes through the
point (- 1, 0). It is also desirable to avoid approaching it too closely because this will
result in motion amplification even though the motions will be stable. Two commonly used criteria for defining adequate safety margins are the gain and phase
margins defined in Fig. 18.28. The gain margin is defined as

Fig. 18.28- Definition of phase and gain margins.

G
m

open loop gain at

~>oH

(18.34)

= - 180

and the minimum acceptable value of G m is generally taken to be 2 which implies that
the open loop gain GH must not exceed 0.5 when the phase is -180.
The phase margin is defined as
~>m

= 180 + open loop phase when

11 + G

HI = 1

(18.35)

and the minimum acceptable phase margin is 30; 60is regarded as very good
practice.
If the system stability is unsatisfactory it can be improved by reducing the overall
gain K 0 or by choosing different values of the sensitivities K 1 , K 2 and K 3 (but still
satisfying equation (18.25)).

Passive tanks

Sec. 18.4]

377

18.3.6. 7 Active roll stabiliser fin performance

Fig. 18.29 shows the performance of a typical active fin roll stabiliser system with
10

...,
CD

Cl
CD

::s

Cl

c:

'0

"'
..c:

CD

I
I
I
~

....

....

Cl)

....s:

\.

"'

Stabilised
----'-

Cl)

u,
0

U2
10

25

Speed (knots)
Fig. 18.29- Effect of speed on roll stabiliser fin performance for a frigate at worst heading.

speed dependent gain. At very low speeds the fins are comple~~ly ineffective because
the overall gain is set to zero (see Fig. 18.20). At speed U1 the fins are switched on
and a substantial roll reduction is achieved. As the speed is increased to U2 the fins
become progressively more effective and the stabilised roll motion decreases. At
speeds above U2 the gain is reduced and the stabilised roll motion becomes nearly
independent of speed.

18.4 PASSIVE TANKS


18.4.1 General principles and types of tanks
The fluid in a partially filled tank in a ship will slosh backwards and forwards across
the tank as the ship rolls. The shifting weight of the fluid will exert a roll moment on
the ship and, by suitable design, this can be arranged to damp the roll motion. Fig.
18.30 shows the desired motion of the '_Yater in relation to the rolling motion of the
ship and it can be seen that we require the motion of the fluid to lead the roll motion
by 90 so that it is in phase with the roll velocity.
Fig. 18.31 shows some of the types of passive tanks which are in current use. The
simplest is the flume or free surface tank which consists of a rectangular tank running
athwartships. Sometimes a limited control is exerteq,over the motion of the fluid by
installing a restriction or baffle in the centre of the tank.

.....:]

00

.....-----....
...,..

I~ I
(a)

t=O

Maximum roll rate to starboard;


maximum stabilising
moment to port

~I ==(b)

t=

rr

(1h4

Maximum roll to starboard;


zero stabilising moment

Maximum roll rate to port;


maximum stabilising
moment to starboard

"'

(c)t=~

2UJ.4

Maximum roll to port;


zero stabilising
moment

~"'
~

"

Fig. 18.30- Passive tank motion.

g
>-'

00

Passive tanks

Sec. 18.4]

(a) Simple flume


tank

(c)

Simple U-tube
tank

379

(b) Flume tank


with baffle

(d) U-tube with


air duct and throttle
valve

-u -u(e) U-tube with


throttle valve

(f) Active U-tube


with pump
r

Fig. 18.31- Types of passive tank.

rl

U-tube tanks have also been fitted in a number of ships. In this case the free
surface is confined to the two arms of the U-tube which are connected by a horizontal
duct. The tops of the vertical arms may be open to the atmosphere or they may be
connected by a horizontal air duct. In this case a throttle valve may be included to
exert some control over the motions of the fluid. Some designs incorporate a throttle
valve or a pump in the bottom duct.
Passive tanks work well at low speeds but they are not usually as effective as a well
designed active fin system at high speed. For this reason they are often specified for
ships like survey vessels or weather ships which must spend the majority of their time
hove to.
Tanks have the advantage that they~have no moving parts (except perhaps for a
pump or controlled throttle valve) and require little maintenance. They also avoid
the small resistance penalty associated with fins and bilge keels. They take up a
considerable volume of the ship's hull but it may be possible to use the fresh water
supply or some of the fuel oiL.as the working fluid so this loss of volume may not be
serious. The optimum tank position high in the ship ~ften makes access along the ship
difficult.

Roll stabilisation

380

[Ch. 18

A major disadvantage is that the free surface always reduces the metacentric
height so that roll stability will be reduced. As a consequence all tanks amplify roll
motions at low encounter frequencies. In certain circumstances this amplification
may become a serious problem and it may be necessary to immobilise the tank by
draining it or filling it completely. This will invariably take a considerable time and
passive tanks are therefore not suitable for ships which are required to change course
frequently (e.g. warships).
18.4.2 Theory for a U-tube passive tank
In spite of the apparent simplicity of the flume tank, no adequate theory for
predicting its performance has been developed. However, Stigter (1966) has developed a theory for U-tube passive tanks and a modified version of this is described
below.
18.4.2.1

Equation of motion for the fluid in the tank

Fig. 18.32 shows a simple U-tube passive tank. The tank is assumed to consist of two

Port
reservo ir --r----

h,

Datum
fluid
level

2
Star board
rese rvoir

;-

T~;;---------___ ---

h,

rd

-'-

vtl
I . w,

1d

Duct

Yd

ilo
I

--- f-.-

...

1--

Wr

Fig. 18.32- Axis system and tank dimensions.

reservoirs and a connecting duct of constant rectangular cross-sections. The length of


the tank (in the fore/aft direction) isx 1 metres. We require to determine the motions
of the fluid within the tank under the influence of the motions applied to the tank by
the ship. These may be analysed using the axis system shown. The origin 0 is at the
midpoint of the connecting duct and an axis y runs along the duct and up the
reservoirs of the U-tube. The fluid velocity along the positive y direction (up the port

Passive tanks

Sec. 18.4]

381

reservoir) is v metres/second. Three additional axes are defined: y ct has its origin at 0
and runs parallel to the duct, positive to port;yrp andyrs have their origins on the duct
centreline and run parallel to the reservoir walls as shown.
n is the width of the tank perpendicular to the y axis. Note that n is a variable
which has different values hct on the duct and wr on the two reservoirs. It is assumed
that there is no flow in the 'n' direction and the motions of a unit mass (1 tonne) of
fluid in the tank will be governed by a simplified version of Euler's equation (2. 7b):

(18.36)

where Y is the external force per unit mass and p 1 is the mass density of the fluid in the
tank.
Now the duct and the reservoirs are assumed to be of constant cross section so we
may write

av =0

ay

everywhere except at the junctions between the duct and the reservoirs. Neglecting
these corner effects equation (18.36) reduces to

av- y - 1-ap
-

at

Pt

(18.37a)

ay

or, since there are now only two variables,

dv _ y 1 dP
- --dt

(18.37b)

Pt dy

If the difference in the height of the fluid level in the two reservoirs is z metres the
velocity in each reservoir will be

d
vr = dt

(z)2 = 2wi-

metres/second

where 't, which is assumed to be small, is the tank 'angle' defined in Fig. 18.32,
w = wd + wr

metrt:s

and the velocity at any point in the tank is

(18.38)

382

Roll stabilisation

[Ch. 18

(a)

(b)

Fig. 18.33- External forces applied to unit mass in (a) U-tube duct and (b) U-tube reservoirs.

Passive tanks

Sec. 18.4]

w. v.
n

w. w t
2n

v = - - =- -

383

metres/second

(18.39)

The external force per unit mass Y is made up of contributions due to the
accelerations applied to the tank and the frictional forces arising from the losses in
any throttle valve, wall friction etc. Figs 18.33(a) and 18.33(b) show these contributions. They are:
(a) The component of the acceleration due to gravity along the y direction

- g cos

<j> 1

metres/second 2

(b) The acceleration due to the roll acceleration

(c) the component of the local lateral acceleration in the y direction


(1) in the duct
y LA = S2

COS X4

= S2

metres/second 2

(18.40a)

(2) in the reservoirs

=s2 sin x 4 = 0
since both s2 and x 4 are assumed to be small
YLA

(18.40b)

(d) The frictional or damping forces. Although these would be expected to be


proportional to the square of the local velocity it is convenient to assume that the
damping can be linearised and is proportional to the velocity v. If the tank length
x 1 is much greater than the normal dimension nit can be shown that the frictional
force per unit mass is approximately

-qv
kN/tonne
n
where q is a coefficient of resistance to be estimated or determined by
experiment.
Equation (18.37b) then becomes

Roll stabilisation

384

WrW 't q Wr W 't


~ + 2n2

,!.,

..

[Ch. 18

+ g cos 'f'l + r x4 sm cf>2- y LA


-1 dP

= Pt dy

metres/second

(18.41)

We now integrate this equation with respect toy to obtain an equation giving the
motion of the fluid in the tank (in terms of the angle 't) as a function of the pressure
difference at the surface in the two reservoirs. Strictly the integration should proceed
from the surface level in the starboard reservoir (negative y) to the surface level in
the port reservoir (positive y). However, the continually varying fluid levels
introduce complications and we therefore obtain an approximate solution by
integrating between the datum levels in each reservoir. We also assume that the
lateral acceleration s2 does not vary appreciably along y. We obtain

Pt Wr W Il 't
2

Pt q Wr W I2 i

+
2
= Ps- PP

+ Pt g

Pt I4 i4

Pt s2 Is

kN/metre 2

(18.42)

where

I1 -

tank

2
dyd JO -+
dYrs Jh'dy
W
dy- Jw/ -+
2..!1?-_
n - - w/2 hd
-h, Wr
0 Wr - hd

+2hr
Wr

(18.43)

(18.44)

I3 = J

cos

cf>l

tank

I4 =

2
dy = x4 Jw/ dyd - Jo dyrs + Jh, dyrp = w x4 metres
- w/2
-h,
0
(18.45)
w/2

r sin cj> 2 dy=

tank

-w/2

rd dyd +

o
J

2 dYrs +

-h

Jh, w

2 dyrp = W (rd + h)r

'

metres

(18.46)

2
dy = - J

Is= - J
duct

where

tank

dyd = - Jw/
duct

w/2

dyd = - w

(18.47)

implies integration along the y axis from the datum level in the

Passive tanks

Sec. 18.4]

385

starboard reservoir to the datum level in the port reservoir and

is confined to

duct

the duct. The angles <\> 1 and <\> 2 are defined for the duct and the reservoirs in Figs
18.33(a) and 18.33(b).
The hydrostatic pressures at the datum levels in the two reservoirs are

w
Ps= -PP= -p1 g2t

kN/metre 2

(18.48)

relative to atmospheric pressure.


Equation (18.42) may now be expressed as an equation giving the motion of the
tank fluid as a function of the moment applied to the tank fluid by multiplying by the
moment of the area of the reservoirs:

metres3

Using equation (13.8) to calculate the lateral acceleration experienced by the


tank located x 81 metres forward of the centre of gravity we obtain
a~ 2 i

+ a~ 4 i 4 + c~ 4 x 4 + a~ 6 i 6 +an 't + bn i- +en t

= 0

kN metres
(18.49)

where the coefficients are


a~2

= - Q1

kN metres/(radianlsecond2 )

a~ 4 = Q1 (rct +h.)
c~ 4

= Q1g

a~6 = -

kN metres/(radianlsecond 2 )

kN metres/radian

Q1x 81

kN metres/(radian!second 2 )

kN: ffi"etres/(radianlsecond 2 )

b~~ = Q

q w. (

;J +:~)

kN metres/(radian!second)

(18.50)
(18.51)
(18.52)
(18.53)

(18.54)

(18.55)

Roll stabilisation

386

[Ch. 18

en= Qt g = C-r4 kN metres/radian

(18.56)

with

Qt = Pt wr2w

Xt

tonne metres

(18.57)

18.4.2.2 Equations of motion for a ship with a passive stabilising tank

The tank angle 't may be regarded as an additional degree of freedom in the equations
of motion for the ship (8.26)-(8.31). Its effects are taken into account by including
additional terms of the form

(i = 1,6)
in these equations.
Many of these coefficients are zero. In particular, the tank has no effect on the
surge, heave or pitch motions so that
(i = 1; i = 3; i = 5)

Simple physical arguments also demonstrate that no sway force or yaw moments can
be caused by a steady tank angle and that the rate of change of tank angle can have
little appreciable effect. So

(i = 2; i = 6)
In addition it is assumed that the rate of change of tank angle has a negligible
influence on the roll moment so that

The lateral plane equations of motion for a ship stabilised with a passive tank are
then:
sway:

(m + azz) Xz + bzz Xz + a24 i4 + b24 X4 + az6 i6 + bz6 i6 + Cz6 x6


+a 2-r 't=Fw2o sin

roll:

(roe

t+y 2) kN

(18.58)

a42 Xz + b42 Xz + (/44 + a44) i4 + b44 i4 + C44 X4 + a46 i6 + b46

+ C46 x 6 -

[a4-r

't + C4-r 't] = Fw4o sin

(roe

t+ Y4)

kN metres (18.59)

(where the expression inside the square brackets is the tank stabilising moment)

Sec. 18.4]

387

Passive tanks

yaw:

a6z Xz + b6z Xz + a64 i4 + b64 + (/66 + a66) i6 + b66 x6

+ c66 x 6 + a 6~ 't = Fw 6o sin (coe t + y6 ) kN metres

(18.60)

The vertical plane equations remain as for the unstabilised ship (equations (8.23),
(8.25) and (8.27)).
We now derive the tank acceleration coefficients a 2 ~, a 4~ and a 6~. These may be
considered as the sway force, roll and yaw moments required to sustain a tank angle
acceleration of 't = 1 radian/second 2 .
Consider the tank shown in Fig. 18.34. If the tank angle acceleration is 't

(.

in,

tv,

m,

.v,

rd

,._

,.,""

md

- l i d ....

.....

r" . . . _ . .

Fig. 18.34- Fluid accelerations and reactions in a passive stabilising tank.

radians/second 2 the fluid accelerations in the reservoirs and the duct are, by equation
(18.39),

w 't vct= wrhwct 't


vr=T,
2

metres/second 2

and the masses of the fluid in the reservOirs and the duct are

The lateral force which must be applied to the tank to sustain these clockwise
(positive) accelerations is

388

Roll stabilisation

[Ch. 18

kN to starboard
so that
(18.61)
If the tank is located x81 metres forward of the centre of gravity the yaw moment

required to sustain the accelerations is


a6~ 't

= x81

a 2 ~ 't

kN metres to starboard

so that
(18.62)
The roll moment required to maintain the acceleration of the fluid in the reservoirs is
kN metres to port
and the roll moment required to sustain the acceleration of the fluid in the duct is

so that the total roll moment is

and we obtain
(18.63)
Finally we obtain the coefficient c 4~ by considering the moment required to
sustain a steady positive tank angle t. The weight of fluid above the datum level in the
port reservoir in Fig. 18.35 is
kN

Sec. 18.4]

Passive tanks
w

389
w

Ptgw,wx,

~
G

w,

w,

Fig. 18.35- Roll moment due to tank angle.

and a similar weight is displaced from the starboard reservoir. So the applied
moment is
kN metres to port
and
e4~

18.4.3

= Q1 g =en= e~ 4 kN metres/radian

(18.64)

Tank natural frequency and damping

The motion of the fluid in the tank is governed by equation (18.49) which may be
rewritten in the form
(18.65)
This has the same form as the equation governing the behaviour of a second-order
linear damped spring-mass system (equation (6.1)) with the right hand side providing the excitation to the tank from the ship. The natural frequency of the tank is, by
equations (6.8), (18.54) and (18.56),

ro -

*t-

)
+ 2hrhd
~(-anen) -- ~(WrW2ghd

radians/second

(18.66)

[Ch. 18

Roll stabilisation

390

The non-dimensional tank damping or decay coefficient is, from equation (6.12),
(18.67)
The tank decay coefficient may be determined with a simple free decay experiment
on a fixed model of the tank. The model should be of fairly large scale and is
conveniently made in acrylic sheet or some other transparent material so that the
oscillations of the fluid may easily be observed. The tank fluid should be displaced
towards one side of the tank and then released. The subsequent decay of the tank
angle oscillations should be recorded and the decay coefficient estimated from
equation (6.22). The dimensional tank damping coefficient is then given by

bn =

~tQt ~[ g (~ + ~J]
Wr

= 2 ~~

Q 1 g kN metres/(radian/second)
co.t

18.4.4

(18.68)

Design of passive stabilising tanks

18.4.4.1 Basic requirements


Fig. 18.36 shows a block diagram representation of a ship stabilised with a passive
tank. The basic requirement for optimum tank performance is exactly the same as for
active fin roll stabilisers. We again require the open loop phase to be zero so that the
stabilising moment applied by the tank is in exact opposition to the roll moment
excitation applied by the waves. Once again it is impossible to achieve this desirable

Roll
movement
due
to waves

Ship

Roll
Roll m oment j
due to tank

Tank

Fig. 18.36- Block diagram for a ship with a passive stabiliser tank.

Passive tanks

Sec. 18.4]

391

state at all frequencies simultaneously and we choose to optimise the performance at


the natural roll frequency where most of the rolling motion occurs. The roll motion
lags the wave excitation by 90 at the natural roll frequency and we therefore require
the stabilising moment to lead the roll motion by 90 at this frequency. It will also be
desirable to arrange for the stabilising moment to be a maximum at the natural roll
frequency.
The roll moment applied by the tank to the ship may be calculated using
equations (18.49) and (18.59). The algebra is considerably simplified if we neglect
the influence of sway and yaw accelerations. We suppose that the ship is rolling in
regular waves and that the roll motion is given by
x 4 = x40 sin (roe t)

radians

and that the resulting tank motion is

't = 'to sin (roe t + en) radians


The stabilising moment applied to the ship is
F14 = F140 sin (roe t + ~:: 12 )

kN metres

Substituting these expressions in equations (18.49) and (18.59) we obtain the tank

moment response to roll motion

and the phase is given by


(18.70)
showing that the tank momentis always in phase with the tank motion.
Fig. 18.37 shows the tank stabilising moment amplitude and phase responses
calculated from these equations for a tank with the following characteristics:
w = 20 m, wr = 3 m, hr = 5 m, hd = 1 m, ht = 10 m, llt = 0.1

giving a tank natural frequency

ro. 1 = 0.529 radians/second

392

Roll stabilisation

[Ch. 18

20

cCll
"0

15

()l

....E:

Q)

10

01

!}!~
0

2.0

w
Q)

Cll

Cll
Q)

en
Cll
..c

a...

Fig. 18.37- Tank stabilising moment characteristics.

This shows that the stabilising moment is a near maximum at the tank natural
frequency Cth 1 and that the phase at this frequency is

Evidently the optimum tank performance will be assured if we arrange for the
tank and ship roll natural frequencies to be the same. This ensures that the stabilising
moment is a maximum and leads the roll motion by 90 at the natural roll frequency.
18.4.4.2

Effect of tank dimensions on tank natural frequency

Equation (18.66) gives the tank natural frequency as a function of the tank
dimensions hct, w, wr and the depth of the fluid hr. As an example of the use of this

Passive tanks

Sec. 18.4]

393

equation Fig. 18.38 shows the effect of changing tank dimensions and fluid depth on
the tank natural frequency. These graphs are for an initial tank design having the
characteristics given above.
The tank natural frequency decreases with the widths w and wrand increases with
the duct depth hct. However, the natural frequency is quite insensitive to the depth of
fluid hr in the tank. It follows that there is little scope for adjusting the natural
frequency after the tank has been designed and fitted to the ship.

18.4.4.3 Fluid depth and maximum tank angle


In practice hr should be selected to give a datum fluid level halfway up each reservoir.
This gives the greatest possible scope for fluid motion and maximises the available
stabilising moment. The maximum possible tank angle is then given by

(18.71)

18.4.4.4 Maximum stabilising moment


The stabilising moment developed at the natural roll frequency is obtained by setting

!\

=1

in equation (18.69) to give

Ft40 __ Qt g {1- a~4/(Qt g)]


X4o

2 ~~

kN metres/radian

I fl

(18.72)

Now we have seen that the natural frequency is determined by the major tank
dimensions w, wrand hct. So the required peak stabilising moment is best achieved by
choosing appropriate values of the remaining parameters p1, x 0 ~~ and 'ct The
stabilising moment increases with the fluid density p1 and the tank length x1
Fig. 18.39 shows the variation of peak moment per metre length of tank with the
vertical location of the duct relative to the ship's centre of gravity for the specimen
tank characteristics given above. This shows thanhe tank becomes more effective if
it is located high in the ship (i.e. rct is small).

18.4.4.5 Loss ofmetacentric stability


As already mentioned, one penalty of passive stabilising tanks is the inherent loss of
roll stability because of the free surface effect described in Chapter 10. This may be
estimated using equation (18.59). If we consider the roll behaviour of the ship at zero
frequency we may write

1.0

1.0

u
=a

Q)

(J)

Q)

"C

[Ch. 18

Roll stabilisation

394

0.5

0.5

3
0

40

w, (metres)

w (metres)

u
=a

1.0

1.0

u
=a

Q)
(J)

Q)

(J)

0.5

~~]
'

h,

3
J

I
4

I
2

0.5

0.5

1.0

2.0

hd (metres)

h,(metres)

Fig. 18.38- Effect of tank dimensions on tank natural frequency.

c
Q)

E
0
E

20

0
rd (metres)

Fig. 18.39- Effect of tank location on peak stabilising moment.

where F4 is some steady applied roll moment. Now the tank angle is, from equation
(18.49),

= - x4

radians

Sec. 18.4]

Passive tanks

395

and using equations (10.8) and (18.64) we find

m g GMs (1 - JJ-1)

X4

= F4

kN metres

where JJ-1 is the fractional loss in metacentric height:


(18.73)
Clearly this loss of stability is undesirable and !Lt is usually limited to about 0.25.
18.4.4.6 Mass of working fluid
The mass of fluid in the tank is given by

(18.74)
and it is usually found that a satisfactory degree of stabilisation can be achieved if m 1
is ofthe order of 1-5% ofthe ship mass.
18.4.4. 7 Tank damping
The tank damping may be adjusted by installing an obstruction or a throttle valve in
the duct as shown in Fig. 18.31. Fig. 18.40 shows the effect of increasing the tank

Unstabilised

10
Speed 0 knots
Beam waves

Q)

"0

.~

c.
E

"'c.

Q)

Stabilised, ~=0.2,
high tank damping

"iii
Q)

>

~"'

a:

\j '\
2

Stabilised,

~=0.2,

'y ' \ / l o w tank damping


\

' ' '


1.0

1.2

Frequency (radians/second)
_,,_

Fig. 18.40- Typical roll transfer functions showing effect of passive tank damping.

Roll stabilisation

396

[Ch. 18

damping on the roll transfer function for a ship in beam waves at zero speed. With
low tank damping the roll response peak at the natural roll frequency is effectively
eliminated, but this is at the expense of resonant peaks at higher and lower
frequencies. These indicate that the tank will amplify the motions at these frequencies, possibly leading to an overall amplification ofthe roll motion, depending on the
shape of the wave energy spectrum. They may be eliminated, or at least reduced, by
increasing the tank damping. The motion at low frequencies is still amplified but this
is a characteristic common to all passive tanks since the loss in stability ensures that
the 'stabilised' roll motion at zero frequency always exceeds the unstabilised roll
motion.
15
0 knots

Cl

----

Q)

Unstabilised roll

::!:!.
~
Cl

c:

"'

e
rn

200

15
Cl
Q)

::!:!.

20 knots

~
Cl

c:

co

e
rn

200

Probability
0.0 of exceeding
1

-0.1

150

200

Heading (degrees)

Fig. 18.41- Passive tank performance.

Tmax

Sec. 18.4]

Passive tanks

397

18.4.5 Performance of passive stabilising tanks


Fig. 18.41 shows the performance of the tank described earlier fitted to a 4000 tonne
ship. The calculations are for a loss in stability of

and a tank mass of 1.87% of the ship mass. At zero speed the tank gives a useful
reduction in roll motion at all headings. This is because the inherent roll damping due
to the hull, bilge keels and other appendages at low speed is small and the damping
provided by the tank makes a substantial additional contribution. At 20 knots the
hydrodynamic damping is much higher and the contribution provided by the tank is
relatively insignificant. So the tank is unable to achieve a worthwhile reduction in roll
motion.
Fig. 18.41 also shows the penalty of the loss of stability at high speed in following
seas. The encounter frequencies are then very low and the tank amplifies the roll
motion.
Also shown in Fig. 18.41 is the rms tank motion for each speed. Equation (18.71)
gives the maximum permissible tank angle as

The rms tank angles corresponding to various probabilities of exceeding this level
may be estimated from Table 17.5:

Probability of
exceeding 24.2

rms tank argle


(degrees)!

0.1
0.01
0.001

11.3
7.4

6.0

and these are plotted in Fig. 18.41. Evidently the tank motion is sufficient to reach
the tops of the reservoirs and the duct about once in every 100 oscillations on the
worst heading in this particular sea condition. This would be regarded as satisfactory
in practice. A more frequent rate of exceedance would invalidate the calculation
which takes no account of any such limits in the tank's stabilising capacity. This could
be rectified by increasing the height of the reservoirs and the depth h. of the working
fluid.

19
Added resistance and involuntary
speed loss in waves

19.1

INTRODUCTION

The speed a ship can achieve in calm water is governed by its resistance, propeller
efficiency and the power of its engines. In rough weather the resistance may be
changed by the action of the waves and the wind and the resulting change in the load
on the propeller usually reduces the propeller efficiency. The speed the ship can
achieve for a given engine power is usually reduced by these effects. This 'involuntary' speed loss does not often amount to more than two or three knots but may still
result in substantial financial losses for merchant ships.
19.2

SIMPLE THEORY FOR ADDED RESISTANCE IN REGULAR HEAD


WAVES

A ship towed in regular waves will have a fluctuating resistance as illustrated in Fig.
19.1. In head waves the mean value of the resistance will be greater than the calm
Resistance
Resistance in waves

Raw
Calm water resistance

Rc~---------------------------L---

Time

Fig. 19.1--- Resistance in waves.

Sec. 19.2]

Simple theory for added resistance in regular head waves

399

water resistance and the difference may be attributed to the effects of the waves.
The simple theory for this added resistance presented here is based on that
proposed by Gerritsma and Beukelman (1971) and has its origins in the strip theory
described in Chapter 9. We shall confine our attention to long crested head waves
which is generally accepted as the most severe case. Only vertical plane motions then
occur.
Consider the relative motion of a strip located x 81 metres forward of the centre of
gravity. At the water surface this is given by equation (13.14) as:

where s 3 is the absolute motion of the strip given by equation (13.9):

Since we are here concerned with the average relative vertical motion experienced
over the draught of the ship, it is appropriate to take the wave elevation at the local
mean draught D (equation (9.31)) rather than at the surface. In this case the relative
motion becomes
r3

= s3 -

exp(- kD)

metres

(19.1)

The force required to sustain this motion is, by analogy with equation (9.16),

(19.2)

Now the relative motion may be written (equation (13.15))

and the work done by the strip in one complete cycle is then

(19.3)
The total work done by the whole~ship in one encounter period is obtained by
allowing ox 81 to approach zero and integrating equation (19.3) over the hull length:
6

Ls (

E -

1Tffie

J
0

d a33
U dx
1

b33-

81

2
r 30

.,_

dx81

kN metres

(19.4)

400

Added resistance and involuntary speed loss in waves

[Ch. 19

This work must be supplied by the ship's engines as an additional quantity


over that required to drive the ship in calm water. If the total resistance of the ship in
waves is written as

where Rc is the calm water resistance and Raw is the added resistance due to the
waves, the additional work required to drive the ship through one wave length is
E = RawA

kN metres

and the added resistance is

(19.5)
Now the relative motion amplitude r 30 is proportional to the wave amplitude ~ 0
and it therefore follows that the added resistance in regular waves must be
proportional to the square of the wave amplitude. A suitable non-dimensional added
resistance 'transfer function' applicable to all wave amplitudes must therefore
include the wave amplitude squared in the denominator. Fig. 19.2 shows a widely
accepted non dimensional form for plotting the added resistance in regular waves.
These results were obtained by Gerritsma and Beukelman (1971) and show an
encouraging comparison between their predictions and measurements on a model of
a fast cargo ship in regular head waves. The added resistance response peak occurs
when the relative motions are a maximum. In very long waves (low wave frequencies) the relative motions are very small (see Fig. 13.23) and the added resistance
tends to zero. In very short waves the relative motion approaches the wave amplitude
as the absolute motions become negligible. The added resistance is then due to wave
diffraction and reflection and approaches some small but finite value.

19.3

ADDED RESISTANCE IN IRREGULAR HEAD WAVES

Consider the narrow band of frequencies centred on some encounter frequency roe in
the encountered wave energy spectrum shown in Fig. 14.2.
If we replace the wave components in this small range of frequencies by a single
sine wave, the amplitude of the sine wave must be, by analogy with equation (4.13))

and the added resistance due to this single sine wave is

Sec. 19.4]

401

Increase of resistance due to wind

Experiment

--Theory

FN=0.25
L5 =152.5 m

0.8

1.0

Wave frequency ltJ(radians/second)

Fig. 19.2-Typical added resistance response for a fast cargo ship in regular head waves. (After
Gerritsma and Beukelman (1971).)

where
= Raw

s6

kN/metre2

(19.6)

is a dimensional added resistance response function. The total added resistance due
to all the wave components in the encountered wave spectrum is obtained by
allowing <>roe to approach zero and integrating to give

(19.7)
Fig. 19.3 shows the results of some typical calculations of the resistance in
irregular waves for a frigate. The resistance rises rapidly with significant wave height,
and the greatest increase relative to the calm water resistance occurs at low speed.

19.4 INCREASE OF RESISTANCE DUE TO WIND


Part of the resistance of the ship in calm water is accounted for by the aerodynamic
drag of the superstructure apd the above water part of the hull. It is customary to
express this drag in the form

Added resistance and involuntary speed loss in waves

402

---../

Waves and wind

[Ch. 19

/
/

/
Q)
()

.""'
Vl

600

o'"'

~'+S'
'].:

Q)

cc

12

0
Significant wave height (metres)

Fig. 19.3 -Resistance in head waves for a frigate.

(19.8)
where
D c is the drag force in kN
PA is the density of air in tonnes/metre 3
U is the speed of the ship in metres/second
As is the maximum cross-section area of the superstructure and above water part of
the hull in metres2 .

The drag coefficient C0 may be determined from wind tunnel tests of a waterline
model of the ship.
Waves are generally accompanied by wind and this increases the aerodynamic
drag to

Sec. 19.5]

Propeller characteristics

403

and the additional drag due to the ambient wind Uw is


(19.9)
There is of course no direct relationship between the wind and the sea state, but
estimates of the aerodynamic drag associated with a particular sea state may be
obtained by calculating the drag for the most probable wind speeds given in Table
5.3. Fig. 19.3 shows the additional drag calculated in this way for a frigate. The
contribution from the wind is quite small compared with the increase in resistance
due to the waves.
19.5

PROPELLER CHARACTERISTICS

The speed attained by the ship for a given resistance depends on the hydrodynamic
characteristics of the propeller. Fig. 19.4 shows a typical set of these characteristics

y
/

Rough water
self-propulsion
0.05

Ko

0.2

0.1

I)

0.04

0.8

0.03

0.6

0.02

0.4

O.ol
I
w
I

Jc
I

0.6

1.0

Advance coefficient J
Fig. 19.4- Typical propeller characteristics.

Added resistance and involuntary speed loss in waves

404

[Ch. 19

obtained from cavitation tunnel tests on a model propeller. The diagram shows, in
non-dimensional form, the thrust developed and the torque absorbed by the
propeller as a function of the advance coefficient. The advance coefficient is a
measure of the 'slip' of the propeller and is defined by

=~

(19.10)

Nd

where
UP is the mean velocity of the flow through the propeller disc in metres/second
N is the number of propeller revolutions/second
dis the propeller diameter in metres.
The velocity through the propeller disc is somewhat less than the forward speed
of the ship because of the effects of the boundary layer on the hull. The two velocities
are related by the Taylor wake fraction which is defined as

WT

u-u...~::P
= __

(19.11)

from which we find that the mean velocity through the propeller disc is

up =

U(1-wT)

metres/second

(19.12)

The Taylor wake fraction is usually of the order of 10% and may be measured in
suitable model experiments.
Combining equations (19 .10) and (19 .12) we see that the advance coefficient may
also be written as

(19.13)

The thrust and torque coefficients are defined as

KT

Ko

T
pNzd4
Q

pNzd5

where Tis the thrust in kN and Q is the torque in kN metres.


The efficiency of the propeller is defined as the ratio

(19.14)

(19.15)

Propeller characteristics

Sec. 19.5]

, =

power delivered
power absorbed

TU(1-wT)
= --'----=-

2TTNQ

405

(19.16)

At high values of advance coefficient the propeller is turning very slowly in


relation to the forward speed of the ship and the thrust developed and the torque
absorbed are both small. The thrust will be less than the resistance and the ship will
slow down. At low values of the advance coefficient the propeller revolutions are
high in relation to the forward speed and the thrust and torque coefficients are both
large. The thrust will then exceed the resistance and the ship will accelerate.
At some intermediate value of the advance coefficient the thrust will equal the
ship resistance and the ship speed will be maintained. This condition determines the
advance coefficient for 'self propulsion'.
The presence of the propeller augments the ship resistance by a small amount a
so that the effective resistance for the self propulsion calculation for calm water is

The total thrust required at self-propulsion in calm water is then

Rc(1 + a) kN

(19.17)

giving a thrust coefficient


(19.18)
where Cc is a constant for a given speed:
(19.19)
Equation (19.18) defines the relationship required between KT and J to propel the
ship at speed U metres/second in calm water. Of all the possible combinations of KT
and J only one can be provided by the propeller. This is determined by the
intersection of equation (19.18) and the KT versus] curve as shown in Fig. 19.4. The
resulting self-propulsion advance coefficient lc determines the propeller revolutions
required to drive the shiJ? at the chosen speed in calm water:

Nc =

U(1- wT)

fed

revolutions/second

The power required to achieve these revolutions is given by

(19.20)

Added resistance and involuntary speed loss in waves

406

[Ch. 19

(19.21)
The characteristics of the propeller are determined for the steady flow conditions
experienced in calm water. In rough weather the waves and the motions of the ship
will cause considerable fluctuations in the flow around the propeller but it is generally
assumed that the operating characteristics of the propeller, the Taylor wake fraction
and the resistance augment will remain unchanged from their calm water values. In
this case the self-propulsion advance coefficient and the power required to drive the
ship at a given speed in waves may be determined in exactly the same way as for calm
water. The resistance is now given by

(Rc + Raw)(1 +a)+ Daw kN

(19.22)

and the required relationship between KT and J is given by


(19.23)
where

(Rc + Raw)(1 +a)+ Daw


pUzdz(1- wT)z

(19.24)

The self propulsion advance coefficient Jw for rough water is determined by the
intersection of equation (19.23) with the KT versus] curve as shown in Fig. 19.4. The
effect of the added resistance is to reduce the self-propulsion advance coefficient so
that the propeller revolutions required to maintain a given speed become:

N
w

1
U( - wT)
fwd

revolutions/second

(19.25)

The propeller efficiency is reduced and the power required to maintain this speed is
(19.26)

19.6 SPEED LOSS


In the previous section it was shown that the advance coefficient is decreased and the
propeller loading increased in rough weather. The consequent reduction in speed
depends on how the engines respond to this change in load. In general we should
expect the propeller revolutions to fall, although the actual decrease in propeller
speed may be difficult to determine. For the sake of simplicity we shall assume here

Sec. 19.6]

Speed loss

407

that the engine delivers constant power at a given throttle setting regardless of the
load.
The power required to drive the ship/propeller combination at a given speed
in a specified wave system may be calculated using the methods described above. Fig.
19.5 shows the results of such a calculation for the ship whose resistance characteristics are given in Fig. 19.3, fitted with the propeller of Fig. 19.4. As expected the
power rises steeply with forward speed and significant wave height.
It is convenient to cross plot these results in the manner shown in Fig. 19.6. This
shows the power as a function of speed for a given significant wave height. It is then a
simple matter to read off the speed that can be achieved at a given power level for a
number of significant wave heights and plot the results as shown in Fig. 19.7.
The speed loss is quite small at high power. At lower power levels much more
dramatic losses occur and in extreme cases the speed may be reduced to zero.

0
Significant wave height (metres)

Fig. 19.5- Power in head waves for a frigate.

[Ch. 19

Added resistance and involuntary speed loss in waves

408

10m

20

Q)

5
0

c...

Significant wave height

10

10

15

20

25

30

Speed (knots)
Fig. 19.6- Power in head waves for a frigate.

20
(/)

0c

:::.
'0

15

Q)
Q)

0.
(f)

Significant wave height (metres)


Fig. 19.7- Speed loss at constant power for a frigate in head waves.

20
Slamming, deck wetness and
propeller emergence

20.1

INTRODUCTION

The relative motions between the ship and the water surface are generally largest at
the ends of the ship. In high waves the motions may be so large that the forefoot and
propeller are exposed and the deck submerged. This occurs most frequently at high
speed in head waves although it is not unknown in <;~ther conditions.
The re-entry of the keel after emergence may result in a substantial impact or
'slam' as the ship's bottom strikes the water surface. Ships with heavily flared bows
may also experience similar, but less severe, impacts under the bow flare even when
there is no keel emergence. These slam impacts may be se'Vtjfe enough to cause local
structural damage to the ship's plating. In extreme cases the loading may be sufficient
to distort the ship's hull permamently and some ships are believed to have broken up
following slamming. Even moderate slamming will cause the hull to vibrate at its
natural frequency (generally of the order of a few cycles per second) and the resulting
fatigue loading will reduce the life of the hull. The vibration following a slam is called
'whipping' and often provides the captain with his first indication that a slam has
occurred.
Deck wetness may occur anywhere along the length of the ship, particularly
where the freeboard is low. However, the most severe deck wetness generally
occurs, like slamming, at the bow at high speed in head waves. In these conditions the
forward speed of the ship accentuates the effects of the water shipped onto the
foredeck, and damage to deck fittings and cargo may occur. Any crew or passengers
on deck may be injured or washed overboard. In extreme cases the ship might even
capsize and sink due to the weight of water taken on board.
Propeller racing will begin to occur when the upper tips of the blades emerge from
the sea surface. The sudden reduction and subsequent increase of torque loading as
the propeller becomes fully Ubmerged again may damage the engine and propeller

shaft or even the propeller itself.

Slamming, deck wetness and propeller emergence

410

[Ch. 20

Clearly these phenomena are undesirable and the prudent captain will try to
avoid them if possible. Since they generally become more severe at high speed they
impose an effective limit on the ship speed in rough weather, especially in head
waves. This aspect is discussed in Chapter 22.

20.2 PROBABILITY OF OCCURRENCE


The probability of occurrence of slamming, deck wetness and propeller emergence
are essentially dependent on the probability of the local relative motion exceeding
the draught, freeboard and the depth of the upper tips of the propeller blades.
Fig. 20.1 shows how the ship adopts a running trim and creates a wave system
F

-7-

*
L~------..-..__,.___~
"f
Xs1M

D,

Designwaterlineat
zero speed

Dpe

Actual running
waterline

r,

Fig. 20.1- Draught and freeboard at forward speed.

which depends on the speed. This gives a steady relative motion r3 even when the
ship is running in calm water. If we assume that this datum relative motion is
unchanged when the ship is in waves, we may regard it as a change to the draught,
freeboard, etc. Thus the effective draught to the keel becomes

Dke

= D + r3 metres

(20.1)

the effective freeboard is

(20.2)

Probability of occurrence

Sec. 20.2]

411

and the effective depth of the tips of the upper propeller blades is
(20.3)
where f 3 is of course determined at the appropriate location on the ship.
The notional relative motion in waves is obtained by subtracting the wave
depression from the absolute motion (see equation 13.14). As the hull dips into the
water the increasing submerged volume causes a local 'swell-up' of the water surface.
The effect disappears as the hull rises. This enhances the relative motion over and
above the notional value. We define a swell up coefficient as

C
s

actual relative motion amplitude


notional relative motion amplitude

(20.4)

and it is found that C5 is a function of hull form, location on the hull, speed and wave
length. At the time of writing no universally accepted method of calculating C 5 has
been developed, but it has been measured in model experiments by several authors.
Probably the most comprehensive set of experiment data was published by Blok and
Huisman (1985). A selection of their results for a small frigate in head waves of
length equal to the ship is shown in Fig. 20.2.
Referring to equation (17.26), the probability . of the local relative motion
exceeding the effective draught (i.e. the probability of keel emergence) is

p ke = exp ( -

2 Ct mo)
Dke

(20.5)

where m0 is the variance of the notional relative motion at the appropriate location
on the ship. A similar expression may be written for the probability of deck
submergence

1 p2 )

Pds = exp ( - 2 Ct emo

(20.6)

and the probability of propeller emergence is

p pe - exp

(-!~)
2 Ct n:Jo

The average period of the peaks may be calculated from equation 4.27 as

(20.7)

Slamming, deck wetness and propeller emergence

412

[Ch. 20

<.J

c
Q)

u
~
Q)
0

(.)

0.
::J

a:;

(/)

0.5

Fig. 20.2- Swell-up coefficient. (After Blok and Huisman (1985).)

(20.8)

where m2 and m 4 are the variances of the notional relative motion velocity and
acceleration at the appropriate location on the ship. The average number of keel
emergences, deck submergences and propeller emergences per hour are then
N

-3600

per hour

(20.9)

per hour

(20.10)

= 3600T p pe per hour

(20.11)

ke-

Pke

N ds- 3600
T p ds
p

pe

Sec. 20.3]

Slamming

413

20.3 SLAMMING
20.3.1 Introduction
Fig. 20.3 shows a frigate at high speed in rough weather. The relative motion is
sufficient to expose a considerable length of the keel and a slam is clearly imminent.
The ship is shrouded in spray from a deck wetting which has just occurred and much
of the water that was shipped is pouring down the sides to return to the sea beneath

Fig. 20.3- A frigate slamming in rough weather. (MoD Photo.)

the keel. Some water may also be drawn up under the keel as it emerges from the sea
surface. This can be seen in Figs 20.4(a) and 20.4(b) which show a model ofthe ship
being tested in a towing tank.
The subsequent slam will be into a mixture of air and water which probably helps
to cushion the impact. Slamming impact loads are also affected by the local hull
section shape, the relative velocity at impact, the relative angle between the keel and
the water surface, the local flexibility of the ship's bottom plating and the overall
flexibility of the ship's structure. A complete prediction of slamming phenomena is a
complex task which is beyond the scoge of any existing theory.
Full-scale measurements of slamming pressures at sea are rare because of the
practical difficulties involved, but Fig. 20.5 shows a slamming pressure time history
recorded by Sellars (1972) on the keel of the US Coast Guard vessel Unimak during a
severe slam. The pressure Fises very quickly after the initial impact: indeed it is
suspected that the measured rise is probably alway~limited by the relatively sluggish
response of the pressure transducer.

414

Slamming, deck wetness and propeller emergence

[Ch. 20

(a)

(b)

Fig. 20.4- (a) Model slamming experiment. (b) View from ahead. (MoD Photo.)

20.3.2 Slamming drop tests


The section shapes of most ships in way of the keel may be approximated by simple
wedge sections of the appropriate deadrise angle defined in Fig. 20.6. The slamming

415

Slamming

Sec. 20.3]
1500

(/)

Q)

:::.
~
::J
(/)
(/)

Q)

a:

0.1

0.2

0.3

Time (seconds)

Fig. 20.5- Slamming impact recorded on US Coast Guard vessel Unimak.


(After Sellars (1972).)

Fig. 20.6- Definition of deadrise angle.

characteristics of these sections may therefore be examined by dropping twodimensional wedge sections into wate&.
Dimensional analysis (Chapter 16) suggests that the peak impact pressure
developed on the wedge during a siam will be given in the form

P = CP i p rj

kNftnetre2

(20.12)

416

Slamming, deck wetness and propeller emergence

[Ch. 20

where CP is the slamming pressure coefficient which is a function of the deadrise


angle ~- This relationship has been confirmed in a number of experiments and Fig.
20.7 shows how cp has been found to vary with ~For deadrise angles above about 25 these results agree well with a theoretical

Symbol

Body

Curved
wedge
20 wedge

D.
0

Source

Cone

Hagiwara and
Yuhara (1974)
Chuang (1970)
Chuang and Milne
(1971)

100

'
o".....
'
5

Deadrise angle[-\ (degrees)

Fig. 20.7- Slamming pressure coefficient.

Slamming

Sec. 20.3]

417

result derived by Wagner (1932) in connection with impact loads on seaplane floats:

(20.13)

Slamming pressures estimated using these results are likely to be too high because
two-dimensional drop tests take no account of the effect of the air/water mixture
likely to be below the keel just before impact; nor do they allow for the effects of the
relative 'pitch' angle between the keel and the water surface. All these effects are
likely to decrease the impact pressure. Nevertheless, the results do give a general
indication of the effect of deadrise angle and confirm the basic physics of the
phenomena involved.
20.3.3 Model experiments in waves
The alternative approach of measuring slamming pressures on scale models in waves
has been pursued by a number of workers in the field. However, scale effects are
likely to be important and it is always difficult to measure the relative velocity at
impact.
Ochi (1964), in a classic paper, described measurements of slamming pressures
on a model of a merchant ship in waves. Although he was unable to measure the
impact velocities directly, he confirmed the relationship given in equation (20 .12) but
found that no appreciable slamming impacts occurred if the relative velocity was less
than a certain value. According to Ochi this critical velocity was about 3.7 metres/
second for a 161 metre ship. Assuming that the critical velocity obeys the Froude
r rl
scaling law' it may be expressed as

r crit = 0.093 V(g L


3

5)

metres/second

(20.14)

Ochi concluded that a slam will occur at a particular location if


(a) the relative motion exceeds the local effective draught Dke
(b) the relative velocity at impact exceeds r3 crit.
Now if the relative motion is considered independent of the relative velocity, the
probability of the relative motion exceeding the effective draught at the same time as
the relative velocity exceeds r3 crit is given by

Pslam

D~e
rlcrit )
= exp ( - 2C?mo- 2C?mz

(20.15)

(see equation (17.35)). m 0 and m 2 in this formula are the variances of the notional

Slamming, deck wetness and propeller emergence

418

[Ch. 20

relative motion and velocity.


The slamming frequency is
N

_ 3600 Pslam
T

slam-

per hour

(20.16)

20.3.4 Estimation of slamming pressures during a typical severe slam


The time history of the relative motion during the short period encompassing an
impact will vary from slam to slam. This makes determination of the severity of the
event difficult, but some progress can be made if we choose to approximate the
relative motion for this short period by
r3 = r30 sin (ror3t + &r)
3

metres

(20.17)

as shown in Fig. 20:8. The relative motion at impact is then

t=O

3 r3

Impact

Keel emerges

Fig. 20.8- Relative motion time history.

- Dke = r 30 sin &r3 metres

(20.18)

Or3 = sin - l

(20.19)

so that

( -

Dke)

radians

r3o

The relative velocity during the motion cycle is given by

Slamming

Sec. 20.3]

r = ror r
3

30

cos (ror3t + Or3 )

419

metres/second

(20.20)

and the relative velocity at impact is given by setting t = 0:


(20.21)
It remains to determine a suitable value for the relative motion amplitude r 30 and
the motion frequency ror3 The probability of the relative motion exceeding a peak
value r 30 is, from equation (20.5),

rio- )
P=exp ( -1- 2 C82 m0

(20.22)

and this peak relative motion will occur once inN oscillations, where

(20.23)

where THis an arbitrary sample period. Hence the relative motion amplitude which
will be exceeded once in TH seconds is

r 30 = Cs cr0

~[

2 loge ( T

~:J J

metres

(20.24)

where cr0 is the rms notional relative motion.


The frequency ror3 is assumed to be the same as the average frequency of the
relative motion peaks (see Equation (4.27)):

ro r3-

-yf(mmz
4

radians/second

(20.25)

Using these equations we may now estimate the peak slamming pressures which
are likely to be exceeded once in T H seconds. Fig. 20.9 shows the results of a
specimen calculation for a frigate in head seas with T Hset at 900 seconds. Fig. 20.9(a)
shows the deadrise angle at the keel and the corresponding slamming pressure
coefficient CP given by equation (20.13). The rms relative motion and velocity for the
chosen speed and wave conditions calculated by strip theory are shown in Fig.
20.9(b). Fig. 20.9(c) shows the derived relative motion amplitude and impact
velocity given by equations (20~24) and (20.21 ). These represent the worst conditions
likely to occur in a period of900 seconds. Note that the impact velocity is a maximum
at the forward end of the ship and falls to zero at some location where keel emergence

420

[Ch. 20

Slamming, deck wetness and propeller emergence


60

60

(a)

01
QJ

~
QJ

Ol

:::J
<J)

40

<Jl+-'
QJC

~QJ

c..-u

co

0);,;:::

QJ

C"'"
- QJ

co

EO
Eo

co
(jj

'0
QJ

,;

,....._..Relative velocity
/

---~--------

Critical velocity

400 (d)

200

Fig. 20.9- Slamming pressure calculation for a frigate in head waves; TH = 900 seconds.

is unlikely. Some of the predicted impact velocities are less than the critical velocity
calculated according to equation (20 .14). It is assumed that no slamming will occur at
these stations.

Sec. 20.4]

Deck wetness

421

Finally Fig. 20.9(d) shows the slamming pressure calculated according to equation (20.12). The pressure is set to zero where the impact velocity is less than the
critical velocity.

20.4 DECK WETNESS


Fig. 20.10 shows a frigate experiencing deck wetness at high speed in rough seas. The
bow is buried in the sea and has thrown a corona of solid water and spray high into the
air. The forward speed of the ship will ensure that some of this water comes onto the
deck although much of it will be cast aside by a well designed bow. The remnants of a
previous wetting can be seen surrounding the ship along the entire length of the hull
and as high as the top of the funnel.

Fig. 20.10- Deck wetness on a frigate. (MoD Photo.)

Analytical prediction of deck wetness frequency and severity is impossible at the


time of writing and seems likely to remain so in the near future. Model experiments
can give useful information but even here the results should be treated with caution.
Shipping 'solid' water ('green seas') is believed to be modelled correctly in a
conventional Froude scaled experiment, but it is clear from visual observations that
the formation of spray is noi.modelled properly. This is because the surface tension
which governs the size of the spray droplets is incorrectly scaled. The model spray
droplets are then much too large and would scale to the size of footballs on the ship.

422
20.5

Slamming, deck wetness and propeller emergence

[Ch.20

FREEBOARD EXCEEDANCE

In considering the frequency of deck wetness it is necessary to distinguish between


occasions when the water rises above the level of the deck but does not come on
board, and true deck wetness where water is actually taken onto the weather deck of
the ship. At low speed freeboard exceedance will almost always be accompanied by
deck wetness, but at higher speeds a well designed bow can throw the water up and
away from the ship as shown in Fig. 20.11. Freeboard exceedance therefore does not
necessarily result in deck wetness, but all deck wettings must be preceded by a
freeboard exceedance.

Fig. 20.11- Freeboard exceedance without deck wetness. (MoD Photo.)

It is at present impossible to calculate the frequency of deck wetness proper, but


some idea of the characteristics of a design can be obtained by calculating the
probability and frequency of freeboard exceedance using equations (20.6) and
(20.10).
Fig. 20.12 shows the results of some calculations of freeboard exceedance using
these equations in head waves. These show how the notional relative motion
increases towards the bow. The swell up effect further amplifies the relative motion
but gives a maximum value some distance abaft the stem.
The effective freeboard is reduced by the bow wave and the running trim of the

Freeboard exceedance

Sec. 20.5]

423

/
Actual relative motion I

E.

"

...... \

0
-~

Q)

>
.;::;

"'
~
II)

0.5

10

E
"2

"'0
..0

Sheer line

Q)
Q)

U:
OWL

0.1

Q)

c 0.4
._co
()

o-o

>Q)

~
:=x
;!:::::

..OQJ

"'-o 0.2

..o~

Oro
~

C....o
Q)

.....~

0
Xs1M/L,

Fig. 20.12 -

Calculation of probability of freeboard exceedance for a frigate at 20 knots;


H 113 = 5.5 m, T0 = 12.4 sec.

ship. Equation 20.6 then gives the probability of freeboard exceedance which is
typically a maximum some way abaft th~ stem.
While this result is typical it will generally overestimate the actual frequency of
deck wetness since most of the freeboard exceedances which occur at this location
will not result in water coming onto the foredeck. Fig. 20.11 shows this pheneomenon very well. Lloyd, Salsich and Zseleczky (1986), in an extensive series of model
experiments, found that the observed frequency of geck wetness was most closely

424

Slamming, deck wetness and propeller emergence

[Ch. 20

correlated with freeboard exceedances at the stem head. In other words a deck
wetting is almost always the result of a freeboard exceedance at the stem head and
freeboard exceedances elsewhere do not result in wetness unless they are also
accompanied by an exceedance at the stem head. It follows that the frequency of
freeboard exceedance at the stem head will probably give a reasonably accurate
estimate of the true deck wetness frequency at least at high speed in head waves.
20.6 EFFECT OF BOW SHAPE
It seems obvious that deck wetness frequency and severity must be affected by the
above water form of the bow. It is therefore suprising to find that there is no universal

agreement on the effects of features like flare, stem rake or knuckles. The available
objective experimental evidence is confused and contradictory and remains the
subject of active research.
High freeboard is the only characteristic which is universally agreed to have a
beneficial effect on deck wetness.

21
Effects of ship motions on
passengers and crew

21.1

INTRODUCTION

Ship motions have two undesirable effects on the people within the ship. They cause
motion sickness and also make it more difficult to move in a controlled and coherent
manner so that the performance of everyday tasks is impaired.
The balance organs located in the inner ear qm detect changes of both the
magnitude and direction of the apparent gravitational acceleration as well as angular
accelerations. Excessive stimulation of these organs will, in most individuals, result
in motion sickness. The condition will to some extent be alleviated if the accelerations are confirmed by visual cues from the eyes. Thus~~ ride on a fairground
switchback railway can be enjoyable and exciting even if large accelerations are
experienced. The same accelerations experienced by a blil;1dfolded rider would
almost certainly result in quite distressing motion sickness. In the same way it is
possible to stimulate motion sickness without any motions being present at all. This
can be done, for some individuals, by showing them a film of a violent fairground
ride.
On board ship it follows that motion sickness is most likely to occur if passengers
or crew are confined below decks so that they cannot see the horizon. Bittner and
Guignard (1985) also showed that motion sickness can be exacerbated by facing
diagonally across the ship. Fore and aft or athwartships seating is to be preferred.
Other factors which may promote seasickness are anxiety, fatigue, hunger, smells
(particularly cooking and fuel oil), greasy food, reading and carbonated or alcoholic
drinks. Nieuwenhuijsen (1958) found that women and young children in the liner SS
Masdam were more susceptible to sea sickness than men. Elderly people were
generally less affected than people of middle age.
Fortunately the symptoms of seasickness usually disappear after a few days. Fig.
21.1 shows this effect in grapj;lical form, based on a study of seasickness in the Royal
Navy. Nevertheless, seasickness remains a deterrent to travel by sea for many people

Effects of ship motions on passengers and crew

426

[Ch. 21

30

CD

CD
""0

c;
c

20 I-

10 I-

Ul
Ul

CD

-"'
u

u;
c

.'2

Day of cruise
Fig. 21.1- Motion sickness incidence: effect ')f acclimatisation. (After Walters (1964).)

and an inconvenience to sailors. This has led to a considerable research effort in


developing drugs to alleviate the symptoms. Rather less effort has been devoted to
establishing the precise nature of the relationship between motions and seasickness.
Motions also impair the ability to work effectively even when there are no
problems with seasickness. Moving around the ship becomes more difficult and the
prudent sailor or passenger will always hang on to some suitable anchorage to
minimise his chances of injury in severe conditions. The old adage 'one hand for the
ship and one for yourself' is sound advice. Even in more moderate conditions
performance at tasks requiring good hand/eye coordination, such as tracking targets
on a radar screen may be affected.

21.2

MOTION SICKNESS INCIDENCE

The principal cause of motion sickness in an individual is believed to be the vertical


acceleration experienced at his locality in the ship. Other motions might, if sufficiently high, also cause motion sickness: but in conventional ships these are usually
too small to offer any significant additional stimulation.
Determining the motion sickness likely to be experienced by an individual
subjected to some random motion response on board ship is a difficult problem.
Individuals differ in their susceptibility to motions so it is immediately clear that
some kind of statistical approach is required with a large number of subjects tested.
An individual's response may also vary from day to day depending on the contributory factors listed in Section 21.1. In particular a person who has a job to do is much
less likely to suffer badly from seasickness than one who has nothing better to do than
to contemplate the agonies of life at sea.
O'Hanlon and McCauley (1974) measured the motion sickness response of over
300 American male college student paid volunteers. None of the students had any
recent acclimatisation to motions. They were tested in pairs in a ship motion
simulator which was capable of driving the small enclosed test cabin through a
vertical sinusoidal motion with amplitudes up to 3.35 metres. The cabin had no

Motion sickness incidence

Sec. 21.2]

427

windows so that the subjects could not receive any visual motion cues and their only
task was to monitor their state of nausea by pressing buttons on a control panel. The
experiments lasted for up to two hours or until the subjects vomited.
O'Hanlon and McCauley found that the 'motion sickness incidence' (the percentage of subjects who vomited within two hours) could be expressed in the form

(21.1)
where

ls3 l is the vertical acceleration averaged over half a motion cycle and
J.LMSI

-0.819 + 2.32(log 10 ro~)

(21.2)

Equation (21.1) may be evaluated with the help of Table 17.3 and is plotted in
Fig. 21.2. This shows how MSI increases with acceleration and is most severe at a

100r---------T---------,---------~

QJ

QJ

"'0
"(3

"'"'
c
QJ

-"

;;;
c
0

."
2:

2
<rle

(radians/second)

Fig. 21.2- Motion sickness incidence. (After O'Hanlon and McCauley (1974).)

frequency of about 1.07 radians/second. This frequency is unfortunately close to the


average frequencies of vertical motions for many ships and this explains why motion
sickness is such a common pwblem at sea.
Application of these results to the real life enviroi},ment of a ship in rough weather

Effects of ship motions on passengers and crew

428

[Ch. 21

requires us to make assumptions about the relationship between the random motions
of the ship and the sinusoidal motions of the simulator. Consider an irregular vertical
acceleration time history measured on board a ship. If we assume that the accelerations are distributed according to the Gaussian probability density function then
the average modulus of the acceleration is given by equation (17.23):
(21.3)
where m 4 is the variance of the vertical acceleration. The average period of the
acceleration peaks is, by inference from equation (4.27),

TP =

2rr

~ (::)

seconds

where

is the variance of the rate of change of acceleration and

is the variance of the second derivative of the acceleration (see equation (4.24)).
These integrals must be evaluated using numerical methods. In practice it may be
difficult to achieve a satisfactory result because the high powers of me will ensure that
the integrals will not converge until the frequency is very high. The final result will
therefore depend on the accurate estimation of the very small motion spectral
ordinates at these very high frequencies. In practice it is better to assume that the
average period of the acceleration peaks is the same as the average period of the
displacement peaks

(21.4)
so the average frequency of the motion may be taken as

~e
UJ

--

'J/(mmz4)

radians/second

(21.5)

Subjective motion

Sec. 21.3]

429

These approximations allow the MSI formula to be used to make an estimate of


the proportion of people who will suffer from seasickness in a given set of conditions
at sea. The estimate may not be very accurate because of the difficulty of allowing for
the many secondary factors which are involved in motion sickness and the various
assumptions made above. Nevertheless, the technique may be used to give some
indication, at least in a comparative sense, of the ride comfort of ships in rough
weather.
As an example Fig. 21.3 shows a calculation of MSI for a passenger ferry at 10
knots in head waves. The MSI is highest at the bow and falls to a minimum a little way
abaft midships. In practice the high MSI at the bow may be immaterial because there
will probably be no passengers or crew at that location. A better impression of the
MSI actually suffered by the occupants of the ship may be obtained by applying a
weighting function to represent the distribution of the occupied spaces within the
ship. The weighted average MSI is then given by

MSI

JMSIWdxBlM
JWdxBlM

~----------

(21.6)

where the integrals are evaluated over the length of the ship. Fig. 21.3 shows a simple
weighting function giving equal weight to all the occupants of the ship over the length
of the passenger accommodation. For this example the weighted average MSI is
1

MSI = 7%

rl

21.3 SUBJECTIVE MOTION

An experienced, well motivated and acclimatised crew will not suffer unduly from
seasickness but may still find that vertical ship motions will inhibit their ability to
work effectively. No satisfactory method of properly estimating these effects has yet
been developed but one technique which has been used is based on some experiments by Shoenberger (1975).
He subjected eight experienced US Air Force pilots to vertical sinusoidal motions
in a simple oscillating chair capable of amplitudes up to 1.5 metres. The pilots
wen~ blindfolded to remove any visual motion cues. After some preliminary
experiments they were subjected to ~a 'standard' reference motion of 0.6g at
1.0 Hz. This motion was assigned value of 10 on an arbitrary 'subjective motion'
(SM) scale. The frequency and amplitude of the motions were then changed and each
subject was asked to assess the severity of the new motion in relation to the standard.
Thus a motion which felt twice as severe was assigned an SM of 20 and one which felt
half as severe was assigned a value of 5.
The subjects were generally able to make their assessments within a minute and

430

Effects of ship motions on passengers and crew

[Ch.21

1.5
c
0

.;o

"'Q;

a;_
UN

""
"'"'
-"'
"'uE
t:>"'

"'
0

.~S:

~~

.s::o

"'~
-"

s:"'"'

::~ i : : : : : : i : ~
0

25

"'

Ill-

~~

20

t5-;
-"
"'c
CCV

.;::o-c
u

oc
:2'-

20

~~
.s::~
.2'UJ

~:2

0
-0.5

Stern

0
Xs1M/Ls

0.5

Bow

Fig. 21.3- MSI calculation for a passenger ferry at 10 knots in head waves;
H113 : 5.5 m, T0 = 12.4 sec.

none suffered from motion sickness during the experiments. Shoenberger obtained
remarkably consistent results which could be expressed in the form

(21.7)

Subjective motion

Sec. 21.3]

431

where A is a parameter which is a function of frequency. Fig. 21.4 shows the


relationship obtained from Shoenberger's results. Apparently humans are least
sensitive to motions at frequencies around 6 radians/second. Sensitivity is enhanced
at both higher and lower frequencies.
The equation
(21.8)
fits the experiment data well and exhibits a maximum at roe = 1.07 radians/second,
corresponding to the most sensitive frequency for MSI.
These experiments were, like the MSI experiments, conducted using sinusoidal
oscillations and it is again necessary to devise some suitable means of applying the
results to the irregular motion environment found on board ship. Lloyd and Andrew
(1977) suggested setting
(21.9)
and the appropriate frequency roe may be obtained from equation (21.5).
Fig. 21.5 shows the relationship between SM and the rms vertical acceleration
according to equation (21. 7) for the worst frequency roe = 1.07 radians/second. Also
shown are semantic descriptions of the resulting motion environment based on the
experience of rough weather trials in two frigates described by Andrew and Lloyd
(1981).
Fig. 21.6 shows an example of a calculation of SM for a frigate in head waves.
The SM reaches 34 (intolerable) at the bow but is only abo1;1~e (serious) some way
abaft midships. Again it is appropriate to apply a weighting function to represent the
relative importance of different parts of the ship. In this case the importance of the
motions in sensitive parts of the ship has been recognised by allocating a weight of 1.5
to the flight deck at the stern and to the bridge and operations room forward of
midships. The weighted mean SM is given by

SM WdxBlM

SM

= .:....___ _ __

(21.10)

where the integrals are evaluated over tne length of the ship. For the example shown
the weighted mean SM is
SM

= 14

432

[Ch. 21

Effects of ship motions on passengers and crew

80r---------,---------,---------,
A

10
P>e

Experiment data

20

30

(radians/second)

Fig. 21.4- Subjective motion parameter A. (After Schoenberger (1975).)

30~----------~----------~----------~

Intolerable

20
c

.9

-----

Hazardous

0
E

-- -- --

Q)

>

Severe : necessary
to 'hang on' all
the time

(.)

Q)

:.0
::J

(/)

---

10

rms vertical acceleration (metres/seconds)

Fig. 21.5- Subjective magnitude and vertical acceleration at 1.07 radians/second.

Lateral force estimator and motion induced interruptions

Sec. 21.4]

433

c:

0
-:;::;

.,

E!
a;_

<.><>
.,.,
(,)N

-~
~E

".,
>
(I)

0~_.--~~~~--._~--~--~~--~

40~~--~~--,-~--~--~~--~~

.,

.2: c:

-o
<>..,_

-0

-JE
en

:2:

en
-o

20

2l

..c:

a;

10

5:
0.5

Stern

Xs1M/Ls

Bow

Fig. 21.6- Subjective motion calculation for a frigate at 20 knots in head waves;
H 113 = 5.5 m, T0 = 12.4 sec.

21.4 LATERAL FORCE ESTIMATOR AND MOTION INDUCED


INTERRUPTIONS
Chapter 13 described the calculation of the lateral force estimator (LFE) which is the
lateral acceleration perceived in the plane of the ship's deck by an object or a person.
It is this acceleration which is responsible for making objects topple or slide across
the deck and people lose their balance and stumbl~,,
Baitis, Woolaver and Beck (1983) defined a 'motion-induced interruption' (Mil)

434

[Ch. 21

Effects of ship motions on passengers and crew

as an occasion when a crewman would have to stop working at his current task and
hold on to some convenient anchorage to prevent loss of balance. They proposed a
relationship between LFE and the frequency of Mil experienced on the flight deck of
a destroyer. This is shown in somewhat modified form in Fig. 21.7.

"'"'

c-

Extremely
hazardous

U;
c
0

g_
E
Q)

c
-a

Severe

Q)
()

:::J

-a
c

Serious

c0

' rms lateral force estimator


(metres/second 2 )

Fig. 21.7- Motion-induced interruptions; natural roll frequency 0.59 rad/sec. (After Baitis,
Woolaver and Beck (1985).)

Baitis's estimates were for a ship with a natural roll frequency of about 0.59
radians/second. Since roll makes a dominant contribution to LFE it follows that the
number of motion-induced interruptions per minute will be increased (for the same
rms LFE) if the natural roll frequency is increased. An approximate correction for
ships with different natural roll frequencies may be applied by using

Mil

(1).4

Milo.s9 0.

59

per minute

(21.11)

Figs 21.8 and 21.9 show how LFE and Mil vary in beam waves with location in the
ship: evidently the lowest levels of Mil are found close to the keel at midships.

Sec. 21.4)

Lateral force estimator and motion induced interruptions

"'E

Q)-

~~
Q)

...

om
~E

~-

.2l

.!!!

Ul

E
...

'tl

Q)

::J

'tl

c:

-~
0

.;::;
0

~
0~_.--~_.

-0.5

Stern

__.__.__

._~--~~~

0.5
Bow

Fig. 21.8 - Effect of longitudinal location on lateral force estimator and motion-induced
interruptions in a frigate in beam waves at 15 knots; H113 = 5.5 m, T0 = 12.4 sec.

435

Effects of ship motions on passengers and crew

436

[Ch. 21

...

.8

"'

-~

Q)_

~N(,)
...

Q)

~~

-E
m._

...
Q)

Ul

"'
c

I
I

::J

2c
""0

g_
4

Q)
(J

::J

""0

c0

'iS
2:

20
Height above keel (metres)

Fig. 21.9- Effect of height on lateral force estimator and motion-induced interruptions in a
frigate in beam waves at 15 knots; H 113 = 5.5 m, T0 = 12.4 sec.

22
Seakeeping criteria and voluntary
speed loss in rough weather

22.1

INTRODUCTION

The methods described so far in this book have been aimed at predicting the
responses experienced by a ship in rough weather. We now need to consider whether
these responses will be acceptable in practice. This requires us to determine limiting
values or 'criteria' for each of the responses we predict.
Ships are required to carry out many different tasks and activities at sea and
criteria for acceptable responses depend on the task in hand. For example, motions
which are acceptable in a warship hunting a submarine woul~ not be tolerated by the
passengers on a cruise liner. Deck wetness which might be acceptable in a frigate
closed down for a high-speed dash in rough weather would. not be condoned for
operations which require men to work on the exposed upper deck.
It is first essential to identify the responses which actually limit performance of
the task. Consider, for example, a helicopter landing on the deck of a small frigate.
This is undoubtedly inhibited if the ship motions are large and we might expect the
following problems to arise in severe conditions:
(a) A large relative velocity between the flight deck and the helicopter might
damage the undercarriage or even the deck itself.
(b) A steady roll angle will "ause the wheels on one side of the aircraft to touch down
before those on the other with consequent greater risk of sliding across the deck.
(c) A high lateral velocity might topple the helicopter on touchdown or even cause it
to land with a wheel over the edge of the deck.
(d) Deck wetness and spray reduces visibility and deck friction making sliding more
likely: spray may also be ingested and damage the aircraft's engines.
For this particular example relative wind conditions will also play an important

438

Seakeeping criteria and voluntary speed loss in rough weather

[Ch. 22

role. Most helicopters can be satisfactorily controlled over only a limited range of
wind speeds and directions.
Criteria should be expressed in terms of these responses (and any others deemed
to be important) and not in terms of responses which have no direct relevance to the
activity being considered. For example, a heave motion criterion would be inappropriate for helicopters (unless the flight deck is over the centre of gravity) because
heave has no direct influence on the helicopter performance.
Having determined the motions which are believed to influence performance, the
next step is to estimate numerical values of the criteria. The ideal method would be to
conduct controlled trials in a variety of sea states to measure performance at the task
as a function of measured ship motions and to determine the conditions in which the
task becomes essentially impossible. Unfortunately this is generally impractical for
the following reasons:
(a) Sea state cannot be varied in a controlled manner and trials covering the
required range of conditions would be very protracted and expensive.
(b) Tasks involving human intervention (as most tasks do) would be heavily
influenced by all the psychological and physiological factors to which humans are
susceptible (motivation, morale, fatigue, acclimatisation etc) and it would be
very difficult to obtain consistent results unless a very large number of experiments with different personnel was conducted. These experiments would
inevitably be conducted in a variety of sea states making a proper analysis very
difficult.
Trials in ship motion simulators have the advantage that the same sequence of
ship motions may be reproduced over and over again so that controlled experiments
on a large number of subjects can be undertaken. However, most simulators are
incapable of reproducing the very large ship motions which are required to inhibit
performance and they cannot usually reproduce all six degrees of freedom (and other
more intangible aspects of life at sea) simultaneously. Furthermore, practical
experiments in most simulators are limited to non-strenuous tasks rather than tasks
requiring great physical effort and manual dexterity.
The only practical method of determining criteria for most tasks is to observe the
apparent performance of actual ships' crews in the ordinary everyday ocean
environment. This can be done in a number of different ways. One of the most
successful is to collect performance data using questionnaires and this is discussed in
Section 22.3 below. An alternative is to monitor ship motions over a long period of
time (months or years) and to correlate the measured motions with the activities of
the ship recorded in the log book. This will eventually build up an envelope of
motions which were actually tolerated while specific tasks were being performed in
the ship.
We may summarise the requirements discussed above in terms ofthree 'rules' for
the determination of criteria:
(1) Criteria must be related to a particular task or mission.
(2) The responses chosen for criteria assessment should be of actual concern to the
mission being considered.

Sec. 22.2]

Equipment criteria

439

(3) Numerical values of criteria should be determined by monitoring the apparent


performance of actual ships at sea.

22.2 EQUIPMENT CRITERIA


Although the criteria for many tasks are dependent on human performance, some
limitations arise from equipment design. In these cases a detailed analysis of the
equipment used in the task can be a useful first step towards determining appropriate
criterion levels. Consider for example a radar antenna. Most antennae are stabilised
so that the radar beam is maintained in the horizontal plane regardless of the pitch
and roll motions of the ship. However the stabilisation system has physical limits so
that the beam will no longer be horizontal and the radar set will cease to function
properly if the angular motions exceed certain critical values or 'criteria'.
The critical motion level will not, of course, be exceeded all the time even in the
most severe conditions. So we still need to decide on some measure of the acceptable
probability of exceedance. Generally the choice is b(;tween an estimate of the
proportion of time or the frequency with which the critical value is exceeded.
The choice depends on the nature of the equipment and the way in which the
performance is degraded. Consider for example a roll stabiliser fin which cavitates if
the modulus of the angle of incidence exceeds a certain value. Let us suppose that
cavitation is of concern because it causes erosion on the fin surface. Clearly we are
then interested in the rate of erosion and the proportion of time the motion exceeds
the critical value. This is given by equation (17.21), derived from the Gaussian
probability density function. If the critical incidence is cxcrit we have

(22.1)

where cr0 is the rms fin motion. The time the fin is being eroded is then given by

t = TH P(lcxl > CXcrit)

seconds

(22.2)

where THis the total time spent in the cavitation prone conditions of the calculation.
On the other hand we r.ight be more interested in the frequency with which the
fin emits bursts of cavitation noise. We are then concerned with the probability that
the peak fin incidence in each motion cycle will exceed cxcrit This may be obtained
from the Rayleigh probability density function by using equation 17.26:

1 ( cx~:t
P( ex > CXcrit) = exp [ "- 2

)2]

(22.3)

440

Seakeeping criteria and voluntary speed loss in rough weather

[Ch. 22

The mean period of the fin motion peaks is, from equation (4.27),

2rr~ (::)

seconds

(22.4)

where m2 and m 4 are the variances of the fin motion velocity and acceleration. The
critical incidence is then exceeded

3600 P( ex> CXcrit)

TP

times per hour

(22.5)

t and N in these formulae may be regarded as criteria for acceptable fin motions.
Numerical values of these criteria are best established by monitoring the actual
performance of ships at sea as required by Rule (3).

22.3 QUESTIONNAIRES
22.3.1

Introduction

Questionnaires provide one of the few practical methods of obtaining data on actual
performance at specific tasks at sea. Part of a typical questionnaire to obtain criteria
relating to damage to deck cargo on container ships is shown in Table 22.1. The
compiler of the questionnaire should always bear in mind that the recipient will
probably not be very interested in the business of criteria determination and the
questionnaire should therefore be made as brief and self explanatory as possible.
Otherwise it is likely to be consigned to the waste paper basket rather than be
properly completed. The questionnaire should therefore open with the minimum
number of questions designed to establish the identity of the recipient and his ship
and follow with a brief summary of the scenario postulated. A statement promising
confidentiality should always be included since some recipients (and their
employers) may consider any confession of damage experience as an admission of
poor seamanship. Questions soliciting opinions on design changes often yield fruitful
results and have the added advantage of boosting the ego of the recipient, thus
encouraging a timely response.
Ship operators will rarely be able to provide reliable direct estimates of limiting
ship motions or other criteria because they generally have no measurement systems
available to monitor the rough weather behaviour of their ship. Sailors are, however,
often reasonably well schooled in estimating wave conditions and much more fruitful
results will be obtained by asking for estimates of the worst sea state in which a
particular task can be completed. Subsequent calculations based on strip theory or

441

Questionnaires

Sec. 22.3]

Table 22.1- Typical seakeeping questionnaire: questionnaire on container damage


in rough weather
Name of captain
Name of ship
How long have you served in this ship?
Container ships running at high speed in rough weather are liable to damage the forward row of
containers due to green sea impacts. Containers have also been lost overboard in heavy rolling
conditions. This questionnaire is designed to obtain data on the incidence of the problem and to
determine the conditions in which it is likely to happen. Your replies will be treated in strictest
confidence.
( 1) Head seas
Deck wetness and risk of container damage is increased at high speed in head seas. In the table
below please indicate the maximum speed you believe you could maintain in each sea state in the
North Atlantic with an acceptable risk of container damage:

Sea state

Max speed (knots)


(2) Beam seas
In beam seas rolling will increase in high sea states. In severe sea states it may be necessary to
change course to avoid a beam sea heading and minimise the risk to the containers. In the table
below please indicate the maximum sea state you believe could be maintained in beam seas at
your normal cruising speed.

Sea state

(3) Design improvements


i 11
Which of the following design improvements would you recommend to alleviate container
damage?

(1)
(2)
(3)
(4)
(5)
(6)

increased freeboard forward


remove forward row of containers
roll stabilisers
stronger lashings
instrumentation to give warning of imminent problems
other (please specify).

Thank you for your assistance. Please now return this questionnaire to:

model tests may then be used to estimate the corresponding motions (or other rough
weather phenomena) in suitable idealised sea spectra. If this technique is used, the
speed and heading must, of course, be-specified in the questionnaire.
22.3.2 Analysis of numerical data
22.3.2.1. Mean value and staqdard deviation
,
Numerical data such as sea state estimates may be analysed to determine mean
values and standard deviations using the formulae given below. If the questionnaires

442

Seakeeping criteria and voluntary speed loss in rough weather

[Ch. 22

yield N estimates of a numerical quantity x (say the limiting sea state for a particular
activity), the mean value is

X=

(22.6)

and the standard deviation is

cro

f(L[(xx)Z])
N -1

(22.7)

\j

22.3.2.2 Student's t test for confidence in the mean value


It is almost inevitable that the number of questionnaire returns will be small and
there will probably be quite a wide divergence in the individual estimates, giving
large values of the standard deviation. We must then consider the possibility that the
derived mean value might be a freak result because the particular sample canvassed
in the questionnaire was biased or untypical in some way. We need to establish
whether is a good estimate of the 'true' mean value which would have been
obtained if it had been possible to obtain estimates from many more ships. This may
be done using Student's t test (see Mack (1966)). Briefly we may state that there is a
95% probability or 'confidence' that the true mean value which would have been
obtained from a much larger sample will lie within the range

where

tcrit

and tcrithjN are given as functions of N in Fig. 22.1.

22.3.2.3 Tests for significant differences between ship classes


Questionnaires are often used to compare the performance of different ship classes.
This will yield two mean values A and 8 and two standard deviations a A and cr 8
from the two samples of N A and N 8 returns. We need to determine whether there is a
significant difference between the two mean values. Two tests are used. We define a
test function

(22.8)
where a A> cr8 so that v'F > 1.
If y'F is greater than the critical value given in Fig. 22.2 there will be a 95%
probability that the two standard deviations are significantly different. If y'F is less

Sec. 22.3]

Questionnaires

443

50
Number in sample N

Fig. 22.1- Critical values of Student's t function for 95% confidence level.

5
Ns=3
4

\ Fcc;t

rl

5
2

10
20
X

Fig. 22.2- Critical values of y F for 95% confidence level.

444

Seakeeping criteria and voluntary speed loss in rough weather

[Ch. 22

than y Fcrit the samples may be combined to give a single estimate of standard
deviation for both samples:

cro

/((NA -1)cri + (N8 -1)cr~)


NA +N8 -2

(22.9)

'V

If the standard deviations are not significantly different and have been pooled in
this way we may apply a further test to establish whether the two mean values are
significantly different. If the test function

(22.10)
exceeds the critical value tcrit given in Fig. 22.1 the two means are significantly
different at the 95% confidence level (i.e. there is a significant difference in the
performance of the two ship classes). N is taken as the total number of returns from
both classes for this test. In this formula Sect is the standard error of the differences
defined by

(22.11)

Example
A questionnaire issued to two ship classes yields the following data for limiting sea
state for a particular activity:

Class A
Class B

Mean value

Standard
deviation

Number in
sample

5.7
4.9

1.2
1.4

7
10

What confidence levels can be attributed to each individual result? Are the results
significantly different?
Using Student's ttest and Fig. 22.1 there is a 95% confidence that the true mean
value for each sample lies within the following ranges:
Class A:

5.7 1.2 x 0.9

5.7 1.08;

i.e. range is 4.62--6.78

Class B:

4.9 1.4 x 0.7

4.90.98;

i.e. range is 3.92-5.88

Using the F test we find that the ratio of the standard deviations is

Questionnaires

Sec. 22.3]

445

and the critical value of yF is, from Fig. 22.2,


VFcrit

= 2.0

so that the two standard deviations are not significantly different. Equation (22.9)
then gives the combined standard deviation as
cr0 = 1.32

and the standard error of the differences is, from equation (22.11),
sed

= 0.65

The test function tis, from equation (22.10),

1.23

and there is a total of 17 samples from the combined questionnaires. Fig. 22.1 gives
the critical value of t as
lcrit

2.04

and the results obtained from the two classes are seen to be not significantly different
at the 95% confidence level. The two samples may therefore be combined to yield a
mean value

X=

5.7x7+4.9x10
17

5.2

and Student's t test may again be applied to find the confidence limits for this
combined result. We find a 95% confidence that the true mean value will lie within
the range
5.2 1.32 X 0.53

5.2 0.7;

i.e. range is 4.5-5.9

It can be seen that combining the samples, where this can be justified, results in an
improved estimate of the mean value.

446

[Ch. 22

Seakeeping criteria and voluntary speed loss in rough weather

22.3.3 Analysis of 'box-ticking' questions


Box-ticking questions yielding 'yes/no' type answers require a different technnique
for determining the reliablity of the results. In principle the returns can be analysed
to give the proportion or 'vote' for a particular opinion: for example, we might find
that out of ten recipients of the questionnaire shown in Table 22.1, seven think that
roll stabilisers would be a useful design improvement. The question is whether this
can be taken as a valid indication with, say, a 95% confidence level that roll stabilisers
would be favoured by a majority of the masters of a (hypothetical) very much larger
sample of ships of the same class. It is quite possible that the seven who favoured
stabilisers might be ill informed and not typical of the population of captains at large
so that the majority vote is a freak result.
To examine this possibility we need to consider whether a majority vote might be
achieved by chance.
As an example consider the possible outcomes of a questionnaire issued to four
ships. If Y means 'yes' and N means 'no' there are 16 possible outcomes to a boxticking question:

Outcome number

Return
number

1
2
3
4

10 11 12 13 14 15 16

N
N
N
N

y
N
N
N

N
y
N
N

y
y
N
N

N
N
y
N

y
N
y
N

N
y
y
N

y
y
y
N

N
N
N
y

y
N
N
y

N
y
N
y

y
y
N
y

N
N
y
y

y
N
y
y

N
y
y
y

y
y
y
y

Suppose that the questionnaire yielded a majority vote (i.e. three or more of the
recipients voted 'yes'). This could be achieved in the five different ways marked with
an asterisk. The probability that this result occurred by chance is therefore 5/16 =
0.3125 and the probability that it was not a chance result is 1-0.3125 = 0.6875. So
the confidence level that the majority of a hypothetical much larger population of
ships' masters would vote positively is 68.75%.
Standard practice in statistical analysis is to demand a confidence level of at least
95% and we can see that this cannot be achieved with a sample of only four returns.
Even if all four recipients respond with a 'yes' vote there is a 1/16 chance that the
result is by chance and the confidence level that this would indicate a majority
opinion in the world at large is only 1-1/16 = 93.75%.
We may extend this approach to an arbitrary number of questionnaire returns as
follows. We require to find the minimum number of votes q for a sample size N which
indicate a majority opinion with a 95% confidence level. Consider a questionnaire
sample of N returns with r 'yes' answers to a particular question. The proportion or
vote for the opinion expressed is

Questionnaires

Sec. 22.3]

447

(22.12)

This vote could be achieved in

Nc,

N!
r!(N-r)!

(22.13)

different ways. The total number of possible combinations of answers is

(22.14)

If the answers to the questions were truly random all pvssible combinations of
answers would be equally likely. Hence the probability of q or more 'yes' answers
would be

P=

(22.15)

and the probability or confidence level that this is not a chance result is

:II
C

1-P

(22.16)

Table 22.2 gives the minimum number of positive responses q required for a 95%
confidence level as a function of the sample size N, calculated using the equations
derived above. From this it can be seen that the absolute minimum number of returns
required is five and they must all vote positively before this can be accepted as a valid
indication of a majority opinion. The required proportion of positive votes falls as
the sample size is increased. Note that the same analysis applies with equal validity to
negative opinions (i.e. not ticking a box).
22.3.4 Increasing the sample size
Evidently there are considerable benifits to be obtained if the sample size is made as
large as possible. Unfortunately it is usually only the largest navies which have more
than, say, half a dozen ships in a class and we have seen that this will generally yield
results of only marginal reliability. For some aspects of seakeeping it may be possible
to pool the results from more than one class of ship, as described above, but this of
course precludes the possibfity of distinguishing any effects of design differences
between the classes.

448

Seakeeping criteria and voluntary speed loss in rough weather

[Ch. 22

Table 22.2- Minimum number of ticks for majority vote at 95% confidence level
Sample size
N

Number of
returns q
ticking a
given box

2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18

5
6
7
7
8
9
9
10
10
11
12
12
13
13

19
20
21
22
23
24
25
26
27
28
29
30

14
15
15
16
16
17
18
18
19
19
20
20

Sample size
N

Number of
returns q
ticking a
given box

31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51

21
22
22
23
23
24
24
25
26
26
27
27
28
28
29
30
30
31
31
32
32

52
53
54
55
56
57
58
59
60

33
33
34
35
35
36
36
37
37

It is sometimes tempting to increase the sample size by canvassing the opinions of


ex-captains. This is not good practice as these individuals may suffer from the so
called 'halo' effect by which experiences of long ago become distorted in the mind
and generally exaggerated.
However, one technique which is acceptable is to repeat the investigation at
intervals over a number of years so that the opinions of successive captains are
canvassed in the same ships. Large sample sizes can be gradually built up in this way.

Sec. 22.4]
22.4

449

Speed loss in rough weather

SPEED LOSS IN ROUGH WEATHER

One 'mission' which has received considerable attention is the ability to maintain
speed in severe head seas. This is often regarded as a general indication of the
seakeeping qualities of a ship since excessive motions, slamming, deck wetness, etc.
force the captain to reduce speed to avoid damaging his ship and its contents and
injuring his crew and passengers.
A typical questionnaire designed to obtain data on speed loss in rough
weather is shown in Table 22.3. Some results for two classes of frigates are shown in
Table 22.3 -

Questionnaire on speed in rough weather

Name of captain
Name of ship
How long have you served in this ship?
Ship speed is limited in rough weather by two factors:
(a) In moderate sea states the action of the wind and waves causes the ship to slow down even if
full power is maintained.
(b) In more severe sea states the captain may decide to reduce power or change course in order to
alleviate slamming, deck wetness, propeller emergence, ship motions, etc.
Imagine that you are required to make a high-speed passage in rough weather. All ship
equipment should be fully operational at the end of the pasage. In the table below please indicate
the maximum speed that you could maintain in your ship in the given sea states. Please also
indicate with a tick the sea state in which you would first reduce power.
Sea state

Max speed (knots)


Sea state for power
reduction
In the table below please indicate the reasons in their order of importance for your decision to
reduce power.
Slamming
Deck wetness
Ship motions
Propeller emergence
Oth~r

(please state)

Fig. 22.3. Estimates of mean speed and standard deviation have been calculated
using equations (22.1) and (22.2). As-expected the ships suffer a dramatic speed loss
as the sea state worsens. The standard deviation increases in high sea states,
reflecting the difficulty of estimating speed loss and perhaps the lack of extreme
rough weather experience of t'he commanding officers'~
Application of Student's t test gives the confidence limits shown as shaded areas

450

Seakeeping criteria and voluntary speed loss in rough weather

[Ch. 22

Not significantly
different

(f)

0c

15

""0

Sea state
for power
reduction

(j)
(j)

c.

(/)

10

Significant wave height (metres)

8
Sea state

Fig. 22.3- Typical results of questionnaire on speed in rough weather.

in Fig. 22.3. The high standard deviation and the small number of ships in Class B
widen the confidence limits considerably so that the estimates of mean speed for this
class are much less reliable than those for class A.
Application of the F test described above shows that the standard deviations are
not significantly different and they may be pooled to give a common value using
equation (22.9). The test function tis greater than the critical value (tcrit = 2.03 for N
= 37; see Fig. 22.1) for all but the highest sea states. So the two results are

Criteria for speed loss

Sec. 22.5]

451

significantly different at the 95% confidence level and we may be confident that the
performance of class A is better than that of class B at least in moderate sea states.

22.5

CRITERIA FOR SPEED LOSS

The questionnaire shown in Table 22.3 asks for the captain's reasons for reducing
speed. For most conventional ships it is usually found that slamming is the primary
cause of speed reduction with either deck wetness or ship motions given as the
secondary factor. Propeller emergence is usually only important for merchant ships
in ballast. For the frigates described above the order of importance found from the
questionnaire was:
(1) slamming
(2) ship motions
(3) deck wetness.
Fig. 22.4 shows predictions of slamming frequency for ship class A in head seas as

0..

u..

.t=

"'
"'

..Q

25

-.J
N

ci

co....
:::J

..r::

c;;
E

"'

(/)

10
12

Significant wave height (metres)

Fig. 22.4- Limiting slamming frequency.

a function of speed and significant wave height (the most probable modal periods for
the North Atlantic have been assumed). Now the questionnaire gives the maximum
permissible speeds in given wave conditions and the results have been plotted in Fig.
22.4 as a locus of acceptable combinations of speed and wave height. This allows us to

452

Seakeeping criteria and voluntary speed loss in rough weather

[Ch. 22

estimate the maximum tolerable slamming frequency: in this case the captains of ship
class A apparently tolerate about 60-80 slams per hour, and }Ve may take this as a

suitable criterion for slamming.


We may follow the same approach for ship motions and deck wetness and similar
diagrams are shown for suitable measures ofthese pheneomena in Figs 22.5 and 22.6.

.E0
E
Q)

u>
Q)

:.0
::J

c"'

"'
E
Q)

""0

2l

.<:::

"'

0
Significant wave height (metres)

Fig. 22.5- Subjective motion.

We see that the captains of these frigates apparently tolerate about 80-100 deck
submergences per hour at the forward perpendicular (equation (20.10)) and a
weighted mean subjective motion of about 11-12 (equation (21.10)).
However, deck wetness and ship motions are not in this case the limiting factors
which force the captain to reduce speed. So these results cannot be taken as criteria
for acceptable levels. All we can say is that the captain would tolerate more frequent
deck submergence and higher levels of motions if slamming were not so severe.
These results may help in the determination of criteria but they cannot be used in the
definitive way which was possible for slamming. We may therefore guess that the

Criteria for speed loss

Sec. 22.5]

453

400
c..

LL

co
~

:::l

0
~

U;
Q)
()

Q)

2l
Q)

..0
:::l
V)

-"'
()
Q)

0
Significant wave height (metres)

Fig. 22.6- Deck submergence frequency.

maximum tolerable deck submergence frequency in this ship is about 120 per hour
and that the maximum tolerable subjective motion is about 15.
Fig. 22.7 shows the speed loss from the questionnaire compared with speed loss

"0
Q)
Q)

c.

(/)

14
Significant wave height (metres)

li'ig. 22.7- Speed loss curves.'

454

Seakeeping criteria and voluntary speed loss in rough weather

[Ch. 22

curves derived from Figs 22.4-22.6 for these estimated criteria. As expected, the
slam limited speeds fit the questionnaire data tolerably well, at least in the middle
range of sea states, while the other curves are higher. Note that all three limits are too
high in high sea states. This may arise because the ship and its crew must then tolerate
severe levels of all three phenomena simultaneously and this will be expected to
reduce their tolerance to each individual factor. In other words severe motions and
deck wetness will reduce the crew's tolerance of slamming and limit the maximum
possible speed.
Table 22.4 lists a selection of seakeeping criteria for speed in rough weather
derived by various authors from questionnaires, trials and intuition.

Table 22.4 - Seakeeping criteria for speed in rough weather


Author

Ship
type

Slamming

Wetness

Kehoe
(1973)

Warship

60/hour
at 0.15L

60/hour
atFP

Ochi and
Motter
(1974)

Merchant

Probability
0.03

Probability
0.07

Shipbuilding
Research
Association
of Japan
(1975)

Merchant

Probability
0.01

Probability
0.02

Lloyd and
Andrew
(1977)

Warship

Lloyd and
Andrew
(1977)

Merchant

Aertssen
(1963, 1966,
1968, 1972)

Merchant

Andrew and
Lloyd
(1981)

Warship

Comstock

Warship

Yamamoto
(1984)

Walden and
Grundmann
(1985)

Propeller
emergence

Vertical
acceleration

Probability
0.1

36/hour

SM = 15

120/hour

Probability
0.03 or 0.04

Probability
0.25
90/hour

SM = 12

20/hour

30/hour

0.2g rms
at bridge

Merchant

Probability
0.02

Probability
0.02 at FP

Probability
of
exceeding
0.4g at
bridge = 0.05

Warship

Probability
0.03

Probability
O.D7

eta/.

(1982)

23
Operational effectiveness

23.1

INTRODUCTION

Operational effectiveness is defined in semantic terms as the ability of the ship to go


to sea and accomplish its mission in whatever weather conditions it may find. In
numerical terms we may express operational effectiveness as the proportion of time
the ship can successfully accomplish its mission in a given combination of sea areas
and seasons. Thus operational effectiveness can be used as a yardstick for comparing

competing ship designs.


We have seen in previous chapters how we may calculate the rough weather
responses of a ship for any combination of weather condition, ship speed and
heading. If we know the limiting values of these response~ ~or a particular task or
mission we may then determine whether the task can be accomplished in that
particular set of circumstances. Generalising the calculation to include all possible
combinations of sea area, season, wave direction, significant wave height, modal
period, ship speed and course allows us to calculate the proportion of time for which
the task is possible.

23.2

SEA AREA AND SEASON

The first essential in any calculation of operational effectiveness is to specify the


ocean environment in which the ship is to operate. This is conveniently done in terms
of the scatter diagrams shown in Fig. 23.1. These show two examples of the
conditional frequency distributions of sea area and season for a cargo ship on a
transatlantic route and a warship operating in the North Atlantic.
Both ships are required to operate throughout the year so the frequency
distribution of seasons !season, is uniform. The merchant ship is required to operate
along a great circle route between Europe and North America and the conditional
frequency distribution of sea ~reas for each season reflects the time it is expected to

[Ch. 23

Operational effectiveness

456

...

Q)

Ol

;::

E
E

....

...
c

Q)

.=::!

...
c

E
E

.=::!

::J

<t:

Ol

Q)

c.

Q)

c.
'IJ

::J
(J)

<t:

(J)

::J
(J)

.10

.10

.10

.05

.03 .03

.03

.20

.20 .20

.10

.20

.20 .20

.10

.04

.04

.04

.07

.04 .04

.04

.07

10

.11

.11

.11

.11

.07

.07

.07

.10

11

.0

.07

.07

.07

.10

.04 .04 .04

.07

.04

.04 .04

.07

.07

.07

.07

.10

.04

.07

::J

(1J

~
(1J
(1J

Q)

(J)

15 .25

.25 .25

16 .44 .44

17

.25

.25 .25
0

23

.07 .07

.07

.07

.04 .04

24

.13 .13

.13

.13

.03

.03 .03

.05

25

.0

.03

.03

.03

.05

.25

.25

.25

.25

.25

.25

.25

.25

Transatlantic
cargo ship

fseason

fseason

Warship

Fig. 23.1- Typical conditional frequency distributions of sea area, !area> for given seasons in
the North Atlantic.

spend in each area along the route. Sea areas are defined according to Hogben,
Dacunha and Olliver (1986) (see Fig. 5.3).
The warship is required to operate throughout the North Atlantic with most of its
time spent in the northern areas 1, 3 and 4. In winter its operational area is moved
south.

23.3 SHIP SPEED AND COURSE


The required frequency distribution of speed and course is estimated in the form
shown in Fig. 23.2. For the transatlantic cargo vessel the required course is easterly
or westerly, depending on the direction of the voyage. Speed is required to be the
maximum economical cruising speed.

457

Calculation of operational effectiveness

Sec. 23.4]

Ship course

30

NE

SE

SW

NW

1.0

1.0

0.5

0.5

25

en
0c

=="C

20
15

Q)

Q)
10
c.
(/)
c.
5
.s::

C/)

0
fcourse

(a) Translantic cargo ship

Ship course
N

NE

SE

SW

NW

.01

.01

.01

.01

.01

.01

.01

.01

en

.10

.10

.10

.10

.10

.10

.10

.10

.50

.50

.50 .50 .50

.50

.50 .50

.30
c. 10
c.
.07
.s::
5
C/)
.02
0
.125

.30

.30

.30 .30

.30

.30

.30

.07

.07

.07

.07

.07

.07

.07

.02

.02

.02

.02

.02

.02

.02

30
25

0 20

15
=="C
Q)
Q)

(/)

.125 .125 .125 .125 .125 .125 .125 f

course

(b) Warship

Fig. 23.2 -Typical condition frequency distributions of speed, f u, for given ship courses.

The warship is again required to be much more versatile. All courses are equally
likely and a wide range of speeds is demanded. Nevertheless, the economical cruising
speed is frequently used. Very high and very low speeds are rare.

23.4 CALCULATION OF OPERATIONAL EFFECTIVENESS


The method of calculating operational effectiveness given here is a simplified version
of that proposed by Andrew, Loader and Penn (1984). The calculation is illustrated
in graphical form in Fig. 23.3.
The operating environmenJ is specified in terms of, a scatter diagram of seasons
and sea areas as already shown in Fig. 23.1. For each combination of sea area and

458

Operational effectiveness

[Ch. 23

Criteria

Fig. 23.3- Calculation of operational effectiveness.

season we may obtain the conditional frequency distribution of wave directions fx


from a suitable wave atlas (for example Hogben, Dacunha and Olliver (1986)). For
each wave direction the same data source gives the joint frequency distributions of
mean zero crossing periods and significant wave heights fTH Then for each wave
condition we may calculate all ship motions for every possible combination of ship
course and speed. (In practice, of course, it is usual to confine attention to those
motions which are relevant to the particular mission being considered.) The required
heading to the waves in these calculations is obtained from the difference between
the ship's course and the predominant wave direction (both referenced to North).
Some of these responses may exceed the maximum permissible values. If this is
the case the operation in question is deemed to be impossible and the ship cannot
achieve its mission.
The proportion of time the ship spends in a given season, sea area, wave
direction, zero-crossing period and significant wave height at a given speed and
course is given by
p

/season /area

fx fTH

/course

fu

(23.1)

where
/season
/area

fx

/course

is the frequency distribution of season


is the conditional frequency distribution of sea areas for a given season
is the conditional frequency distribution of primary wave directions relative to
North for a given season and sea area
is the joint frequency distribution of mean zero-crossing periods and significant wave heights for a given season, sea area and wave direction
is the frequency distribution of ship courses relative to North

Sec. 23.4]

fv

Calculation of operational effectiveness

459

is the conditional frequency distribution of ship speed for a given ship course

The ship can successfully execute its mission if all the responses are less than the
critical levels for the mission being considered. The proportion of time for which this
is the case (the operational effectiveness) is obtained from the weighted sum of all the
possible values of P. Mathematically this may be written as
(23.2)
where :Eseason :Earea :Ex, :ETH :Ecourse and :E 0 imply summation over all seasons, areas,
wave directions, zero crossing periods and significant wave heights, ship courses and
speeds respectively. r n is a counting functional defined by
for rn < rn,ril
for rn > rn";'
where rn is the nth response of interest (e.g. roll, subjective magnitude etc.) and rn .
'"'
is its critical value.
Table 23.1 shows an example of a calculation of operational effectiveness taken

Table 23.1- Operational effectiveness of a frigate in


the North Atlantic
(After Andrew, Loader and Penn (1984))

Unstabilised
Pitch only
Roll only
Pitch and roll

0.976
0.650
0.649

0.976
0.851 '
0.851

rms pitch limit = 2.0 degrees.


rms roll limit = 3.0 degrees.

from Andrew, Loader and Penn (1984). This shows the operational efffectivness of a
frigate with and without roll stabilisers in the North Atlantic. For this example it is
assumed that the frigate's mission will be impossible if the rms pitch exceeds 2.0
degrees and the rms roll exceeds 3.0 degrees.
Considering first the pitch motions in isolation we see that the ship is able to
achieve its mission almost all the time. Roll stabilisers, of course, have no influence
on this result. Roll motions have a much greater effect on the ship's ability to achieve
its mission and the unstabilised ship is effective for little more than half the time.
These effects are considerably alleviated by the stabilisers.

24
The effect of hull size and form on seakeeping

24.1

INTRODUCTION

The methods outlined in this book allow the designer to quantify and assess the
seakeeping qualities of a new design before the ship is built. If the predicted
performance is inadequate the designer will need to change the size and/or shape of
the hull to effect the necessary improvements.
The designer therefore requires some guidance on the performance improvements which are likely to result from changes to hull form and size. The methods
described earlier may be used to provide information on these trends and this chapter
gives the results of some specimen calculations. These are necessarily specific to a
particular basic hull form and should not be used to give numerical estimates of the
changes in performance of other hull forms. However, the trends described will be
found to be generally applicable and the results may be used to suggest suitable
changes in size and shape to a wide range of hulls. The actual performance of a
particular design should always be estimated from strip theory calculations or model
tests.
We will only consider the effect of hull form and size on motions in the vertical
plane and confine our attention to head waves. This heading generally gives the worst
vertical plane motions and it is found that a form which has low motions in head
waves nearly always has a satisfactory performance at other headings. The most
important motions to be considered in this context are the absolute vertical
accelerations and the relative motions at the bow. The former are an indication of the
severity of the motions experienced by the crew and passengers and we choose to
calculate the acceleration at x81 MILs = 0.15, a typical location for the bridge of a
warship or the passenger accommodation on a ferry. In general it is found that a ship
which has satisfactory motions at this location will also have acceptable motions at
other locations occupied by passengers or crew.
We shall calculate the relative motion at x81 M/L 5 = 0.3, a typical location for a
severe slam, and use this result to estimate the probability of keel emergence at this
station.

Sec. 24.3)

Effect of hull size

461

Lateral plane motions are also influenced by changes in hull form and size.
However, it is not generally considered worthwhile to optimise hull form to achieve
small motions in the lateral plane. These are more effectively reduced by installing
suitable roll reduction devices such as bilge keels, active fins or passive tanks and, of
course, the rudder.
24.2

PARENT HULL FORM

There are an infinite number of ways in which the size and shape of a given hull may
be changed and we must choose suitable constraints on the changes we shall
consider. We first need to select a parent hull form to provide a basis for our
calculations. We choose a typical frigate having the following characteristics:
draught/length ratio

DPILP

0.034

beam/length ratio BPI LP = 0.120


forward waterplane area coefficient

Cwfp

= 2 Awtpi(Bp Lp) =

0. 700

We shall examine the effects of changing the size and shape of the frigate hull
form as follows:
(a) changing the size (length) of the hull while keeping the shape constant
(b) Changing the shape of the hull while keeping the length constant.
24.3

EFFECT OF HULL SIZE: CHANGING THE SIZE OF THE HULL


WHILE KEEPING THE SHAPE CONSTANT

The effect of hull size may be studied by calculating the responses of a series of
geometrically similar ships all having the same hull shape ~t differing lengths as
shown in Fig. 24.1. Changing the length ofthe hull while keeping the shape constant
results in proportional changes to all the linear dimensions (~earn, draught, freeboard, etc.) and the displacement varies as the cube of the length. So these hulls may
be regarded as scale models or geosims of the parent.
Transfer functions for heave and pitch of the parent form in head waves are
shown in Fig. 24.2. These are given in non-dimensional form as functions of the nondimensional wave frequency roy (L.fg), wave length A./ L, and Froude number. In this
form these transfer functions apply to all the derived hull forms because they all have
the same geometrical shape. All the motion responses are essentially unity in waves
which are much longer than the ship and more or less negligible in waves shorter than
a critical length which is about three-quarters of the ship length. In other words ships
tend to contour very long waves but do not respond to very short waves, as already
discussed in Chapter 13.
Transfer functions for a given ship length and speed may be derived from these
results and examples for 20 knots are shown in Figs 24.3 and 24.4. These figures show
the calculation of the rms motions in a typical long crested Bretschneider wave
energy spectrum using the s\}.ort (wave frequency domain) method described in
Chapter 14 (Section 14.4).
Consider first an infinitely long ship. All the waves in the seaway are shorter than

462

The effect of hull size and form on seakeeping

[Ch. 24

[----------=::::>
~----------L--~/7
_c____________

~-

L-?
t::::::::_ _ _ _ _ _
_~/

c
~~--------~--~/?
Fig. 24.1-- Geometrically similar ships.

the critical wave length, the heave and pitch transfer functions are zero over the
entire range of frequencies and the ship does not respond to the wave input at all.
As the ship length is reduced, some of the waves in the seaway begin to exceed the
critical length and the transfer functions adopt appreciable values in the range of
important wave frequencies. So the heave and pitch responses are increased, as
shown in Figs 24.3(c) and 24.4(c).
A very short ship with Ls = 0 metres finds that all the waves in the seaway are
longer than the ship and the transfer functions are unity over the entire range of
frequencies. The ship heave response 'output' Sx,(ro) is then the same as the wave
energy spectrum 'input' S;(ro). So the rms heave is the same as the rms wave
depression. Using equation (4.32)

cr0 = Vm0 = 0.25 H 113

metres

In the same way the pitch response output for a very small ship is the same as the
wave slope energy spectrum input. In other words a very small ship contours all the
waves in the seaway and suffers large heave and pitch motions.
Fig. 24.5 shows the rms heave and pitch motions obtained from estimating the
area under the response curves shown in Fig. 24.3(c) and 24.4(c). These clearly
demonstrate that small ships suffer from increased absolute motions in a given
seaway.
Fig. 24.6 shows a similar calculation for the vertical acceleration at xB 1M/Ls =
0.15.

Effect of hull size

Sec. 24.3]

463

1.5

1.0
0

x."'

0.5

6
(IJ

Y(L 5 /g)

1.5

1.0
0

JJ

><
0.5

0
(I)

V(Lsfg)

Fig. 24.2- Non-dimensional heave and pitch transfer functions for p~rent hull form in head
waves.

The acceleration response ordinates in Fig. 24.6(c) are given by

Ss/co) = ( cot:30

Sc,

(co)

(metres/second2_)2f(radian/second)

(see equation (14.11).


Again the displacement transfer function for an infinitely long ship is zero at all
frequencies and the ship does not respond to the seaway at all. As the ship length is
reduced the transfer functions increase and the response output becomes appreciable. The effects of encounter frequency become progressively more important as
the ship length is reduced (in other words small ships respond more vigorously to the
shorter waves). In the limit a very small ship with Ls = 0 metres will contour all the
waves and suffer very large accelerations.
Fig. 24.7 shows a similar calculation for relativ~,motion at x 81 M/Ls = 0.30. We

Fig. 24.3- Effect of shp size on heave motion in head waves at 20 knots;
H 113 = 5.5 m; T0 = 12.4 sec.

have seen that the infinitely long ship does not respond to the waves and it follows
that the relative motion must then be the same as the wave depressiont (see equation
(13.14)). This is confirmed in Fig. 24.7 where we see that the infinitely long ship has a
relative motion transfer function which is unity over the entire range of frequencies.
t Apart from any distortion caused by swell-up effects.

Effect of hull size

Sec. 24.3]

0
1.5

(b)

1.0
0

JJo

"'

)(

0.5

:X:

0.008

Q)

::0"'
f:'

%
f:'

J
:X:

0
Wave frequency'w (radians/second)

Fig. 24.4- Effect of sniP size on pitch motion in head waves at 20 knots;
H 113 = 5.5 m, T0 = 12.4 sec._,.

465

466

The effect of hull size and form on seakeeping

[Ch. 24

Ship mass (tonnes)

(i)

(i)

Q)

a;
E_

"'
~
Q)

Q)

..c

>

"'

B
a.

Q)

..c
Ul

Ul

Ship length (metres)

Fig. 24.5- Effect of ship size on rms heave and pitch in head waves at 20 knots;
R 113 = 5.5 m, T0 = 12.4 sec.

So the response is the same as the wave energy spectrum and the rms relative motion
is

0.25 H 113

metres

As the length is reduced, the ship begins to contour the longer waves and the
transfer function encompasses a smaller range of frequencies. However, peaks
appear in the transfer function and the response is amplified if these peaks coincide
with the peak of the wave energy spectrum. For very small ships (Ls = 0 metres)
contouring all the waves the transfer function is everywhere zero and there are no
relative motions.
These trends are summarised in Fig. 24.8.
The probability of keel emergence calculated using equation (20.5) is shown in
Fig. 24.9. For the seaway considered here keel emergence is common for ships in the
range 70-200 metres and is most common for ships about 120 metres in length.
Smaller ships are less susceptible because of their small relative motions. These ships
will, however, suffer from high vertical accelerations which would make life on board
intolerable at the speed considered in this example.
The greater draught of the larger ships ensures that keel emergence is unlikely for

Effect of hull size

Sec. 24.3]
6

467

(a)

uQl
(/)

=o

1:

~J

(b)

We/
/

JJ

(/)

=o

"'

L5 =0 m

0
80

uQl

(c)

uQl
(/)

=o
~

N-

'1;
Ql
(/)

E
:::;
3
M

VJ~

2.0

Wave frequency w (radians/second


Fig. 24.6- Effect of ship size O!! absolute vertical acceleration at 20 knots;
x 81 MJL, = 0.15, H113 = 5.5 m, T0 = 12.4 sec.

these vessels in spite of their appreciable relative motions. Increasing the size of the
ship also gives a dramatic reduction in vertical acceleration. Large ships are generally
more comfortable than small Qnes in rough weather.

468

The effect of hull size and form on seakeeping

c:;-

[Ch. 24

L5 =150m

Ql
1/)

=c
~

l
3

~
Om
0

2.0
Wave frequency w (radians/second)

Fig. 24.7- Effect of ship size on relative bow motion at 20 knots; H 113 = 5.5 m, T0 = 12.4 sec.

24.4 EFFECT OF HULL SHAPE


24.4.1 Introduction
Detailed changes of shape, such as easing the radius of curvature of the bilges or
changing the deadrise angle at the keel, have little discernible effect on ship motions
in the vertical plane. The designer seeking an improvement in seakeeping performance must think in terms of changes to the overall proportions of the ship rather
than piecemeal modifications.
Seakeeping performance assessment must therefore be considered at an early
stage in the design process before the major proportions have been settled.

Sec. 24.4]

Effect of hull shape

469

Ship mass (tonnes)

""0

Cii

~Q>

"

Q>

(/)

.s

(i;

"EQ>

Relative motion

~
1i)

.s

Q>

...,0

""'
0.

(/)

'0

"'Q;
Qi

(/)

"
"'"

(/)

0
0

500
Ship length (metres)

Fig. 24.8 -

Effect of ship size O!!_ absolute and relative motions at 20 knots in head waves;
H 113 = 5.5 m, T0 = 12.4 sec.

Ship mass (tonnes)

500 5000 10000

o.2or-~-~~--.....;..;.~..--.=;;.....;........,--.,

Q)
{.l

Q)

2'
Q)
E
Q)

Q)

.~

:.a

"'

.0
0

r.t

500
Ship l_~ngth (metres)

Fig. 24.9- Effect of shig_ size on keel emergence at 20 knots in head waves;
H 113 = 5.5 m, T0 = 12.4 sec.

The effect of hull size and form on seakeeping

470

[Ch. 24

We shall examine the effects of changing the draught/length and beam/length


ratios and the forward waterplane area coefficient. These particular parameters have
been selected because they have appreciable and well defined effects on seakeeping
performance. Other parameters, such as block coefficient, transom beam, etc. have
relatively minor effects and it would not normally be considered worthwhile
changing them to improve seakeeping performance.
24.4.2 Draught/length ratio
Fig. 24.10 illustrates changes to the draught/length ratio. Since we choose to keep the

!.....-=:=::::::==============-zL-

_I

L-

- c=::::::-----:;,/'-

-c::::-----yL

Fig. 24.10- Draught/length ratio variations.

block coefficient constant, increases in draught result in proportional changes to the


displacement. The effects on the motions are shown in Fig. 24.11. These are given in

Relative motion

II

.......

.------

Vertical
acceleration

Fig. 24.11- Effect of draught/length ratio on rms motions in head waves;

T~

= 3.5, FN = 0.3.

Effect of hull shape

Sec. 24.4]

471

non-dimensional form which makes them more easily applicable to ships of different
sizes. They are presented for a Froude number of 0.3 and a non-dimensional modal
period defined as
T~ = T0 Y(g!L,) = 3.5

which corresponds approximately to a modal period of 12.4 seconds for a ship length
of 125 metres. Similar trends are found for other modal periods and ship lengths.
Reducing the draught increases the added mass and damping coefficients, as
shown in Fig. 11.10, and this has the generally beneficial effect of reducing both the
absolute and relative motions. Fig. 24.12 shows the corresponding effects on the

0.15

Fig. 24.12- Effect of draught/.Iength ratio on probability of k~~l emergence;


T0 = 3.5, FN = 0.3.

probability of keel emergence. This rises dramatically as the draught is reduced, in


spite of the associated reduction in relative motion.
24.4.3 Beam/length ratio
Figs 24.13-24.15 show similar results for beam/length ratio. Again increases to the
beam are associated with increased displacement and the larger added mass and
damping coeficients would be expected to yield reduced motions. However, the
excitation from the waves is also increased because of the larger waterplane area and
the resulting changes to the motions are not dramatic. A large beam/length ratio is
seen to reduce absolute vertical accelerations at the expense of increased relative
motions. The probability of keel emergence is greatest for ships with beam/length
ratios in the range 0.1-0.2 (typical of many modern designs).
24.4.4 Forward waterplane area coefficient
The effects of changing the fQrward waterplane area coefficient are shown in Figs
24.16-24.19. A large forward waterplane area coeffi~ient increases the local beam at

472

The effect of hull size and form on seakeeping

[Ch.24

c____________________

_______

-----------------

Fig. 24.13- Bearn!length ratio variations.

0.6

0.4

- 0.4

0.6

- -- Relative
motion

l:f
'o
0

0.2

Absolute-- 0.2
acceleration

0.1

0.2

tl')
M

l:f

:r"
0

0
0.3

8 5 /Ls
Fig. 24.14- Effect of beam! length ratio on rms motions in head waves;

0.1

0.2

T~ =

3.5, FN = 0.3.

0.3

Fig. 24.15- Effect ofbearnllength ratio on probability of keel emergence;

T;, = 3.5, FN = 0.3.

Summary

Sec. 24.5]

473

L
_______
:>
Fig. 24.16- Forward waterplane area coefficient variations.

DWLI
I

I
I

I
I
I

Fig. 24.17- Effect on section shape of increasing the forward waterplane area coefficient.

the bow and gives favourable changes to the hydrodynamic coefficients in this region,
again at the expense of increased wave excitation. If the displacement is kept
constant it also leads to more favourable section shapes with higher deadrise angles
as shown in Fig. 24.17.
A large forward waterplane is clearly beneficial, reducing both absolute and
relative motions and the probability of keel emergence. The more favourable
deadrise angles will alleviate slamming when it does occur.
24.5

SUMMARY

In summary these results show that a large ship will generally be more comfortable
than a small one. Increasing ship size will almost always result in improved
seakeeping performance.

474

The effect of hull size and form on seakeeping

[Ch.24

Fig. 24.18- Effect on forward waterplane area coefficient on motions in head waves;
T,; = 3.5, FN = 0.3.

Fig. 24.19- Effect of forward waterplane area coefficient on probability of keel emergence;
T;, = 3.5, FN = 0.3.

If the ship length is already determined, low levels of vertical acceleration can be
achieved with a shallow draught/wide beam hull form. This may, however, suffer
from frequent keel emergence and slamming. Immunity from slamming can best be
achieved by increasing the draught at the penalty of increased vertical accelerations.
It is for the designer to decide on the best compromise for these conflicting
requirements. A large forward waterplane area coefficient is always beneficial.

Bibliography

The following abbreviations are used in the bibliographical details of the items listed.
AM
ASEM
ATTC
BMT
DMVW
DTMB
HMSO
HSV
!MechE
lOS
ISP
JSNAJ
NACA
NEJ
NSRC
NSRDC
ONR
PRS
QJMAM
SRAJ
SRI
STAR
TRINA
TSNAME
USDC
USNI
UCEP

Aerospace Medicine (USA)


Aviation, Space and Environmental Medicine (USA)
American Towing Tank Conference
British Maritime Technology Limited (UK)
First International Symposium on the Dynamics of Marine Vehicles in
Waves, !MechE, London, 1974
David Taylor Model Basin (USA)
Her Majesty's Stationery Office, London (UK)
Hamburgische Schiffbau-Versuchsanstalt (West Germany)
Institute of Mechanical Engineers, London (UK)
Institute of Oceanographic Sciences, Wormley (UK)
International Shipbuilding Progress (Netherhilnds)
Journal of the Society of Naval Architects of Japan
National Advisory Council for Aeronautics (USA)
Naval Engineer's Journal (USA)
Netherlands Ship Research Centre TNO
Navy Ship Research and Development Center (USA) (formerly
DTMB)
Office of Naval Research Hydrodynamics Symposium (USA)
Proceedings of the Royal Society (UK)
Quarterly Journal of Mechanics and Applied Mathematics (UK)
Shipbuilding Research Association of Japan
Ship Research Institute, Tokyo (Japan)
SNAME Ship Technology and Research Symposium
Transactions of the RoySJ.l Institution of Naval Architects (UK)
Transactions of the Society of Naval Architects and Marine Engineers
(USA)
US Department of Commerce National Weather Service
US Naval Institute
University of California EngineerinKPublication (USA)

476

Bibliography

Aertssen, G. (1963) Service, performance and seakeeping trials on MV Lukuga.


TRINA 105.
Aertssen, G. (1966) Service, performance and seakeeping trials on MV Jordaens.
TRINA 105.
Aertssen, G. (1968) Labouring of ships in rough seas with special emphasis on the
fast ship. SNAME Diamond Jubilee International Meeting.
Aertssen, G. & van Sluys, M. F. (1972) Service, performance and seakeeping trials
on a large container ship. TRINA 114.
Andrew, R. N. & Lloyd, A. R. J. M. (1981) Full-scale comparative measurements of
the behaviour of two frigates in severe head seas. TRINA 123.
Andrew, R. N., Loader, P. R. & Penn, V. E. (1984) The assessment of ship
seakeeping performance in likely to be encountered wind and wave conditions.
RINA International Symposium on Wave and Wind Climate Worldwide,
London.
Baitis, E., Woolaver, D. A. & Beck, T. A. (1983) Rudder roll stabilization for Coast
Guard cutters and frigates. NEJ 95 No.3.
Bales, S. L., Lee, W. T. & Voelker, J. M. (1981) Standardized wind and wave
environments for NATO operational areas. DTNSRDC Report SPD-0919-01.
Bittner, A. C. & Guignard, J. C. (1985) Human factors engineering principles for
minimizing adverse ship motion effects: Theory and practice. NEJ 91 No. 4.
Bledsoe, M.D., Bussemaker, 0. & Cummins, W. E. (1960) Seakeeping trials on
three Dutch destroyers. TSNAME 68.
Blok, J. J. & Huisman, J. (1985) Relative motions and swell up for a frigate bow.
TRINA 127.
Cartwright, D. E. & Longuet-Higgins, M.S. (1956) The statistical distribution ofthe
maxima of a random function. PRS (Series A) 237.
Chuang, S. L. (1970) Investigation of impact of rigid and elastic bodies with water.
NSRDC Report 3248.
Chuang, S. L. & Milne, D. T. (1971) Drop tests of cones to investigate the three
dimensional effects of slamming. NSRDC Report 3453.
Comstock, E. N., Bales, S. L. & Gentile, D. M. (1982) Seakeeping performance
comparison of air capable ships. NEJ.
Crapper, G. D. (1984) Introduction to Water Waves. Ellis Horwood, Chichester.
Cummins, W. E. & Bales, S. L. (1980) Extreme value and rare occurrence statistics
for Northern Hemisphere shipping lanes. 5th STAR, Coronado, California.
de Jong, B. (1973) Computation of the hydrodynamic coefficients of oscillating
cylinders. NSRC Report 145S.
Gerritsma, J. & Benkelman, W. (1967) Analysis of modified strip theory for the
calculation of ship motion and wave bending moments. NSRC Report 96S.
Gerritsma, J. & Benkelman, W. (1971) Analysis of the resistance increase in waves
of a fast cargo ship. Laboratorium voor Scheepsboukunde Report No. 334.
Technische Hogeschool Delft.
Gilhousen, D. B., Quayle, R. G., Baldwin, R. G., Karl, T. R. & Brines, R. 0.
(1983) Climatic Summaries for NOAA Data Buoys. USDC.
Grim, 0. (1959) Oscillation of buoyant two dimensional bodies and the calculation
of the hydrodynamic forces. HSV Report 1171.

Bibliography

477

Hagiwara, K. & Yuhara, T. (1974) Study of wave impact load on ship bow. Japan
Shipbuilding and Marine Engineering 8 No.4.
Hoerner, S. F. (1965) Fluid Dynamic Drag. Published by the author, 148 Busteed
Drive, Midland Park, New Jersey 07432.
Hogben, N., Dacunha, N. M. C. & Olliver, G. F. (1986) Global Wave Statistics.
BMT.
Hogben, N. & Lumb, F. E. (1967) Ocean Wave Statistics. HMSO.
Hsiung, C. C., Friis, D., Milne, W., Peters, G. R. & Weber, H. W. (1983) New
towing facilities at Memorial University of Newfoundland. 20th ATTC.
Kato, H. (1958) On the frictional resistance to the roll of ships. JSNAJ 102.
Kehoe, J. W. (1973) Destroyer Seakeeping- Ours and Theirs. USNI.
Lamb, H. (1932) Hydrodynamics. Cambridge University Press.
Lee, W. T., Bales, S. L. & Sowby, S. E. (1985) Standardized wind and wave
environments for North Pacific Ocean areas. DTNSRDC Report SPD-0919-02.
Lewis, F. M. (1929) Theinertiaofthewatersurroundinga vibrating ship. TSNAME.
Lloyd, A. R. J. M. (1975) Roll stabiliser fins: a design procedure. TRINA 117.
Lloyd, A. R. J. M. (1977) Roll stabiliser fins: inkrference at non-zero frequencies.
TRINA 119.
Lloyd, A. R. J. M. & Andrew, R.N. (1977) Criteria for ship speed in rough weather.
18thATTC.
Lloyd, A. R. J. M., Salsich, J. 0. & Zseleczky, J. J. (1986) The effect of bow shape
on deck wetness in head seas. TRINA 128.
Mack, C. (1966) Essentials of Statistics for Scientists and Technologists. Heinemann
Educational Books, London.
Marine Information and Advisory Service (1982) Catalogue of Wave Data. MIAS,
lOS.
Massey, B. S. (1986) Measures in Science and Engineering, their expression,
relation and interpretation. Ellis Horwood, Chichester!. 11
Nieuwenhuijsen, J. H. (1958) Experimental Investigations on Sea sickness. Drukkerij Schotanus and Jens, Utrecht.
Nordenstrom, N. (1969) Methods for predicting long term distributions of wave
loads and probability of failure for ships, AP11. DNV Report No. 69-22-S.
Ochi, M. K. (1964) Extreme behaviour of ships in rough seas- slamming and
shipping of green water. TSNAME 72.
Ochi, M. K. & Bolton, W. E. (1973) Statistics for prediction of ship performance in a
.
seaway. Parts 1-3. ISP Nos 222, 224, 229.
Ochi, M. K. & Motter, L. E. (1974) Predictions of extreme ship responses in rough
seas of the North Atlantic. DMVW.
O'Hanlon, J. F. & McCauley, M. E. (1974) Motion sickness incidence as a function
of the frequency and acceleration of vertical sinusoidal motion. AM.
O'Neill, M. E. & Chorlton, F. (1986) Ideal and Incompressible Fluid Dynamics.
Ellis Horwood, Chichester.

Pierson, W. J., Tick, L. J. & Baer, I. (1966) Computer based procedures for
preparing global forecasts and wind field analyses capable of using wave data
obtained by spacecraft. ~th ONR, Washington !)C.
Porter, W. (1960) Pressure distribution, added mass and damping coefficients for
cylinders oscillating in a free surface. UCEP Series 82-16.

478

Bibliography

St Denis, M. & Pierson, W. J. (1953) On the motions of ships in confused seas.


TSNAME.
Salvesen, N., Tuck, E. 0. & Faltinsen, 0. (1970) Ship motions and sea loads.
TSNAME.
Schmitke, R. T. (1978) Ship sway, roll, and yaw motions in oblique seas, TSNAME
86.
Sellars, F. H. (1972) Comparison of model and full-scale slamming impact pressure
data. Report MPR-351. MPR Associates Inc., Washington DC.
Shipbuilding Research Association of Japan (1975). Report on the seakeeping
performance of a super high speed container ship. 125th Research Committee,
SRAJ Report No. 211.
Shoenberger, R. W. (1975) Subjective response to very low frequency vibration.
ASEM.
Smith, W. E. (1967) Equation of motion coefficients for a pitching and heaving
destroyer model. NSRC Report 98S.

Stigter, C. (1966) The performance ofU-tanks as a passive anti-rolling device. ISP 13


No. 144.
Takaishi, Y., Matsumoto, T. & Ohmatsu, S. (1980) Winds and waves ofthe North
Pacific Ocean 1964-1973. SRI.
Tanaka, N. (1960) A study on the bilge keels (Part 4 - On the eddy making
resistance to the rolling of a ship hull). JNSAJ 109.
Tasai, F. (1959) On the damping force and added mass of ships heaving and pitching.
JSNAJ 105.
Tasai, F. (1960) On the damping force and added mass of ships heaving and pitching.
UCEP Series No. 82.
Ursell, F. (1949a) On the heaving motion of a circular cylinder in the surface of a
fluid. QJMAM 2 218-231.
Ursell, F. (1949b) On the rolling motion of cylinders in the surface of a fluid.
QJMAM 2 335-353.
Vugts, J. H. (1968) The hydrodynamic coefficients for swaying, heaving and rolling
cylinders in a free surface. NSRC Report 112S.
Wagner, H. (1932) Landing of seaplanes. NACA TN 622.
Walden, D. A. & Grundmann, P. (1985) Methods for designing hull forms with
reduced motions and dry decks. NEJ.
Walters, J. D. (1964) Motion Sickness, Medicine and Travel. Documenta Geigy. J.
R. Geigy SA, Basle.
Whicker, L. F. & Fehlner, L. F. (1958) Free stream characteristics of a family of low
aspect ratio all moveable control surfaces for application to ship design. DTMB
Report933.
Yamamoto, 0. (1984) Fuel saving obtained by a navigation under economical ship
speeds. Nippon Kokan (Japan) Technical Report (overseas) No. 41.

Glossary

Figs G.l and G.2 give a diagrammatic illustration of most of the terms defined.

~
::J
(.)

'6

"'

u:i
u

c.
Q;
c.

;;:"
~ c

ro-

Q)

OQ)
u..C.

Q;
c.

~
.... -L/2-- - - - - -

L/2

Superstructure

Freeboard

Bow

Draught
Forefoot
Keel

Length (between perpendiculars)

Fig. G.l- Hull

After perpendicular (AP)


Beam

defin_~tions:

zero speed; calm water.

Vertical line drawn through the intersection of


the stern and the waterline
Maximum width of the hull measured at the
waterline in calm water at zero speed

480

Glossary

Waterline at zero speed

Sinkage

Running
trim

Fig. G .2- Sinkage and running trim.

Bow
Draught
Fetch
Forecastle
Forefoot
Forward perpendicular (FP)
Freeboard
Ideal fluid
Length
Sheer line
Sinkage

Stem
Stemhead
Stern
Superstructure
Swell

The foremost part of the ship


Local depth of keel below waterline at zero speed
in calm water
Length (in direction of wind vector) of ocean
exposed to wind
The raised part of the hull at the bow
The keel close to the bow
Vertical line drawn through the intersection of
the stem and the waterline
Distance from the water surface to the edge of the
weather deck
A fluid with no viscosity, compressibility or surface tension
Waterline length of ship in calm water at zero
speed
Side elevation of the edge of the weather deck
Increase in draught at forward speed measured
relative to the undisturbed water surface at
midships
The line joining the keel and the deck at the
foremost part of the ship
The top of the stem
The aftermost part of the ship
The part of the ship above the weather deck
Wave system generated by a distant storm and
unrelated to local wind conditions; swells are
usually composed of quite regular long period
waves

Glossary
Transom stern
Trim
Wall-sided
Waterline
Weather deck

481

Stern truncated at a flat transverse section


Steady pitch angle developed at forward speed in
calm water
The sides of the hull are vertical
The plane of intersection of the hull and the water
at zero speed in calm water
The main deck exposed to the weather

Numerical values

Mass density of fresh water


Mass density of sea water
Mass density of air at 15C
Viscosity of fresh water at 15C
Viscosity of air at 20C
Acceleration due to gravity
1 knot

1.0 tonnes/metre 3
1.025 tonnes/metre 3
0.001225 tonnes/metre 3
1.14 x 10- 6 kN seconds/metre2
1.808 x 10-s kN seconds/metre2
9.81 metres/second 2
0.515 metres/second

Index
Added resistance, 398
irregular head waves, 400
regular head waves, 398
wind, 401
Aertssen, 454
Andrew, 11
Andrew, Loader and Penn, 457
Andrew and Lloyd, 277, 282, 454
appendage,231
aspect ratio, 60
Baitis, Woolaver and Beck, 433
Bales, Lee and Voelker, 126
bilge keels, 344
Bittner and Guignard, 425
Bledsoe, Bussemaker and Cummins, 277
Blok and Huisman, 411
BM, 193
bow shape, 424
boundarylayer,58,293
Cartright and Longuet-Higgins, 104
chord, 59
Chuang, 416
Chuang and Milne, 416
circular cylinder
heaving in free surface, 196
in uniform stream, 49
Cobra, HMS, 26
coefficient
hydrodynamic, 173
in heave and pitch equations, 174
in lateral plane equations, 177
hydrostatic: 191
lateral plane, 192
vertical plane, 191
in equations of motion, 158
lift, 61
Comstock et al., 454
confidence levels, 442
box ticking, 446
Student's t test, 442
F test, 442
conformal transformation, 52

coupling, 163, 169, 215, 217


Cox and Lloyd, 345
criteria, 437
deck submergence, 453
equipment, 439
slamming, 452
speed loss, 451
subjective motion, 452
Cummins and Bales, 118
damping, 162,166,204,212,214,217,218,223
decay coefficient, 136, 139
deck wetness, 281, 409, 421, 452
degrees of freedom, 153
de Jong, 196
dimensional analysis, 286, 294
dimension ratio, 289
dipole, 46
doublet, 46, 52
drag, 60
eddy,59,225
electronic filter analogy, 263
equation
Euler's, 29
Bernoulli's, 38
continuity equation, 33
fin stabilised ship in regular waves, 350
Laplace's, 40
ship motions in regular waves, 151, 154, 168
passive tank fluid motion, 380
tank stabilised ship in regular waves, 386
error function, 334
excitation
in regular waves, 181, 235
roll, 188
sway and yaw, 185
vertical, 183, 235
fetch, 94, 106
flow
circular cylinder in uniform stream, 49
elliptical cylinder in uniform stream, 56
laminar, 58,293
separation, 59,293

484
two-dimensional, 29, 64
turbulent, 59, 293
uniform stream, 43
force
exciting, 157
Fourier analysis, 97
freeboard
effective, 410
exceedance, 422
free decay, 139
frequency
encounter, 145
model, 291
natural, 137,235,251, 389
of deck submergence, 412, 452
of keel emergence, 412
of propeller emergence, 412
of slamming, 418,451
wave, 73
Friesland class destroyer, 162, 165, 237
Froude-Kriloff hypothesis, 170
Froude number, 287,292
Froude, William, 26, 307
fluid
ideal, 28
inviscid, 29
Gerritsma and Beukelman, 170,399
Gilhousen et al., 129
GM, 194
green seas, 421
Grim, 196
GZ, 194
Hagiwara and Yuhara, 416
harmonic response, 133
heading, 144, 292
helicopters, 437
histograms, 327
Hoerner, 63
Hogben and Lumb, 122, 124
Hogben, Dacunha and Olliver, 124,458
hull shape, 468
hull size, 461
incidence, 60
Kato, 230
Kehoe, 454
Lamb, 66
Lee, Bales and Sowby, 127
Lewis forms, 206
heaving, 212
hydrodynamic properties, 211
permissible forms, 208
rolling, 217
swaying, 214
lift, 60
coefficient, 61
curve slope, 61
linear system, 133
Lloyd, 353, 356

Index
Lloyd and Andrew, 454
Lloyd, Salsich and Zseleczky, 423
Loader, II
local hydrodynamic properties: 196
measurement, 218
Mack, 442
mapping function, 53, 55
Marshfield, II
Martin, 345
mass
added, 161, 165, 204, 212,214,217,218
moment of inertia, 155
virtual, 162
Massey, 287
metacentric height, 194
MIAS, 129
model experiments
ballasting, 308
beaches,297,300
bifilar suspension rig, 310
carriage, 297
compound pendulum rig, 309
forced oscillation, 158, 165
irregular waves, 296,318
model materials, 307
model restraint systems, 299
instrumentation, 303
open water, 297
regular waves, 294, 312
scaling Jaws, 286, 289
seakeeping basin, 298
slamming, 414, 417
tank wall interference, 323
trimming, 308
towing tank, 297
wave makers, 300, 319
moment of inertia
added mass, 165
mass, 155, 165, 290, 309
product, 156
virtual, 165
multipole, 48
Nieuwenhuijsen, 425
NOAA, 129
non linearity, 133, 166, 223, 262, 275
Nordenstrom, 122
Ochi, 417
Ochi and Bolton, 104
Ochi and Motter, 454
O'Hanlon and McCauley, 426
O'Neill and Charlton, 28
operational effectiveness, 455
outreach, 59
PAT-86 seakeeping computer program, 27
Pierson, Tick and Baer, 126
Porter, 196
potential
complex, 52

Index
velocity, 35
Probability: 327
density function: 329
Gaussian, 333
Normal, 333
Rayleigh, 337
of deck submergence, 411
of keel emergence, 411, 466, 471,472
of occurrence, 410
of propeller emergence, 411
of slamming, 417
joint, 342
propeller
advance coefficient, 404
characteristics, 403
efficiency, 405
racing, 409
self propulsion, 405
thrust coefficient, 404
torque coefficient, 404
questionnaires, 440
radius of gyration, 165,291,310
Reynolds number, 230,287, 293
Ridjanovic, 346
StDenis and Pierson, 26, 263
Salvesen, Tuck and Faltinsen, 170
scatter diagrams, 124
Schmitke, 170, 225,227,230
sea state code, 121
second order, 133
Sellars, 413
Shipbuilding Research Association of Japan, 454
ship motions
absolute, 250
accelerations, 257, 269,295,462,470, 471,472
acclimatisation, 425
axes, 151
beam waves, 243
contouring waves, 462
coupling, 163,169,235
definitions, 151
effect of hull shape, 468
effect of hull size, 461
energy spectrum, 266
following waves, 240
head waves, 235
heave, 152,235,236,244,462
irregular waves, 263, 296
lateral force estimator, 259, 433
matching wave spectrum and transfer function,
271
measurements, 279
motion induced interruptions, 433
motion sickness incidence, 426
oblique waves, 243, 252
phase shift due to wave probe location, 313
pitch, 153, 235, 236

regular waves, 151, 156, 234, 294


relative, 257, 463

485
notional, 411
resonance, 248
rms values, 268
roll, 153,243,247,248
short crested waves, 272
subjective motion, 429, 452
surge, 152,235
spectral moments, 270
sway, 152,243,243,249
transfer functions, 234, 312
velocities, 257, 269
wave reflection, 314
yaw, 153,243,249
Shoenberger, 429
sink, 44,52
skin friction, 228, 293
slamming, 281,409,413,451
critical velocity, 417
drop tests, 414
frequency, 418
pressure, 418
probability, 417
Smith, 162, 165, 237
source, 44, 52
speed loss
involuntary, 398, 406
voluntary, 437,449
criteria for, 451
spring-mass system, 132
Stabilisation, 343
active fins, 349
boundary layer losses, 354
constraints on design, 350
control systems: 362
gain margin, 376
overall gain, 370
phase margiil,~B70
sensitivity setti'ngs, 365
speed dependant gain, 363
stability, 372
transfer function, 362
design recommendations, 358
equations of motion for fin stabilised ship, 350
fin-bilge keel interference, 357
fin-fin interference, 355
fin servo, 370
forced rolling trial, 366
hydrodynamic losses, 353
overall effectiveness, 358
performance, 377
sway-yaw effect, 360
bilge keels, 344
passive tanks, 377
design, 390
dimensions, 392
equations of motion for tank fluid, 380
equations of motion for tank stabilised ship,
386
fluid depth, 393
flume tank, 377
Joss of metacentric stability, 393
mass of working fluid, 395

486
performance, 397
stabilising moment, 393
tank damping, 389, 395
tank natural frequency, 389
U tube tanks, 379
stagnation
point, 51, 57, 58
pressure, 40
stall angle, 61
standard deviation, 97
Stigter, 380
stream function, 41
strip theory, 170
superposition, 64
swell-up, 257, 411
system with no stiffness, 141
Takaishi, Matsumoto and Ohmatsu, 126
Tanaka, 228
Tasai, 170, 196
Taylor wake fraction, 404
transformation, 206
transition point, 59, 293
trials, 277
courses, 282
forced rolling, 366
run lengths, 282
Unimak USCG, 413
units, 23
Ursell, 26, 196
variance, 97, 101,269,270
viscosity, 58
coefficient of, 59
Vugts, 165
Wagner, 416
Walden and Grundmann, 454
Walters, 426
waves
amplitude, 64, 94
significant single, 96
bandwidth, 104, 108
breaking, 94
buoys, 129, 279
celerity, 64, 71, 73
characteristics, 73
energy, 86
fully developed, 94
frequency, 73
average, 103
encounter, 145
generation, 93
group velocity, 89
height, 64, 95
characteristic, 107, 108

Index
significant, 96, 105, 339
length, 64, 291
long crested, 113
measurement, 279
number, 70, 292
ocean, 93
Atlantic, 126, 127
atlases, 124
Baltic Sea, 126
Black Sea, 126
fair weather bias, 126
Gulf of Mexico, 126
hindcasting, 126
measured data, 129
Mediterranean Sea, 126
Pacific, 126, 127
North Sea, 126
scatter diagrams, 124
statistics, 121
visual observations, 121, 124
orbits, 74
period, 65, 95
average, 103, 107
mean period of peaks, 104, 108
mean zero crossing period, 104, 108
modal, 108
model, 292
phase velocity, 71
potential function, 65
pressure, 66, 86
reflection, 314
regular, 64
short crested, 114
slope, 65, 72
spectral
moments, 101, 103, 107
ordinate, 100
spectrum, 99
Bretschneider, 107, 112
directional, 114
encountered, 264
idealised, 106
ITTC (two parameter), 107
JONSWAP, 109, 112
slope spectrum, 112
spreading, 113
steepness, 65
synthesis, 98
swell, 94
surface profile, 66
Whicker and Fehlner, 62
whipping, 409
wind, 93, 127,401
WMO, 121
Yamamoto, 454