Cavitation of Hydraulic Machinery

SERIES ON HYDRAULIC MACHIWEW - VOL.
1 I
Hydraul
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
Copyright World Scientific Publishing Co.

. ..+$
7
r
+
I.4
- ImperiaI College Press

-
Provided by IHS Markit under license with WILEY Licensee=Jacobs Enterprise Wide/3219500143, User=Roncal Jaico, Julio Cesar
No reproduction or networking permitted without license from IHS Not for Resale, 02/26/2019 09:03:23 MST
Cavitation of
Hydraulic Machinery
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

HYDRAULIC MACHINERY BOOK SERIES
- Hydraulic Design of Hydraulic Machinery
Editor: Prof H Radha Krishna
- Mechanical Design and Manufacturing of Hydraulic
Machinery
Editor: Prof Mei Z Y
- Transient Phenomena of Hydraulic Machinery
Editors: Prof S Pejovic, Dr A P Boldy
- Cavitation of Hydraulic Machinery
Editor: Prof Li S C
- Erosion and Corrosion of Hydraulic Machinery
Editors: Prof Duan C G, Prof V Karelin
- Vibration and Oscillation of Hydraulic Machinery
Editor: Prof H Ohashi
- Control of Hydraulic Machinery
Editor: Prof H Brekke
The International Editorial Committee (lECBSHM):
Chairman: Prof Duan C G Secretary: Prof Li S C
Treasurer: Dr R K Turton Dr A P Boldy
Committee Members:
Prof H Brekke (Norway) Prof V P Chebaevski (USS)
Prof E Egusquiza (Spain) Prof M Fanelli (Italy)
Dr H R Graze (Australia) Prof R Guarga (Uruguay)
Prof P Henry (Switzerland) Dr H B Horlacher (Germany)
Prof V Karelin (USS) Prof G Krivchenko (USS)
Prof Li Sheng-cai (China) Prof Liu D K (China)
Prof M Tadeu de Almeida (Brazil) Prof C S Martin (USA)
Prof M Matsumura (Japan) Prof Mei Zu-yan (China)
Prof A Mobarak (Egypt) Prof H Murai (Japan)
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
Prof H Netsch (Canada) Prof H Ohashi (Japan)

Prof S Pejovic (Yugoslavia) Prof D Perez-Franco (Cuba)
Prof H Petermann (Germany) Prof H C Radha Krishna (India)
Prof C S Song (USA) Prof C Thirriot (France)
Prof Hans Ingo Weber (Brazil) Prof G Ziegler (Austria)
Honorary Members:
Prof B Chaiz (Switzerland) Prof J Raabe (Germany)
Committee Chairman: C. G. Duan Series Editor: S. C. Li
Cavitation of
Hydraulic Machinery
Editor
S. C. Li
University of Warwick, U.K.
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
-fl^
Provided by IHS Markit under license with WILEY
No reproduction or networking permitted without license from IHS
Imperial College Press
Licensee=Jacobs Enterprise Wide/3219500143, User=Roncal Jaico, Julio Cesar
Not for Resale, 02/26/2019 09:03:23 MST
Published by
57 Shelton Street
Covent Garden
London WC2H 9HE
Distributed by
World Scientific Publishing Co. Pte. Ltd.
P O Box 128, Farrer Road, Singapore 912805
USA office: Suite IB, 1060 Main Street, River Edge, NJ 07661
UK office: 57 Shelton Street, Covent Garden, London WC2H 9HE
British Library Cataloguing-ln-Publlcation Data

A catalogue record for this book is available from the British Library.
CAVITATION OF HYDRAULIC MACHINERY

Copyright © 2000 by Imperial College Press
All rights reserved. This book, or parts thereof, may not be reproduced in any form or by any means,
electronic or mechanical, including photocopying, recording or any information storage and retrieval
system now known or to be invented, without written permission from the Publisher.
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
For photocopying of material in this volume, please pay a copying fee through the Copyright
Clearance Center, Inc., 222 Rosewood Drive, Danvers, MA 01923, USA. In this case permission to
photocopy is not required from the publisher.
ISBN 1-86094-257-1
Printed in Singapore by Uto-Print

CONTENTS
Preface xiii
Foreword of the Editor xv
Contributing Authors xix
1 Introduction 1
SCLi
1.1 Cavitation 1
1.1.1 Discovery 1
1.1.2 Classification 1
1.2 Hydraulic Machinery and Cavitation 3
1.2.1 Problems Caused by Cavitation 3
1.2.2 Combating Cavitation 5
References 8
2 Cavitation and Cavitation Types 9
A J Acosta
2.1 Cavitation Phenomenon 9
2.1.1 Cavitation 9
2.1.2 Hydrodynamic Cavitation 10
2.2 Types of Cavitation 11
2.2.1 Cavitation Index 11
2.2.2 Scale Effects and Cavitation Types 12
2.3 Cavitation Effects 24
2.3.1 General Phenomena 24
2.3.2 Influences on Machine Performance 25
2.3.3 System Stability 28
2.4 Cavitation Nucleation and Inception 30
2.4.1 Inception 30
2.4.2 Nuclei Measurement 34
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
2.4.3 Concluding Remarks 38

References 40
3 Bubble Dynamics 47
Part 1: Single Bubble 47
A Shima
3.1 Rayleigh Analysis 47
3.1.1 Besant's Problem 47
V
vi Contents
3.1.2 Rayleigh's Solution 48

3.2 Vaporous and Gaseous Bubbles 49
3.2.1 Classification 49
3.2.2 Stability 49
3.3 Viscosity and Compressibility Effects 50
3.3.1 Effects of Viscosity and Surface Tension 50
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
3.3.2 Compressibility Effect 50
3.4 Bubble Rebound 53
3.4.1 Introduction 53
3.4.2 Studies of Rebound 53
3.5 Nonspherical Collapse and Micro-jet Formation 54
3.5.1 Collapse near Solid Wall 54
3.5.2 Collapse in Contact with Solid Wall 55
3.6 Pressures Generated at Collapse 57
3.6.1 Impact Pressure and Their Modes 57
3.6.2 Temperature Effect 60
References 61
Part 2: Multi-Bubbles (Stochastic Behaviour) 65
SCLi
3.7 Origins of Bubble Stochasticity 65
3.7.2 Bubble-Boundary Interaction 67
3.7.3 Bubble-Bubble Interaction 73
3.7.4 Bubble-Flow Field Interaction 94
3.7.5 Remarks 115
3.8 Stochastic Models of Cavitation Bubbles 117
3.8.2 Single-Event Model 117
3.8.3 Multi-Event (Cluster) Model 118
3.8.4 Comprehensive Model 120
3.9 Power Spectrum 121
3.9.2 Spectrum of Single-Bubble Collapse 121
3.9.3 Spectrum for Sequence of Single Events 127
3.9.4 Spectrum for Sequence of Clusters 137
3.9.5 Spectrum for Comprehensive Sequence 148
References 153
157

Contents vii
4 Cavitating Flow
4.1 General Features 157
H Murai
4.1.1 Sheet Cavitation 157
4.1.2 Travelling Bubble Cavitation 160
4.1.3 Vortical Cavitation 161
4.1.4 Hydraulic Loss Caused by Cavitation 161
4.2 Hydrofoil and Hydrofoil Cascade 166
H Murai
4.2.1 Cavitation Characteristics of Hydrofoil 166
4.2.2 Cavitation Characteristics of Conventional
Airfoil Sections 170
4.2.3 Cavitation Characteristics of Ogival Hydrofoils 171
4.2.4 Cavitation Characteristics of Hydrofoil Cascade 175
4.2.5 Cavitation Characteristics of Conventional Airfoil
and Ogival Profiles 176
4.2.6 Hydrofoil Profile Suitable for Decelerating
and Accelerating Cascades 180
4.2.7 Computer Simulation of Partially Cavitating Foil 182
4.2.8 Supercavitating Hydrofoil 184
4.2.9 Supercavitating Hydrofoil Cascade 186
4.3 Control Valves 187
E Outa
4.3.1 General Features of Control Valve Cavitation 187
4.3.2 Cavitation Pictures of Globe Valve Flows 194
4.3.3 Cavitation Inception due to Vortex Growth 197
4.3.4 Cavitation Erosion and Anti-Cavitation Valves 202
References 205
5 Cavitation P h e n o m e n a in Hydraulic Machinery 211
5.1 General Features of Turbine Cavitation 211
H Tanaka
5.1.1 Cavitation in Francis Turbines 211
5.1.2 Cavitation in Propeller Turbines 220
5.1.3 Cavitation in Pelton Turbines 225
5.1.4 Cavitation in Francis Pump-Turbines 226
5.2 General Features of Pump Cavitation 229
R K Turton
5.2.2 General Effect on Pumps 229
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

viii Contents
5.2.3 Net Positive Suction Head 232

5.2.4 Definition of Critical NPSHR 235
5.2.5 Implications for Pump Design 236
5.2.6 The Role of the Inducer 240
5.3 Pump Cavitation Similarity 242
V Chebaevsky and V Petrov
5.3.1 Problem Nature and Similarity Conditions 242
5.3.2 Thermodynamic Property Simulation 248
5.4 Cavitation Detection Techniques 251
E Egusquiza
5.4.2 Generation of Noise and Vibration 252
5.4.3 Propagation 253
5.4.4 Background Noise 254
5.4.5 Cavitation Detection in Frequency Domain 255
5.4.6 Cavitation Detection in Time Domain 259
5.4.7 Cavitation Detection with Time-Frequency Analysis 263
References 265
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
6 Cavitation D a m a g e t o Hydraulic Machinery 269

6.1 General Mechanism of Cavitation Damage 269
Y Iwai and T Okada
6.1.2 Cavitation Bubble Collapse Pressures and Damage 269
6.1.3 Correlation between Erosion Resistance
and Mechanical Property 276
6.2 Cavitation Damage in Turbines 277
SCLi
6.2.2 Propeller and Kaplan Turbines 285
6.2.3 Francis Turbine 286
6.2.4 Pelton Turbine 290
6.2.5 Cavitation Guarantee 292
6.3 Cavitation Damage in Pumps 295
SCLi
6.3.2 Axial Flow Pumps 295
6.3.3 Centrifugal Pumps 296
6.3.4 Pump-turbines 298

Contents ix
6.3.5 Cavitation Guarantee 299

6.4 Silt-Laden Water Effect 302
SCLi
6.4.2 Silt Erosion 302
6.4.3 Synergism of Silt and Cavitation Erosions 307
6.5 High Resistance Materials 314
SCLi
6.5.2 Fused Materials 314
6.5.3 Non-fused Materials 330
6.6 Repair of Machine 334
SCLi
6.6.2 Cause of Damage 334
6.6.3 Main Concerns in Repair 336
6.6.4 Examples 343
References 353
7 Cavitation Caused Vibrations 359
7.1 Cavitation Pressure Pulsation in Turbines 359
7.1.1 Blade Cavity Pulsations 359
J Sato
7.1.2 Blade Wake Cavitation Pulsations 364
J Sato
7.1.3 Draft Tube Vortex Core Cavitation Pulsations 364
P Henry
7.2 Cavitation Induced Pulsations in Pumps 369
R K Turton
7.2.2 Characteristics of Cavitation Induced Pulsations 372
Y Tsujimoto
7.2.3 Mechanisms of Cavitation Induced Pulsations 376
Y Tsujimoto
7.2.4 Cavitation Characteristics - Mass Flow Gain Factor
and Cavitation Compliance 380
Y Tsujimoto
7.3 Influence of Operating Conditions 383
P Henry
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

x Contents
7.3.1 Preliminary Discussion 383

7.3.2 Part Load Precession 383
7.3.3 80% Load Oscillations 388
7.3.4 Draft Tube Free Oscillations 388
7.3.5 Auto-oscillations 398
7.3.6 Influence of the Test Head 401
7.3.7 Thoma Number 403
7.4 Cavitation Resonance in Hydraulic Machinery Installations 405
P Henry
7.4.2 Prototype Installation 405
7.4.3 Model Tests 409
7.4.4 Stability of the Prototype Installation 413
References 417
8 U n s t e a d y Cavitation Flows Caused by Machine Transients 423
Part 1: Turbine Transients 423
C S Martin
8.1 Introduction 423
8.2 Types of Turbine Cavitation 424
8.2.1 Francis Turbines 424
8.2.2 Kaplan Turbines 432
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
8.3 Draft-Tube Column Separation Incidents 433
8.4 Physical Modelling of Cavitating Transient Flows 436
8.5 Two-Phase Flow Modelling in Conduits 440
8.5.1 Acoustic Velocity 441
8.5.2 Slug Flow 444
8.6 Analytical Modelling of Cavitating Transient Flows 444
References 445
Part 2: P u m p Transients 451
H Tsukamoto
8.7 Introduction 451
8.8 Transient Cavitation in Discharge Lines 451
8.8.1 Transient Cavitation Type 451
8.8.2 Water Column Separation 453
8.9 Transient Cavitating Flow in Turbopumps 453
8.9.1 True Total Pressure Rise 453
8.9.2 Transient Behaviour of a Cavitating Pump 455
8.9.3 Transient Characteristics of Pump 457

Contents xi
8.10 Computer Simulation of Cavitating Transient Flows

in Pump System 460
8.11 Concluding Remarks 461
References 463

--``,```,,`````
This page is intentionally left blank
--``,```,,``````,,,,

PREFACE
It is a privilege to be asked to write a prefatory note on this remark
able reference book, the latest in the Hydraulic Machinery book series. It is
unusually wide in scope, ranging from basic research to engineering applica
tions. Its publication marks the achievement of a large team of specialists
in cavitation research and in the design and operation of hydro-turbines and
storage pumps.
Although the problem of cavitation in machines is an old one, dating back
at least to the time of Galileo (16th century), it has not yet been completely
solved. This is because hydraulic machines are faced with increasingly se
vere performance requirements demanded by economic pressures. The need
to provide cheaper designs involving smaller and more powerful machines
rotating at higher speeds continues to heighten the danger of cavitation.
Despite improved techniques that have been developed for detecting and
measuring cavitation, and thereby assessing the damage capacity of cavitat-
ing flows, they do not necessarily solve the problem, but may nevertheless
provide useful guidance for avoiding trouble.
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
The International Editorial Committee of the book series decided to ini

tiate an up to date review of the state of the art of coping with practical
cavitation problems in hydraulic machinery. An imposing team of twenty
specialists was formed to cover topics ranging from basic cavitation research,
machine design and performance, cavitation and abrasion damage and its
repair, and also cavitation induced machine vibration and transient oscil
lations. A study of the chapter headings will show the particularly wide
variety of topics covered in one volume; it is also a welcome trend that they
range from basic science/applied physics to engineering operations. This
has required a well conceived plan and efficient organisation by a patient
and indefatigable Editor who has also written several of the sections. With
exemplary co-operation he and the other authors have been engaged for
twelve years in producing this comprehensive review of current knowledge.
Congratulations and thanks are therefore due to all concerned with launch
ing and completing this monumental task. I am sure that all readers will be
most grateful to the international team's outstanding achievement in assem
bling this treasure trove.
S.P. Hutton
Emeritus Professor
University of Southampton
xiii
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

Foreword of the Editor
The present book Cavitation of Hydraulic Machinery is a volume in the
Hydraulic Machinery book series. It covers cavitation related subjects from
fundamental mechanisms to practical applications in turbines and pumps 1 .
Cavitation is one of the most unwanted phenomena in hydraulic machin
ery despite the fact that it does have some positive effects in other fields. Al
though a great advance in understanding this phenomenon has been achieved
in the last 100 years, our knowledge about cavitation is not good enough to
precisely predict and completely solve this problem. Nowadays, most of the
cavitation related work in hydraulic machinery still, to a great extent, relies
on previous experience, model test and prototype observation.
After introducing cavitation and its relationship to hydraulic machines,
the rest of this book falls naturally into two parts. The first part, Chapters
2, 3 and 4, deals with the fundamental knowledge necessary for understand
ing the cavitation involved in hydraulic machines. It includes: cavitation
types, scale effects, nucleation and inception, single bubble dynamics, multi-
bubble dynamics (bubble-bubble, bubble-boundary and bubble-flow inter
actions), stochastic models of cavitation bubbles, noise spectra, cavitating
flows of hydro-foil and cascade, cavitating characteristics of valves and other
hydraulic elements. The second part, Chapters 5, 6, 7 and 8, deals with
cavitation related themes in turbines and pumps such as cavitation features,
similarities, cavitation detection techniques, cavitation damage mechanism
and features, synergism of cavitation with silt erosion, material resistances to
cavitation damage, cavitation-damage repair, cavitation-induced pressure-
pulsations, cavitation resonance, cavitating transient flow and computer sim
ulation. As it is impossible to include all relevant subjects in a single volume,
careful selection has been necessary. Only those basic concepts and new de
velopments which are not covered by existing books and review articles 2 are
given in-depth treatment. Extensive lists of references and footnotes are
thus included to support the presentation and assist readers who want to
dig deeper.
A total of 17 authors from 7 countries, all experts in their chosen fields,
have made contributions to the book. By drawing upon wide resources and
'As agreed at the Inaugural Meeting of the International Editorial Committee, the
Book Series would deal only with hydro-turbines and pumps.
2
Such as Cavitation by Knapp/Daily/Hammitt (1970), Cavitation by Young (1989),
Cavitation and Bubble Dynamics by Brennen (1995) and 'Cavitation in Fluid Machinery
and Hydraulic Structure' (Ann. Rev. Fluid Mech) by Arndt (1981) etc.
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
xv
xvi Foreword of the Editor
experiences from North America, Europe, China, Russia, Japan etc, the
book aims to give a more balanced view of the various topics. The editor
is greatly indebted to all the authors for their valuable contributions and
effective cooperation over last 12 years. Without their great efforts, the
book would not have been possible. In particular, I would like to say a few
words in memory of Professor Pierre Henry who died on 26th April 1994
from cancer. We miss him very much, he was only 54 and had built up
the Lausanne Laboratory and established a brilliant reputation in unsteady
operating conditions due to cavitating vortices. The sections of §7.1.3, §7.3
and §7.4 he wrote were mainly based on that remarkable work.
The structure and content of this book was originally proposed by the
editor in 1983 as a monograph to be written in collaboration with Professor
Fredrick G. Hammitt (University of Michigan, USA) as an effort to bridge
the existing gap between fundamental cavitation phenomena and cavitation-
related subjects in hydraulic machines. This book, together with another
proposed joint book on the topic of transients in hydraulic machines, sub
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
sequently provided the inspiration for the Hydraulic Machinery book series.
The International Editorial Committee (IECBSHM) was established for this
purpose in 1986 in Beijing. Unfortunately, a deterioration in his health
denied Professor Hammitt the opportunity to participate in the book. Nev
ertheless, his enthusiastic response to the editor's initial motion was a vital
support and encouragement. I would like to take this opportunity to say a
special word of thanks in memory of Fred, who passed away in 1989.
In order to provide up-to-date information to our readers, authors were
allowed to amend their contributions right up to the last minute, squeez
ing the editing and compilation into the last few months. This presented
me with a very complicated and intensive job. The support of the follow
ing persons was invaluable in achieving this. Many experts were invited
to review manuscripts. Their constructive criticism, comments, discussion
and suggestions are highly valued by the authors and the editor. They are
Professor Peter W Carpenter (Warwick University, UK), Professors Allan
Acosta, Dr. M L Billet and Professor C E Brennen (California Institute of
Technology, USA), Dr. R K Turton (Loughborough University, UK) and Mr.
Harland Topham (Water Turbine and Pump Consultant, UK). Thanks also
go to Professor Duan C G and Dr A P Boldy of IECBSHM for their support;
to Professor H Murai for his assistance in coordinating with some authors in
the early stage; to Dr. Tony Price and Dr. Wanda Lewis (Warwick University,
UK) for their advice; to Mrs. Wendy Murray (IT Services, Warwick Univer
sity) for her assistance in preparing computer-editable source files from some

Foreword of the Editor xvii
manuscripts; and to Mr. Alan Hulme (the Engineering Computer Manager)

and the secretaries for their effective support.
The constant support and encouragement received from Professor David
Anderson (the Chairman, School of Engineering) and Professor Peter W Car
penter (the Head of the Division, Civil & Mechanical Engineering), and the
advice given by Dr S P Hutton (Professor Emeritus, University of Southamp
ton, UK), who also wrote the preface, were greatly appreciated.
During the publishing process, the advice received from the commission
ing editors, Dr John Navas, Mr Anthony Doyle, and the desk editor, Mr
Yeow Hwa Quek, was vital that made this volume eventually available to
our readers.
My work on the book is supported in part by the National Research
Grant No. 87022190 of Water and Power (China), the EPSRC (Engineer
ing and Physic Science Research Council, UK) Grants (GR/F57977 and
GR/L74729), and the internal funds of the School of Engineering, Warwick
University (UK).
The figures in the book have come from a variety of sources. The ac
knowledgement appears in the caption as a parenthetical reference keyed by
the name(s) of the original author(s) and a number to the list of references
at the chapter end where the source is cited in full. A particular thank you
goes to Mr. R Stahel (Sulzer Hydro Ltd, Switzerland) who kindly supplied
me with the original photos of Figures 6.14, 6.23, 6.24, 6.26 and 6.27.
Finally, the editor is in great debt to his family members for their un
derstanding, support and forbearance during the lengthy process of writing
and editing this book.
LI S C, Editor
University of Warwick
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

Contributing Authors
Shengcai Li, Professor,

IRCHM (Beijing);
and North China University of Water Resources,
China.
Principal Research Fellow (Warwick University,
UK). Founding member of lECBSHM. Distinction-
graduation specialised in hydraulic machinery
from the Beijing University of Hydraulic Engi
neering. PhD (self-guided) from Warwick Uni
versity. Research Engineer, Ministry of Water
Resources & Power, 1964-77. Since 1977, asso
ciated with the North China University of Wa
ter Resources. Visiting Scientist to the Univer
sity of Michigan (US A) 1981-83. Research areas:
cavitation in hydraulic machinery, stochastic be
haviour of cavitation, transient flows of flow sys
tems, developments of pioneering turbines and
governors, turbine-test techniques.
Allan Acosta, Hayman Professor Emeritus,

Division of Engineering & Applied Science,
California Institute of Technology,
Pasadena, California 91125, USA.
Born in 1924, Anaheim, California, USA. B.S.
1945, M.S. 1948 and PhD 1952 from California
Institute of Technology; then, Assistant Profes
sor, Associate Professor and Professor. Execu
tive Officer for Mechanical Engineering 1988-93;
Richard L. and Dorothy M. Hayman Professor of
Mechanical Engineering 1990-93. Research ar
eas: cavitation flow in turbomachines, hydrody
namics, heat transfer. Life Fellow of the Amer
ica Society of Mechanical Engineering, Fellow of
the America Association for Advanced Science,
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
XIX
xx Contributing Authors
Member of the National Academy of Engineers

(USA).
Vadim Chebaevsky, Professor,

Moscow State University of Environmental En
gineering, Moscow, Russia.
Doctor of Technical Science. Graduated from
the Moscow Aviation Institute in 1949. Author
and co-author of 5 monographs, a branch ref
erence book, a textbook, training aids, 90 sci
entific articles and 56 inventions. Scientific ac
tivities: improvement of cavitation character
istics for blade pumps with different working
media; hydraulic improvement and cavitation
erosion/tear prediction of large blade pumps;
and, automation of technical process for land-
reclamation stations.
Eduard Egusquiza, Professor,
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
Department of Fluid Mechanics,
Polytechnic University of Catalonia, (UPC) at
Barcelona, Spain.
Born in Barcelona. Industrial Eng. and Dr.
Eng. at UPC. Assistant and Associate Professor
in 1977-83 at the Faculty of Industrial Engineer
ing (ETSEI) in Terrassa. Professor in 1983 at
the University of Oviedo and 1988 at the ETSEI
in Barcelona. Research on unsteady flows, flow
induced vibrations and condition monitoring in
turbomachinery (axial flow fans and hydraulic
machinery).

Contributing Authors xxi
Pierre Henry, (Deceased) Professor,

Federal Institute of Technology,
Lausanne, Switzerland.
Professor Henry died on the 26th April 1994
from cancer. The Sections, §7.1.3, §7.3 and §7.4,
he prepared are included in this volume as a trib
ute to Pierre.
Yoshiro Iwai, Professor,

Department of Mechanical Engineering,
Pukui University, Fukui, Japan.
Born in Japan in 1949. Graduated from the
Graduate School of Kyoto University in 1977
and received a Doctor degree in Engineering from
Kyoto University in 1980. Professor at the De
partment of Mechanical Engineering, Fukui Uni
versity since 1991. Research on wear and cavi-
tation erosion.
C. Samuel Martin, Professor,

School of Civil Engineering,
Georgia Institute of Technology,
Atlanta, Georgia 30332-0355, USA.
Born in USA in May 22, 1936. Ph.D (1964) and
M.S. (1961) from Georgia Institute of Technol
ogy; B.S. (1958) from Virginia Polytechnic Insti
tute and State University. Assistant (63-67), As
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
sociate (67-74) and full Professor (74-present) at

Georgia Institute of Technology. Guest Profes
sor (Technical University of Munich, West Ger
many, 1984-85; University of Karlsruhe, Ger
many, 1970-71). Ford Foundation Faculty Res
ident, Harza Engineering Company (1966-67).

xxii Contributing Authors
Designer of Hydraulic Machinery, Newport News

Shipbuilding and Dry Dock Company (1959-60).
Professional Engineer, State of Georgia. Re
search experience: Cavitation characteristics of
control valves using signal analysis, development
of numerical methods for representing charac
teristics of hydraulic machinery (pump-turbines,
Francis turbines and Kaplan turbines), hydraulic
model testing (pump intakes, bifurcations and
spillways). Honours: Alexander von Humboldt
U.S. Senior Scientist Award (1984-85), Ameri
can Society of Mechanical Engineers Fellow Elec
tion (1983), American Society of Civil Engineers
J. C. Stevens Award (1977), American Society of
Mechanical Engineers John R. Freeman Scholar
(1970-71), Fulbright Travel Grantee (1970-71),
Ford Foundation Faculty Fellow (1966-67) and
Ford Foundation Fellowship (1961-62).
H Murai, Professor Emeritus,

Tohoku University, Sendai, Japan.
Born in Hiroshima in 1922. Graduated from
Tohoku University in 1944. Master (1946) and
Doctor (1957) of Engineering from Tohoku Uni
versity. Lecturer (1946), Assistant Professor (1947),
Professor (1960-1985) and Director (1974-1979)
of the Institute of High Speed Mechanics. Pro
fessor of Hiroshima Institute of Technology from
1985 to 1992. Technical advisor (1968-1985 and
1995) of National Aerospace Laboratory. Exec
utive Director (1977-1978) of the Japan Society
of Mechanical Engineers. President (1985-1986)
of the Visualisation Society of Japan. Member
(1988) of the Engineering Academy of Japan.
Research Fellow (1992-1995) of Tsuru Promo
tion Society for Education & Research.
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

Contributing Authors xxiii
Tsunenori Okada, Professor Emeritus,

Pukui University, Pukui, Japan.
Born in Japan in 1931. Graduated from Osaka
City University in 1954 and received a Doctor
degree in Engineering from Kyoto University in
1967. Professor at the Department of Mechan
ical Engineering, Fukui University, in 1971-97.
Visiting Professor at the University of Michigan
in 1979-1980. Research on cavitation erosion.
Eisuke Outa, Professor,

Waseda University, Tokyo, Japan.
Born in Kobe, Japan. B.S. (1963), M.S. (1965)
and Dr. Engr. (1976) from Waseda Univer
sity. Assistant and Associate Professor in 1968-
1975 at the Department of Mechanical Engi
neering, Waseda University; and, Professor since
1975. Research interests include: experimen
tal and computational studies of unsteady flows,
turbomachinery noise, high speed flow and shock
waves in gas and multi-phase media, cavitation
in control valves and pump cascades, noise re
duction of supersonic jet flow etc.
Vladimir Petrov, Professor,

Russian Academy of Space Technology,
Korolyov Town, Russia.
Academician of Russian Academy of Space Tech
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
nology, Doctor of Science. Graduated from the

Moscow Bauman Technical University in 1958.
Fields of research: hydrodynamics of one and
two phase fluid flows with cavitation, calcula
tion and design of hydraulic equipments of fuel
supply systems for space industry. Author and

xxiv Contributing Authors
co-author of 7 monographs, 70 scientific articles

and 50 inventions.
Joshiro Sato, Senior Chief Engineer,

Hitach Ltd. Hitach Works,
Japan.
Born in Japan in 1939. BS (1962) and Dr. Eng.
(1991) from Tohoku University. Engaged in re
search, development, design and project man
agement of pump-turbines and conventional tur
bines in the Turbine Design Department since
1962. Specialised in cavitation and silt erosion.
Akira Shima, Professor Emeritus,

Tohoku University, Sendai, Japan.
Born in Morioka, Japan, in 1930. Bachelor in
Engineering from Iwate University in 1953 and
Doctor of Engineering from the Graduate School
of Tohoku University in 1959. Associate Profes
sor at the Institute of High Speed Mechanics,
Tohoku University, in 1959. Senior Research
Fellow at the California Institute of Technology,
USA, in 1969-70. Professor at the Institute of
High Speed Mechanics in 1970-89 and Professor
at the Institute of Fluid Science in 1989-1994,
Tohoku University. Councillor of Tohoku Uni
versity in 1987-89. Since 1994, Professor Emeri
tus of Tohoku University. Research areas: bub
ble dynamics, cavitation, water jet and nozzle.
Hiroshi Tanaka, Chief Fellow Specialist,

Energy System Group, Toshiba Corporation,
Yokohama, Japan.
Born in Japan in 1933. Graduated from the
University of Tokyo in 1956 and serving with
Toshiba Corporation since then. For the first 20
years of service, research and development for
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

Contributing Authors xxv
both conventional turbines and reversible pump-

turbines. Chief Engineer in charge of hydraulic
machinery in 1987, supervising all R& D and de
sign/manufacture of hydraulic turbines. Since
1991, Chief Fellow Specialist. Members of In
ternational Association of Hydraulic Research,
International Hydropower Association; member
and Ex. Council member of Japanese Society of
Mechanical Engineers; member and Ex. Presi
dent of Visualisation Society of Japan; member
and Ex. Vice President of Turbomachinery So
ciety of Japan.
Yoshinobu Tsujimoto, Professor,

Faculty of Engineering Science,
Osaka University,
Toyonaka Osaka, Japan.
Born in Japan in 1949, Bachelor (1971) and Doc
tor (1977) of Engineering from Osaka University.
Research Associate (1977), Associate Professor
(1986) and Professor (1989) of Osaka University.
Visiting Associate at the California Institute of
Technology from 1983 to 1984.
Hiroshi Tsukamoto, Professor,

Kyushu Institute of Technology,
Kitakyushu, Japan.
Born in Aichi, Japan, in 1948. Bachelor in En
gineering from Waseda University in 1972. Mas
ter of Engineering in 1974, and Doctor of Engi
neering in 1977, from University of Tokyo. Lec
turer at University of Tokyo in 1977. Associate
Professor at Kyushu Institute of Technology, in
1978. Professor at Kyushu Institute of Tech
nology since 1992. Research areas: hydraulic
machines, pump, unsteady flow, unsteady flow
measurements.
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,

xxvi Contributing Authors
Keith R. Turton, Retired, Senior Lecturer,

Loughborough University of Technology,
Leicestershire, England.
Visiting Fellow, School of Mechanical Engineer
ing, Cranfield Institute of Technology. Design
consultant to a number of companies on prob
lems in pumping and fluid dynamics. Author
of Principles of Turbomachinery published by
E & F N Spon and Principles of Centrifugal
Pump Design published by Cambridge Univer
sity Press.
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

Chapter 1
Introduction
S.C. Li
1.1 Cavitation
1.1.1 Discovery
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
Although the word CAVITATION was originally coined by R.E. Froude and
firstly cited by Barnaby & Thornycroft in 1895 [1.2], the phenomenon was
suggested much earlier by Euler in his theory of hydro-turbines in 1754 [1.3].
However, actual cavitation was firstly discovered and investigated by Barn
aby & Parsons in 1893 [1.1] when they found that the formation of vapour
bubbles on blades was responsible for the propeller failure of a British high
speed warship (HMS Daring with a design speed of 27 knots). In 1895,
Parsons established the first water tunnel 1 for cavitation study, and discov
ered the relation between cavitation and its damage on the propeller [1.8].
It was Rayleigh who laid the theoretical foundation for cavitation study by
solving the problem of the collapse of an empty cavity in a large mass of
liquid in 1917 [1.9].
Following the milestone work done by both Parsons and Rayleigh, more
than ten thousand articles and several books on this subject have been pub
lished. Nevertheless, our knowledge of cavitation is still very limited.
1.1.2 Classification
Cavitation is normally defined as the formation of bubbles filled with vapour/
gas or their mixture and subsequent activities (such as growth, collapse and
'This tunnel with a 12 in diameter test section is now at the University of Newcastle
upon Tyne, UK.
1
2 Cavitation
rebound) in liquids. According to the content of bubbles, cavitation can be

classified as vaporous cavitation and gaseous cavitation. Cavitation differs
from boiling by its generating mechanism. That is, it is a phenomenon di
rectly related to the pressure reduction below a certain critical value. Usu
ally, there are two ways by which the pressure reduction is caused. One is
by a fluid flow, which is often referred to as hydrodynamic cavitation. The
other is by an acoustic field, which is often referred to as acoustic cavitation.
However, there are also other cavitations generated either by photons of laser
light or by other elementary particles (e.g. protons in a bubble chamber).
These cavitations are achieved in nature by local energy deposit rather than
by tension in liquid. Therefore, they are often referred to as optical cavita
tion and particle cativation respectively. These two types of cavitation do
not occur in hydraulic machines.
According to the mechanisms by which cavitations are generated, the

classification by Lauterborn, 1979 [1.5], is shown in Figure 1.1. For more
information, the review by Hutton, 1972 [1.4], is suggested.
CAVITATION
ENERGY DEPOSIT TENSION
PARTICLE OPTIC ACOUSTIC HYDRODYNAMIC
Figure 1.1: Classification of cavitations
Nowadays, cavitation study is being performed in various fields such

as (single/multi/stochastic) bubble dynamics, acoustic cavitation, hydrody
namic cavitation, cavitation luminescence, cavitation noise and cavitation
erosion etc.
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

Introduction 3
1.2 Hydraulic Machinery and Cavitation

1.2.1 Problems Caused by Cavitation
Cavitation is almost always an unwanted phenomenon in hydraulic machin
ery although it has some favourable effects in other fields2. Actually, it is a
main obstacle to the development of high-performance machines.
Cavitation will erode machine parts (e.g. Figure 1.2); deteriorate ma-
1
i i i i
Figure 1.2: Typical sponge-like erosion-pattern caused by a severe leading-

edge-cavitation attack (Li, 1987 [1.7])
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
chine performance; cause noise, vibration and even entire system oscillation;
and enhance corrosion/silt erosion through synergism mechanisms. As exam
ples, such cavitation-damaged turbine and pump are shown in Fig. 1.3. The
2
They are used in some physical/medical/industrial applications. For, detail, see
Young, 1989 [1.10].

4 Cavitation
Figure 1.3: Cavitation damaged Francis-turbine runner (top) and

centrifugal-pump impeller (bottom). (Courtesy of IMHEF, Lausanne)
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

Introduction 5
repair of cavitation damage is a costly work, which reduces the capability-

availability and the lifetime of machines.
1.2.2 Combating Cavitation

The past 100-year history shows that the discovery and study of cavitation is
strongly associated with the development of hydraulic machinery 3 . Under a
given application environment, the higher performance (e.g. higher specific
speed, ns) machines are more prone to cavitation and its damage 4 . There
fore, cavitation becomes a main obstacle in developing high-performance
machines.
To combat cavitation, appropriate measures should be carefully consid
ered and balanced throughout the planning of hydro schemes, machine se
lection and parametric design, machine (hydrodynamic) design and material
selection, mechanical design, determination of machine setting level (i.e. the
plant cavitation number, ap, for turbines and the required net positive suc
tion head, NPSHreq, for pumps), and machine operating/maintenance/repair
Thus, a comprehensive and optimum anti-cavitation approach can then be
developed for a given project, which requires a joint effort cutting cross many
disciplines.
There are basically two concerns in combating cavitation. One is how
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
to avoid cavitation in the first place. On the other hand, if cavitation is
not avoidable, measures should be employed to minimise it to an acceptable
level and to reduce future repair cost by easing the repair/replacement of
the parts that are likely to be damaged.
When selecting or designing a machine for a given scheme, previous ex
perience or databases should be used as a prime reference since theoretical
calculation and prediction of cavitation occurrence in a machine, particu
larly for off-design operation, is not accurate and not always reliable. The
advanced flow-simulation technique, such as the 3D A; — e turbulence flow
model 5 , should be employed to optimise the hydraulic design of machine.
3
In broad definition, any machines that use a fluid as the working medium either to
change energy from one type to another or to transmit energy are referred as hydraulic
machinery. Based on their functions, they can be further categorised as four types: hydro
driving machines (e.g. hydro turbines), pumping machines, hydro transmissions and hydro
propulsions (e.g. ship propellers).
4
For example, the cavitation erosion rate increases with the 6th power of relative flow
velocity.
5
A complete simulation of a Francis turbine (i.e. from the inlet of the spiral case to
the exit of the draft tube) with a k — e turbulent model has been reported recently by

6 Cavitation
It is vital to avoid the most damaging cavitations (such as the vortex cav
itation) and unsteady cavitating flows (such as cavitated vortex shedding
and cavitated vortex core in the draft tube). A proper model-test technique
should be adopted to detect critical flow structures prone to cavitation and,
by using a precise scaling law, to predict possible cavitation inception and
its development on the prototype. The proper selection of the machine
setting level is vital to avoid blade cavitation but it has little effect on non-
blade related cavitation. A safety factor is required for choosing the value of
NPSHreq or op to ensure an acceptable degree of cavitation 6 . However, this
will increase capital cost since a deeper setting of the machine requires more
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
civil work. Therefore, a careful trade-off is needed. The use of materials with
high-cavitation resistance and low cost (in terms of both the material itself
and the machining/repairing process) in the areas prone to cavitation attack
is effective in mitigating cavitation damage. From an operation view point,
if possible (e.g. for the turbines providing base load), avoiding off-design
operation is crucial. Besides, anti-cavitation devices such as air injection,
blade fin (for Kaplan/propeller machines) are the effective measures often
used. Novel designs are also sought for developing low-cavitation and high-
energy machines. For example, the idea of using a double-row cascade runner
was proposed by Li, 1964 [1.6], and has been proved successful in producing
such machines 7 . As concerns maintenance and repair, appropriate inspec
tion and maintenance programme and repair approach (which involves repair
Keck et al ( Sulzer Technical Review 1/97, pp26-29). This approach uses the technique of
'circumferential mixing plane interface' to simulate the interaction between the rotor and
the stator so that it can produce a full hill chart of machine performance. It is claimed t h a t
for retrofit-turbine projects this approach can design new runners matching the existing
components (for say, the stay ring) much better t h a n t h e costly model test approach does.
6
T h e machine value of NPSH or a is obtained by the test against a certain energy (e.g.
efficiency, pumping head etc) drop which is caused mainly by blade cavitation. Indeed, at
this stage, cavitation has already developed into a severe status.
7
This idea was tested with favourable results on the Francis turbine (Yunfeng Power
Station, China, in late 1960's) and on pumps (by Shanghai P u m p Factory in early 1970's).
Recently, a novel Francis turbine (unit capacity 150 MW) using such a double-row struc
ture has been developed by Kvaerner Brug A / S (Norway) for the Lubuge Power Station,
China. By using a double-row (the blade chord-length of the second row is shorter) cas
cade runner, particularly for low n , Francis machines, machine cavitation number a can
be remarkably reduced owing to: (a) the favourable alteration of the pressure distribution
on the main blades; (b) the postponement of boundary-layer transition and separation
in some operating conditions which in turn delays cavitation inception; (c) the mitiga
tion of secondary-flow formation during part-load operation. Apart from a reduction, the
high-energy (efficiency) zone becomes much broader owing to the better flow conditions
achieved.

Introduction 7
frequency, damage-cause analysis, repair method and material selection etc)

are vital to the minimisation of cavitation and its damage. By an effective
repair approach, cavitation problem on a machine can be minimised and
even solved completely. Otherwise, the damage will become a progressive
process. For a retrofit project, it is often seen that severe cavitation and
damage is introduced by retrofitting because the desire for a higher energy
performance (output and efficiency) often tempts one to overlook the risk of
cavitation. For example 8 , a retrofitted new Francis runner usually possesses
a larger diameter, fewer blades and a much sharper corner of the bottom
ring. This will create a flow environment prone to cavitation. Therefore, for
retrofit design, precautions must be taken in balancing the energy gain and
cavitation risk.
Having implemented an appropriate comprehensive anti-cavitation ap
proach, such as mentioned above, it is possible to produce a virtually cavita
tion (damage)-free turbine of 60-year lifetime, capable of operating 30 years
without major maintenance. For pumps, a ten-year lifetime may be possible
if cavitation intensity is controlled below the material resistance.
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
See §6.6.4 Example.

8 Cavitation
References
1.1 Barnaby, S. W. and Parsons, C. (1897), Trans. Inst. Naval Arch. 38.
1.2 Barnaby, S. W. and Thornycroft, Sir John, (1895), 'Torpedo Boat

Destroyers', Proc. Inst. Civ. Engrs. 122, 57.
1.3 Euler, L. (1754), 'Histoire de l'Academie Royale des Sciences et Belle

Letters', Memo. R.10, Berlin, pp266-267.
1.4 Hutton, S. P. (1972), 'The Saga of the Singing Kettle', Inaugural

Lecture, University of Southampton, UK.
1.5 Lauterborn, W. (1979), 'Cavitation and Coherent Optics', Cavitation

and Inhomogeneities in Underwater Acoustics, Springer-Verlag, pp3-
12.
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
1.6 Li, S. C. (1964), 'Hydrodynamic Approach for Improvement of Hy-
droturbine Cavitation Performance', Graduation Thesis, Beijing Uni
versity of Hydro Engineering; also, Proc. 2nd National Conference on
Turbomachinery, Harbin, China.
1.7 Li, S. C. (1987), 'Cavitation Damage on Pump-turbine', Consultant

File.
1.8 Parsons, C. (1912), Trans. N. E. Coast. Engrs. & Shipbuilders.
1.9 Rayleigh, Lord (1917), 'On the Pressure Developed in a Liquid during
the Collapse of a Spherical Cavity', Phil. Mag. 34, pp94-98.
1.10 Young, F. R. (1989), Cavitation, McGraw-Hill, UK, pp389-408.

Chapter 2
Cavitation and Cavitation

Types
A.J. Acosta
2.1 Cavitation Phenomenon

2.1.1 Cavitation
The word cavity is derived from Latin and it has the meaning of a hollow
or empty place within a solid1. In the technical literature the word "cavita
tion" has come to have several related meanings; for example, the formation
of voids in a solid due to stress in one. The term is also used for the su
personic expansion of a gas into a vacuum. In engineering concerned with
applications to naval hydrodynamics and hydraulic machinery, the term cav
itation refers to the formation of voids, hollows, or simply "cavities" within
a liquid medium. The cavities in each of these cases are filled with a gaseous
phase; this phase may consist of a mixture of the vapour of the original
liquid medium, a gaseous phase of a non-condensible component dissolved
in the liquid, or perhaps a different gaseous phase injected or entrained into
the cavity either deliberately or by the action of the flow. For example, a
marine propeller or lifting hydrofoil operating near the free surface of the
ocean may entrain air from the surface into the low-pressure regions of the
foil forming large cavities which have a severe effect on the performance of
the lifting surface. The propeller is then often said to "cavitate". In this
case the contents of the cavity are principally that of the surrounding air;
some writers prefer to use the term "ventilation" for this term (see Acosta,
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
'Oxford Compact Unabridged Dictionary, 1971.
9
10 Cavitation
1973) or "artificial" cavitation. Indeed, the deliberate introduction of ven

tilating air to cause cavities is a well-known means of modifying the lift on
foils/propellers and has been used for that purpose.
2.1.2 Hydrodynamic Cavitation

But it is not necessary to have a neighbouring free surface or artificial means
to provide the contents of these cavities in the liquid; that can be done from
the liquid medium itself by vaporisation. Indeed, this is the usually under
stood meaning of the term "cavitation" in a hydraulic machine; namely, the
cavitation results from the liquid being subjected to pressures less than the
vapour pressure of the liquid through the action of the fluid flow. That is
to say, the pressure change is brought about by the Bernoulli effect and
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
is therefore a dynamic process. The phase change that results provides the
vapour of the liquid together with that of any other dissolved component.
There are further distinctions to be made, however, depending upon the
relative importance of the dissolved gas content. It may be recalled that a
thermodynamic system consisting of a pure liquid and a dissolved noncon-
densible gas (that is the most common situation in hydraulic machines and
for marine applications) may have an equilibrium composition at a given
temperature proportional to the partial pressure of the non-condensible gas.
(This relationship is known as Henry's Law.) It follows that if the mixture
total pressure at a given temperature and composition is less than equilib
rium one, the dissolved gas phase will come out of solution, thereby forming
a "cavity" or gaseous phase. This equilibrium pressure at which the gas
phase appears is called the "bubble point". Plainly, if the liquid pressure is
less than the bubble point, "cavitation" from this source can occur; if the
pressure is less than the vapour pressure of the pure liquid, both components,
gas and vapour of the liquid, can appear within the cavity 2 .
Although these effects are well-known, they can and do cause difficulty
and uncertainty in the interpretation of specific cavitation phenomena, as
will be mentioned subsequently. Professor J.W. Holl, in his influential article
(Holl, 1969) clearly distinguishes between the cavitation phenomena arising
from situations where the cavity contents are primarily non-condensible gas
on the one hand and condensible vapour on the other (it is assumed that
2
More complex multi-component systems are common in petro-chemical pumping ap
plications. Even sea water is different than the tap water commonly used in test facilities
and, as pointed out by Edstrand (1984), the presence of CO2 in sea water can affect the
cavitation occurrence and types expected from model tests.

Cavitation and Cavitation Types 11
ventilation or artificial cavitation is ruled out). The first of these he terms

"gaseous cavitation" and the second "vaporous cavitation." Both occur in
technically important flows and, as will be seen, there are important physical
consequences of these two types.
Thus far in this section we have described one physical basis for the
formation of a gaseous/vaporous phase in a liquid by dynamic action of the
fluid flow. It is clear that at a given system pressure the liquid temperature
could equally well be raised to form gaseous evolution of dissolved non-
condensibles as well as "boiling" of the pure liquid component. Indeed, there
are many close similarities to boiling and vaporous cavitation occurring in
"bubbly" flows for they are in essence the same phenomenon. But there
are differences, too, because in the fluid flow typical of hydraulic machines,
and hydrodynamics in general are the additional features of a complicated,
three-dimensional, unsteady viscous flow which has a profound effect upon
the physical form the cavitation "cavities" take. We discuss these various
forms and types, some of their effects, and mechanics of cavitation inception
in the next section.
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
2.2 Types of Cavitation
2.2.1 Cavitation Index
The cavitation index used herein and throughout the field is the dimension-
less pressure coefficient called the cavitation number
a = (Poo-Pv)/(pU20O/2) (2.1)
where p^ is the static reference pressure of the fluid, pv, the equilibrium
vapour pressure of the pure liquid medium, p the fluid density and C/oo the
reference speed. In hydraulic machinery applications, I / ^ is the inlet rela
tive velocity; sometimes the inlet total pressure pa (at the site of cavitation)
is used and the difference (pti - pv) is called the net positive suction pres
sure, NPSP. For vaporous cavitation to occur p m j„ < pv where p m j n is the
minimum pressure in the fluid. The value of Eq. 2.1 when the onset or in
ception of cavitation occurs is called cavitation inception index and denoted
0{. Sometimes the pressure is known (or computed) within the flow or on
the wetted surface and made dimensionless in coefficient form as
Cp = ( p - P o c ) / ( p ^ / 2 ) (2.2)

12 Cavitation
and clearly then

—
0~i < Cp min
is a condition for vaporous cavitation inception. As Holl (1969) shows, if

Oi > —Op min, then gaseous cavitation occurs by diffusion of dissolved gas.
For most technical applications the actual cavitation number is less than
the inception value and less than —Cp m j n although this may not be known.
2.2.2 Scale Effects and Cavitation Types

In most applications in hydraulic machinery similarity of operation follows
from the affinity laws for liquid turbomachines and it follows that the cavi
tation behaviour of geometrically similar turbomachines would be expected
to depend only on the cavitation number. When this does not happen, there
is said to be a "scale effect". Cavitation is a complicated physical process
involving a phase change with heat and mass transfer and the extent and
development of the cavitation in a machine depends in addition upon the
concentration of the cavitation nuclei within the fluid itself. Holl and Billet
(Morgan and Parkin, 1979, p.12) classify these scale effects into two types.
Because these effects do cause important differences in the type and appear
ance of cavitation, we list them here before discussing the range of cavitation
types (paraphrased).
• T y p e 1 These effects change the external flow field by changes in the

Reynolds number, Fronde number, or geometric effects such as surface
roughness and includes turbulent pressure fluctuations.
• T y p e 2 These effects change the physical processes governing the

phase change of the liquid and include the time scale, heat and mass
transfer, surface tension, concentration of cavitation nuclei.
These are useful distinctions because, as they mention, Type 1 affects

the value of Cp m j„ whereas Type 2 effects may locally alter growth rates of
microbubbles and cavities and the vapour pressure, pv, of the pure liquid
medium. This latter effect is called the "thermodynamic" effect and it is
often important in the pumping of liquids when the density ratio of the liquid
and vapour states is not great as it is for cold water, for example, in which
this effect is negligible. With this as a background and drawing on similar
tabulations (Acosta and Parkin, 1980), we tabulate forms of cavitation in
Table 2.1 occurring at and below inception in a wide variety of situations. All
of these various types of cavitation can occur on a pump or turbine. Often
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

Item Type Description

1 Travelling bubble Spheroidal bubbles travel over the sur
face at near free-stream speed
2 Sheet cavitation A somewhat clear sheet from near
the nose of hydrofoils or axisymmetric
bodies
3 Super cavitation Similar to (2), a "fully developed" cavity
usually longer than the chord
4 Bubble ring A highly intermittent form in a ring near
the body surface - in a separated region
5 Band cavitation Consists of many small bubbles forming
a region of macroscopic cavitation
6 Fixed patch cavitation An attached form of cavitation often
wedge shaped starting near Cp m j„
7 Travelling patch cavitation Similar to the above but moves down
stream, gradually merging with travel
ling bubble forms
8 Spot cavitation Similar to patch cavitation - usually
more like a string. Often due to local
roughness
9 Cloud cavitation A collection of cavitating bubbles, of
ten seen in wakes of bluff bodies, down
stream of unsteady collapsing sheet cav
ities on propellers or hydrofoils. Also in
shear flows and jets
10 Tip vortex cavitation Most clearly seen on propeller or hydro
foil tips; highly rotational flow. Also
occurs in tip clearance or gap flows in
turbomachines, the hub of propellers or
turbines at off-design
11 Vortical cavitation A reference to the cavitation seen in the
wakes of bluff bodies, both primary and
secondary (streamwise) cavitation is of
ten seen. Also downstream of unsteady
collapsing sheet cavities on partially cav
itating foils
12 Shear flow cavitation Within the wall shear layer, free shear
layers of jets
Table 2.1: Types of cavitation
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

14 Cavitation
in the pump literature these several types are lumped together into simply
"blade surface" cavitation for items 1, 2, 3, 5, 6, 8 because it is not often
easy to observe these flows in machines. Cloud cavitation can take place in
the jet-like flows through wearing rings and in the mixing region with the
reverse flow and the oncoming meridional velocity as described by Stoffel and
Ludwig (1990). If a pump is operated far off-design, say at 25 percent design
flow rate, then strong reverse flows occur both at inlet and discharge and
then a severe cavitation environment may exist at the inlet with cavitation
and cavitation damage then occurring on the normal pressure surface of the
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
vane. Based on the observations of Giilich (1989), it is probable that an
intermittent collapsing cloud cavity is responsible for this damage.
These various kinds of cavitation, their description, origin and connection
to the underlying viscous fluid flow, population of micro-bubbles, and two-
phase flow mechanics has been studied intensively by the major hydraulic
and hydrodynamic laboratories of the world for the last twenty years and
this effort continues today with unabated effort. There have been several
conferences which highlight the many reasons leading to the long catalog
of Table 2.1. Among these may be mentioned the various International
Symposia in the References from which these various types of cavitation are
drawn. Of these various sources the review by Rood (pp. 1-22, Morgan and
Parkin, 1989) is most helpful, particularly for those in the hydrodynamics
field concerned with cavitation inception and the less available International
Towing Tank Conferences (Huang, 1990).
Figure 2.1: Cavitation on a "Schiebe" body having Cp m ; n = 0.75; the same

body of 50.8 mm diameter is shown in two different water tunnels showing
travelling bubble cavitation in one case a = 0.46 and a fixed "patch" of
cavitation in the other a = 0.6 (Gates and Billet, 1980 [2.29])

l.Rome 2. AEW 3. Delft
4. NPL 5. Cal. Tech. 6. Cal. Tech.
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
7. SSPA 8. SSPA 9. SSPA
Figure 2.2: Different forms of cavitation on a smooth body with a laminar

separation (Johnsson, 1969 [2.41])
We mention again the two types of effects causing a scale effect; the first
of these is a real fluid, that is, a viscous effect. It is now realized in the
cavitation community, somewhat belatedly, that flows common in test lab
oratories may have a laminar separation present. The bubble ring, band,
and sheet cavitation forms all occur within a laminar separation. Some bod
ies do not have a laminar separation; an excellent example of one having
a deep minimum pressure coefficient and still no separation is the so-called
" Schiebe" body, named for Frank Schiebe, who was the first to use these par
ticular half-bodies for research (Schiebe, 1972). Figure 2.1 shows cavitation
on such a body in two different facilities. One has travelling bubble cavita
tion, the other an attached fixed patch; many intermediate forms can exist.
On the other hand, Fig. 2.2 shows the well-known ITTC standard head form

16 Cavitation
Bubble-ring cavitation on the hemisherical nose Band cavitation on the hemispherical nose
a =0.626. Voo= 18.3mps, a=9.5ppm a=0.610, Vo<pl8.3mps, a=9.2 ppm
Figure 2.3: Bubble ring (left) and band cavitation (right) on the hemisphere
body (Holl and Carroll, 1979 'Observations of the Various Types of Limited
Cavitation on Axisymmetric Bodies', Intl. Symp. on Cavitation Inception,
ASME, pp.87-101).
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
which is (now) known to have a laminar separation for the test conditions
of Johnsson's round-robin tests. There are clear and important differences
between the two types of bodies and the kinds of cavitation revealed. The
hemispherical head form, possibly the most tested body in hydrodynamics,
also contains a laminar separation and Fig. 2.3 shows the ring/band kind
of cavitation there. These important viscous effects were first pointed out
by Arakeri et al. (1973) and it then became clear that these viscous effects
could be manipulated by boundary layer excitation through polymer injec
tion, trips/roughness. More importantly, it may be that the prototype on
a large scale may not have a laminar separation while the laboratory scale
might. Curiously no travelling bubble cavitation was seen by Arakeri or Holl
in their work on the hemisphere body but in other experiments on the ITTC
headform Kodama et al. (Morgan and Parkin, 1979) was able to observe
them by supplying additional nuclei to the test fluid even for conditions far
beyond inception when a developed or sheet cavity had formed.

The presence of laminar separation and laminar transition is a Type

1 or external influence on the flow. Indeed, boundary layer stimulation
by roughness elements may be used to preclude some features of this scale
effect (Huang in Morgan and Parkin, 1984) and this technique has also been
applied with apparent success to propeller blade leading edges, (Kuiper,
1979; Huang and Shaw, 1986).
The foregoing discussion has originated largely from considerations near
inception (except for Fig. 2.4). As the cavitation index is decreased at
Sheet cavity on n iC body with no added nuclei U00=6.8 m/s, a =0.60
Bubble cavities on ITTC body with added nuclei U 00 =6.8m/s, a =0.60
Figure 2.4: The same body as in Fig. 2.2 with different nuclei concentra
tion (Kodama et al, 1979 'The Effect of Nuclei on the Inception of Bubble
and Sheet Cavitation on Axisymmetric Bodies', Intl. Symp. on Cavitation
Inception, ASME, pp.75-78).
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,
constant speed from inception the extent of cavitation in all cases increases.
Items 2, 4, 5, 6, 8 of Table 2.1 are all attached and limited forms of cavi-

18 Cavitation
tation according to Holl (1969). And with continued decrease in cavitation

number, these regions expand from a bubbly, foamy region to form an at
tached sheet cavity, the contents of which are largely vapour and dissolved
gas, however, the terminus of this region is highly unsteady on all bodies.
This condition prevails until the cavity becomes quite long; in the case of half
bodies or hydrofoils, the cavity may become longer than the body. There
then results a clear sheet cavity and this was labeled "supercavitation" by
Marshall Tulin many years ago. This sequence of events is amply illustrated
in the monograph by Knapp et al. and because the flow except at the cavity
terminus approaches that of a potential flow has led to extensive mathemat
ical and numerical analysis. See, for example, the review by Wu (1973), and
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
Figure 2.5: An attached cavity formed on the Schiebe head form made in
acrylic resin: the cavitation index is a = 0.32 (Ceccio and Brennen, 1990a
[2.16]).
Note the different appearances of the cavitation seen in Fig. 2.1 on the same
body.

the papers by Murai (1968) and Puruya (1975) as samples of this extensive
subject.
The sequence of events is not so clearly understood when more extensive
cavitation is developed from the travelling bubble inception form. In that
situation, with a decrease in the cavitation index, more bubbles form which
tend to cover the surface of the test body until a condition is reached when
an attached cavity or a sheet is formed on the body. This process occurs
even on surfaces not having a laminar separation. This process was clearly
observed on the non-separating Schiebe body by Gates et al. (1976) and
more recently by Ceccio and Brennen (1990a) shown in Fig. 2.5. Implicit in
the formation of an attached sheet cavity is an upstream laminar separation
(Arakeri, 1975; Franc and Michel, 1985); that is, the formation of cavitation
surface itself requires the viscous layer upstream itself to separate. It would
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
seem then that the presence of the many travelling bubbles give rise for
the conditions of a laminar separation to occur on the body which would
otherwise not have the separation. This supposition remains to be shown,
however.
■ KCAO'OftM SUR/ACC
-£ Htkorottu suffice
9»0 Q0
V- KOUCM V suHMCt
- 5 mm
(a)
Figure 2.6: Schematic diagram of a cavitating bubble evolution (a) the non-
separating Schiebe body, and (b) from the ITTC head form of Fig. 2.2
(Ceccio and Brennen, 1990b [2.17])
The process of travelling bubble cavitation even when the bubbles are
far apart, as first photographed by Knapp, have been recently studied ex
perimentally again by Ceccio and Brennen (1990b) who show by systematic

20 Cavitation
photography that the growth and collapse history of individual bubble cavi
tation is quite different depending on whether or not the body has a laminar
separation. In both cases the bubble is very non-spherical as shown in Figs.
2.6a,b; the presence or absence of a separation has an important effect upon
the process of cavitation re-bound and as a consequence upon the noise signa
ture of the collapse. It is particularly interesting that the trailing streamers
seen in Fig. 2.6b, which have been widely seen before, are now shown to
occur in the pressure of a laminar separation underneath the cavitating flow.
The presence of cloud cavitation has already been mentioned as a promi
nent feature near the terminus of a sheet cavity or in the wake of a recircu-
Figure 2.7: Sequence of photographs showing progressive states of cavita

tion on an NACA 64-309 oscillating three-dimensional hydrofoil. The foil
is oscillating 2° about a 7° mean angle of attack at a reduced frequency,
k — wc/2U, of 0.88 and cavitation number of 0.5. The sequence shows a
leading edge cavitation cavity forming at a 9° angle of attack (top left).
Photographs taken in 0.5° increments show cavitation beginning as travel
ling bubbles progressing to a leading edge sheet cavity and then collapsing as
an intense cloud cavitation as the foil oscillates to 7° angle of attack (bottom
ring). The foil has an aspect ratio of two and 15 cm chord (Hart et al, 1990
[2.34])
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

lating shear flow and it is as a consequence an important feature in many

pumps and propellers. Van der Muelen et al. (1990) have shown that the
collapse of these cavities can be very energetic and suggests this is a source
of propeller cavitation erosion. The intensity of this process is attributed
to vorticity being perpendicular to the cavity (i.e., in the spanwise direc
tion). This is the case shown in the cloud collapse shown on the hydrofoil
of Fig. 2.7; the sheet cavity (paraphrasing van der Muelen) on a propeller
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
is however normally three-dimensional and the sheet cavity terminus has a
pronounced sweep back which streams smoothly off into a tip vortex and an
intense cloud cavity collapse is avoided. The unsteady cavitating flow past
hydrofoils continues to be a matter of great concern (e.g. Narasaka et al.,
1990) and the models of these flows based on a bubble "cloud" concept have
attracted much attention. Recent numerical studies (Kubota, 1988) show
remarkable similarity to experiments. The idea of a cloud cavity as a means
of greatly simplifying hydrodynamic interactions had already attracted con
siderable attention (Chahine (1982), d'Agostino and Brennen (1983)) and
further development of this approach may be expected.
Figure 2.8: Tip vortex cavitation in an axial flow pump with a rounded blade
tip (right) and a squared-off one (left). (Caltech photo)

22 Cavitation
A tip vortex flow and tip vortex cavitation is often a highly visible and
even spectacular feature of propellers and foils of finite aspect ratio. The
bound vorticity of the foil shed into the flow organises itself into a rolled
up vortex core forming low pressures in which cavitation can begin. The
flow can be exceedingly complex (see the review by Rood, 1989), is not
necessarily steady, and depends sensitively on viscous properties of the flow
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
as well as the dissolved and free gas content. It is thought that inception
can be either vaporous or gaseous depending upon the length of time of
exposure to cavitating conditions. Pumps and hydraulic machines in general
may experience tip vortex cavitation in the tip clearance space of unshrouded
machines as shown in Fig. 2.8 or in the leakage shear flow of the wear rings of
shrouded impellers as sketched in Fig. 2.9. In his survey paper Grein (1978)
Figure 2.9: Location of cavitation in a pump inlet (Stoffel and Ludwig, 1990
[2.72])
shows photographs of the cavitation development in the eye of a shrouded

centrifugal pump impeller. At inception, attached forms of cavitation at the
blade leading edge are seen which develop into a glassy sheet cavity and on
the shroud surface travelling bubble cavitation is seen as well as in the wear
ring leakage flow; these latter forms appear to be "cloud" cavities at lower
pressures.

The term vortical cavitation is used to designate the strongly rotational

flow found downstream of bluff bodies, behind backward facing steps in slots
(Cimbala and Billet, 1989). In all these cases, the cavitation index at incep
tion is significantly greater than the time averaged or instantaneous value of
—Cp min on the surface of the body. The implication is that fluctuating pres
sures within the fluid because of the unsteady flow give rise to the cavitation.
But it is not usually known if the cavitation in these regions is gaseous or
vaporous because it is exceedingly difficult to measure fluctuating pressures
within the flow field. One method to do (Ooi et al., 1983) uses the volume of
initially uniform microbubbles in a turbulent non-cavitating flow as a mea
sure of local pressure. Based on ideas from such measurements, Cimbala and
Billet (1989) suggest that the slot cavitation is gaseous because the pressure
fluctuations are sufficient to cause air diffusion (i.e., the transient pressures
are below the bubble point) but are not strong enough to force vaporous
cavitation.
The last item of Table 2.1, although similar in respects to number 11,
is characterised by the absence of a body scale length; the relevant lengths
are those characterising the spatial and temporal structure shear layer itself.
These flows are almost exclusively turbulent and are unlike flows around
bodies; cavitation is driven by the character of these mixing regions (the
Type one effects) and the population of the cavitation nuclei available to
react to the flow. Such flows have been extensively studied by Arndt (1969,
1981).
In closing this section it should be mentioned that, although some of
the main ideas of cavitation types have been described here, there is not
necessarily a universal acceptance of these concepts. Also, attention has,
been directed towards the "onset" or inception of cavitation rather than its
"disappearance" or desinent condition. Both quantities are important and,
as described by Holl (1969) and even more so by Keller (1979), the liquid en
vironment and the test procedure can significantly influence inception (and
desinent) results. This is in part because, as Keller points out, the popu
lation of nuclei within recirculating test facilities is not constant with time.
And, in addition, as is increasingly recognised (e.g. Li et al., (1986)), cav
itation is a stochastic process. Many of these kinds of considerations until
recently have been largely overlooked in the hydraulic industry in the test
ing of pumps and turbines for performance and cavitation endurance. But
with the increasing emphasis on long life and reliable operation of pumps for
power plants and industrial applications, they are now in the mainstream of
the industry (Giilich (1989), Gopalakrishnan (1985), Avellan et al. (1986)).

--``,```,,````
24 Cavitation
2.3 Cavitation Effects

Knapp (1952) states that "Cavitation is one hydraulic phenomenon in which
the effects are predominantly undesirable and often destructive". Despite
the few constructive uses that can be made of cavitation in the design of
special self-regulating pumps (as pointed out by him) or its use in cleaning
or cutting devices, most hydraulic engineers today would agree with Prof.
Knapp's early observation and it is generally recognised that cavitation ad
versely changes the performance of hydraulic equipment, may result in an
unsteady flow causing noise, mechanical vibration, internal flow surging and
material damage termed cavitation "erosion". Indeed, this erosion is a se
vere problem in all hydraulic equipment in hydropower plants, thermal power
plants, industrial pumps and it has been with us from the first applications
of ship propulsion by propellers (Knapp et al., 1979) and it continues so to
the present day in hydraulic machines (Chapter 6 to follow). More extensive
recent reviews describing the effects of cavitation in hydraulic machinery and
structures (Arndt, 1981; Arndt and Webb, 1985) are recommended reading.
2.3.1 General Phenomena
As the cavitation index is decreased from inception the extent of cavitation

increases. On lifting surfaces such as a hydrofoil the increasing extent of
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
cavitation alters the lift coefficient and generally increases the drag. If the
cavitation is a sheet cavity starting near the leading edge, the lift coefficient
may at first rise with increasing length of the cavity and then decrease to
a much lower value when the cavity becomes longer than the foil; several
examples are shown in Fig. 2.10. Generally, the cavitating flow is not steady
and typically when the sheet cavity becomes about 1/3 to 1/2 chord length
oscillations develop in which the entire cavity or perhaps only portions of
the cavity grow and collapse nearly periodically (Murai and Itoh, 1985) with
resulting fluctuations in lift force, that can be an appreciable fraction of the
average lift depending upon the foil type. These force fluctuations cease
when the cavity is significantly longer than the chord. The frequency of
the fluctuations is "high' just after inception and decrease with increasing
cavity length. The dimensionless reduced frequency k = u>c/U where w is
the angular frequency, c the chord and U the free stream velocity may range
from 5 for the high frequencies to about 0.1 for the low but there may be a
broad band of frequencies (Murai and Itoh loc cit).

r— 1
O-2.07S
IK
^~^v
—r
>6
1
'
A, i
1
o-1.2*9
« I
90 ' r
<s-0M6
( M
J il 100
12)
~\—^_i_
MO ».«
Crr^t
nvac
1.0 1.6
«►
13 1J
-*•
5.4
OJM
Hz CAVITATION INDEX, a
Figure 2.10: The effect of cavitation on lift and drag of a foil showing un
steady effect on lift and the frequency of these effects (Murai and Itoh, 1985
[2.61])
2.3.2 Influences on Machine Performance

Very similar processes can occur within a pump. That is, there is a per
formance change of head coefficient and efficiency as shown in Fig. 2.11 on
a medium specific speed pump, Ns ~ 0.75 somewhat similar to that of a
hydrofoil3. It may be seen there that a small "loss" of performance due to
cavitation is acceptable. The amount of this loss varies from 1 percent to
3 percent and depends upon the expected life due to cavitation erosion. It
is typical that lower cavitation indices result for lower flow coefficients for
centrifugal impellers; the sharp decline of performance seen in Fig. 2.11 is
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
termed cavitation "breakdown" and it is due to the profound effect of the

large extent of cavitation in the impeller inlet and it tracks to a great degree
the performance of the hydrofoils shown in Fig. 2.10. Some applications
3
Here N, = QQi/2/(Apt/p)3/4 where Q is the angular speed in rads/s, Ap ( the total
pressure rise, p the density, Q the volumetric flow rate all in consistent units.

26 Cavitation
-
—r — I — T 1 F—|-—•" ~r~ — 1 — 1 "
—"- ♦,-OJS
\
0.43
" ■ * * . ,
0.41 -
I ACCVTAIU — <
1 KXTOCMANCE
LOSS •
^»
■_ _ L 1_ 1 1 1 1 ' l , . 1 .
0 OJ 1.0 !>
OvU»6oo Index Btted on Inlet Relative Velocity, a
Figure 2.11: Cavitation performance of an efficient centrifugal pump at con

stant inlet flow coefficient vs the cavitation index based on the inlet relative
velocity. The flow coefficient, <f>\, is the ratio of inlet axial velocity to inlet tip
speed U\ (adapted from Blom, C , 1950 'Development of Hydraulic Design
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
for the Grand Coulee Pumps', Tran. ASME, Vol. 72, p. 53).
may require operation with significantly lower values of cavitation index

than those of Fig. 2.11, even at the same specific speed. This may be ac
complished if the inlet portions of the pump are provided with lightly loaded
hydrofoil cascade sections which may be an integral part of the impeller or a
separate portion attached to the main impeller, usually called an " inducer".
The sole function of the inducer is to provide the total pressure rise needed
to make the main portion of the impeller operate satisfactorily; that is, the
inducer dimensionless pressure rise should just about equal 4 the inlet cavi
tation index needed for acceptable operation of the main impeller. Figure
2.12 shows the cavitation performance of a typical inducer pump (Kamijyo,
1975); the cavitation performance of a pump impeller (such as that in Fig.
2.11) fitted with an inducer upstream of the impeller on the same shaft and
with no intervening stationary blading is similar to that of the inducer alone
at the design point of the two units and total pressure rise is essentially the
same as that of the original pump impeller. Such inducer pumps have been
intensively studied around the world (e.g. Kamijyo et al., 1975; Velensek,
1983; Wood and Whippen, 1969; Janigro and Ferrini, 1973; Anton, 1975;
Takamatsu et al., 1980) to mention only a few. With tip blade angles rang-
4
If both are based on the same reference dynamic pressure.

N« 10.000 tyre
^•aora
0.106
8-H 0.117
■
*" OOJI-
t OJOJ 010 0.1) 020

CAVITATION INDEX, o » ( p r pY) /pU? / 2
Figure 2.12: Cavitation performance of a conical inducer pump (Kamijyo et

al, 1975 [2.42]).
There are three blades, inlet tip angle = 7.5°, outlet hub ratio= 0.29 and
the tip solidity is 2.5.
ing from 6° to 15°, solidities 5 of 1.5 to about 3, values of the cavitation index
as low as 0.02 are readily achieved in fluids such as water and liquid oxygen.
When the fluid state becomes relatively nearer the critical point an addi
tional "thermodynamic" effect appears due to the adiabatic evaporation of
liquid to fill the vapour cavities causing a well-known depression of the local
vapour pressure and operation at a lower value ofCT,than would otherwise
be the case. This is a Type 2 similarity effect (Holl, 1969); see Arndt (1981)
for more detail.
These more extreme suction conditions frequently required for operation
of a pump are far removed inception itself. The ratio of the cavitation index
from inception to that of breakdown for medium specific pumps (0.55 <
Ns < 1.4) ranges from a low of 2.5 to over 10 (McNulty and Pearsall, 1979).
The conditions near inception are still of intrinsic interest, however, and we
shall return to this point in the next section.
The unsteady cavitation oscillations seen on individual hydrofoils also
occur within complete pumps and gives rise to unsteady pump discharge
and suction pressures (Murai and Itoh, 1985) as well as cavitation noise,
an example of which is shown in Fig. 2.13 for an inducer pump. It is now
' R a t i o of developed blade length to spacing in the peripheral direction.
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

28 Cavitation
2*0 ~\ r
£
220 -O 0 O-
200 J i I I L
110 1 1 1 1 -i r »' ■
id
20- 40KH*
40- 60KHz
{<* IOO - 60- 80KHz
S0-120KHz
ST MI- ~
K
fr-"*-!**-' o -
1
70 -
1" -
1
<>§ eo
30 i
K) 20 30 40
NPSH IN FEET OF WATER
Figure 2.13: Variation of acoustic output at various frequency bands and the
total developed head vs Net Positive Suction Head for a 5 in inducer pump
rotating at 3600 rpm; the flow is 600 gpm (U.S.) (adapted from Gopalakr-
ishnan, 1985 [2.31])
generally agreed that cavitation noise, readily detected for hydrofoils, can
be separated from the machinery noise of cavitating pumps and used as a
diagnostic test for the presence of cavitation (Gulich, 1989; McNulty and
Pearsall, 1979) and even erosion (van der Muelen, 1986).
2.3.3 System Stability

A further and most important feature of the cavitating flow in a hydraulic
machine, pump or turbine is the possibility of system stability; system os
cillations associated with such machines have had a long history and are
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

discussed at length in Chapter 7. In these preliminary remarks it is per

haps only important to note that the unsteady surging oscillations reported
in the literature for many inducer pumps is an interactive phenomenon of
the pump (or turbine) itself, the attached hydraulic system and of course
any associated control system. It was realized early on that analysis of this
dynamic phenomenon required a dynamic representation of a cavitating hy
draulic machine. It should be recalled that the cavitating performance tests
of Figs. 2.11, 2.12 are for steady inlet pressure, and constant rotative speed
and flow rate (these may be difficult to realize in some situations), and that
these conditions do not prevail when system oscillations do occur. This led
many workers in the 1960s to represent the dynamic perturbations of pump,
for example, about its average operating point in the following form
( Z ) -I* ( Z ) ■ ("'
Here Ap represents the fluctuating dimensionless total pressure increase, m?
the fluctuating discharge flow coefficient, p, the fluctuating inlet cavitation
number and M, the fluctuating inlet flow coefficient. All these fluctuations
are supposed to be small compared to average quantities. The term [Z] is
called the transfer matrix whose coefficients are complex in time and are
functions of a reduced frequency (e.g. k = w/fi) where w is the oscillation
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
frequency and ft, is the shaft rotational speed as before. (Additional terms
are needed if the shaft speed is not constant.) Certain of the terms of [Z] may
be estimated if k « 1, that is, the oscillations are quasi-steady, from nor
mal steady pump performance data (see Sack and Nottage, 1965 for one such
example). Measurements of all four (complex) components of [Z] were made
at Caltech about 1975 (see the summary paper of Brennen, 1978) and it was
found that all four complex terms could be determined; they were found to
be a function of reduced frequency and to be dependent upon the cavitation
index. In other respects they follow normal pump similarity laws. In ad
dition, for the more extensive cavitation found in inducer pumps, Brennen
and Braisted (1980) found that the cavitating flow pump could put energy
into the oscillations, that is, the pump became an "active" element rather
than a "passive", consisting of capacitive, resistive and inertance lumped
parameter elements as earlier works. More recent developments at Sulzer
Bros, with a different experimental approach (Stirnemann et al., 1987) have
made similar dynamic measurements on a centrifugal pump with Ns ~ 0.6
typical of a boiler feed pump stage. They find more recently (Bolleter et

30 Cavitation
al., 1990) that in the cavitation domain typical of these pumps (less than
3 percent head loss), that the very simple passive representation mentioned
above was sufficient in analysing the acoustics of hydraulic systems and, in
effect, revert to earlier lumped-parameter estimates. It may be mentioned,
anticipating the results of Chapter 7, that these quasi-steady lumped param
eter terms have the following physical interpretations: The real part of Zne
corresponds to the slope of the standard cavitation performance curve (Fig.
2.11), and the real part of Z12 corresponds to the slope of the normal pump
characteristic curve. The effects of inertia due to liquid oscillation through
the pump correspond to the imaginary part of Z12. The outlet and inlet mass
flow rates can differ if the effective density within the pump changes with
time; this includes both the effect of the volume of the cavitation formed
as a function of pressure as well as the liquid compressibility and elasticity
of the housing. Thus the term Z21; and finally the angle of attack on the
blade inlet edges can change the cavity volume and thereby one is led to
the term Z22. These terms are the focus of the work reported by Brennen
(loc cit) and more recently by Nishiyama (Nishiyama and Nishiyama, 1980).
In essence, Bolleter's observation is that for large hydraulic systems which
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
dominate the dynamics, and for a "small" amount of cavitation a cavitating

pump is "passive"; quasi-steady measurements from the pump characteris
tic are sufficient for real Z\2, the liquid compressibility is nearly sufficient
for Z21, Z\\ and Z22 may be neglected and a geometric estimate of liquid
inertia is sufficient for imaginary Z\2- This is not the case, however, for the
cavitating behaviour of inducer pumps where all four terms of the transfer
matrix must be evaluated.
2.4 Cavitation Nucleation and Inception

2.4.1 Inception
The variability of the cavitation inception index is readily imagined with
the different types of cavitation exhibited on one body such as the ITTC
test body of Fig. 2.2. Inception indices for this body are shown in Fig.
2.14 and they all reveal a significant scale effect (i.e. a + Cp min 7^ 0). Many
similar reports are given in Arndt's (1981) review for related bodies (e.g. the
hemisphere body) and for non separating bodies (e.g. Fig. 2.1, see also very
similar results from Xu, Chen and Pan, 1986) and for the effects of roughness
8
Using the usual matrix notation

Corr**f4 rvtotti*
tt
Figure 2.14: Values of cavitation inception index on the test body of Fig.
2.2 (Johnsson, 1969 [2.41])
on otherwise smooth bodies, wall shear layers and free shear layers (jets).
These are all complicated and inter-related effects due to the two types of
scale effect previously mentioned. At present from Arndt (1981), Rood and
other sources, there seems to be a general consensus that cavitation that
forms in pre-existing regions of a laminar separation (Types 2,4,5 on smooth
bodies) has an inception index of the form
'Cps Cpt (2.4)
where —CpS, denotes the mean pressure coefficient near the end of the sep
aration region and Cpt is a " measure" of the unsteady pressure fluctuations.
Generally,CT;< —Cp min except when mass diffusion (Holl, 1969) drives the
process and gaseous cavitation results can be seen in Fig. 2.15 for the cases
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

32 Cavitation
125
15 63 IS IS
RELATIVE IMPELLER FLOW RATE, Q/Qta
Figure 2.15: Cavitation inception data at the inlet of the pump shown in
Fig. 2.5 (Stoffel and Ludwig, 1990 [2.72]).
These data were obtained visually for one set of gap clearances.
j j > —Cp min. The characteristics of the turbulent pressure fluctuations are
not well known, although rms measurements for flows near turbulent reat-
tachment locations and laminar/turbulent boundary layer transitions may
exceed several percent of the dynamic pressure. In the absence of a pre
existing laminar separation either travelling bubble (Type 1) or attached
cavities (Types 6, 7, 8) may appear at inception. In these cases including
travelling bubble cavitation
where Cp m j„ is the body minimum pressure coefficient and in some cases it

has been suggested that
°i ~ -Cptr - Cptt (2.5)
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
where Cptr is the pressure coefficient at the site of transition (itself not a
constant in real flows) and Cptt an equivalent unsteady term caused by tur
bulent pressure fluctuations. Again, although these may approach 20% of
the dynamic pressure, little is known of the statistical spatial or temporal
properties of these fluctuations. The last case, 12, pure wall shear flow is a
limiting case and Arndt (1969) gives the relationship
CTj = l6Cf (2.6)

Where cj is the turbulent shear stress coefficient.

Travelling bubble cavitation inception on non-separating bodies is a rate-
dependent phenomenon and clearly depends upon the number density distri
bution of microbubbles. The attached forms of inception are not so depen
dent on these nuclei and are relatively independent of dissolved air content.
Xu et al.(1986) clearly are the first to show quantitatively the effect of mi-
crobubble concentration on the transition of a travelling bubble cavitation
o n a c p m j n = —1.0 Schiebe body with nuclei (about 2-5 per cc of 40 mi
crometer diameter) to a "patch" cavity with about one-half as many nuclei
of less than about 30 micrometers diameter. The nucleation effects become
serious in extreme conditions of deaerated liquid such as the vacuum towing
tank at Wageningen (Marin). There it shows clearly (Kuiper, 1979) that
freestream nuclei must be supplied and if these are not enough, local nuclei
must be supplied from artificial roughness (Figure 2.16). Artificial rough-
Without electrolysis With electrolysis
Figure 2.16: Photographs showing the effect of electrolysis-generated cavi

tation nuclei in a de-aerated towing tank (Kuiper, 1979 [2.50])
ness as a source of nucleation is not without its problems, but in a different

context may be used to insure turbulent flow on model scales avoiding any
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

34 Cavitation
separation and then the inception rule tends to follow Eq. 2.5 (Huang, et
al., 1986; Huang, 1984) but is still nuclei dependent. Despite these advances
in technique and nuclei measurement methods it does not yet appear pos
sible to predict beforehand the onset of attached cavitation or travelling
bubble cavitation on non-separating arbitrary bodies with arbitrary nuclei
concentrations and dissolved gas content.
All of these considerations apply to hydraulic machines except that the
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
additional complexities of these flows call into play the other types of cavita
tion as shown in Fig. 2.15; here as mentioned before the "blade" cavitation
includes attached as well as travelling bubble types. Determination of the
inception point is not readily possible in most hydraulic machines hence
the growing interest in acoustic detection already mentioned (Gopalakrish-
nan, 1985; Gulich, 1986, 1989; McNulty and Pearsall, 1979; McNulty, 1985)
and there does appear to be a useful correlation between noise and erosion.
But, as Gopalakrishnan emphasises, erosion is so important for long life in
commercial/industrial application that visual observation of the blade cavi
tation is often required as an acceptance procedure. Gulich, in an extensive
series of tests, correlates the relationship between the "bubble length" of
blade cavitation at a pump inlet, erosion and empirically finds a cavitation
noise-bubble length correlation for one type of pump in a specific acoustic
measuring location (relative to the impeller inlet). But it is not often pos
sible in a commercial development visually to observe cavitation inception
in a pump as reported by Grein, Gulich and others and alternative means
based on the erosion of "soft coatings" have been widely used in recent years
(Gopalakrishnan, 1985, among others) to infer the presence of cavitation on
the inlet portions of a pump. Such "witnesses" to cavitation coincide with
the collapse of travelling bubbles (leading to the "pits" described in Knapp
et al., 1979) or the termination region of a sheet cavity already known to
be a source of great cavitation intensity, and they fulfill a useful but limited
role in the prediction of the subsequent erosive wear.
2.4.2 Nuclei Measurement

The role of "cavitation nuclei" as explained in the famous monograph by
Knapp et al. points out the necessity of nucleation but understates if any
thing just how important and subtle these effects are for the inception and
development of cavitation and its "effects". This growing appreciation has
led to a number of different approaches in trying to define what nucleation
is in liquids, its thermodynamic basis for pure fluids, the effects of micropar-

ticulates (organic, inorganic), properties of the solid surface of test bodies,

chemistry and many other physical effects beyond our present scope. There
seems to be agreement that some of solid particulates normally found in test
facilities and the natural waters of oceans and rivers serve as nuclei but quan
tifying this possibility has remained elusive. At the same time it is certain
that gas bubbles within the liquid are suitable nucleation sites for travelling
bubble cavitation and the other forms as well. Accordingly, a dual approach
has been developed over the years; one is to measure by means of a standard
cavitation experiment the susceptability of the liquid to cavitation (thereby
including all types of nuclei), and the other is to deduce by direct measure
ment the concentration and size distribution of the particulates within the
liquid and to classify them as either "solid" or microbubbles. Many examples
of both approaches can be found in the literature and they are systematically
reviewed in the International Towing Tank Tri-annual Conferences (Huang,
1990). The early development of "susceptibility" devices is traced in Arndt
(1981); the basic idea is to detect the frequency of travelling bubble col
lapse events as a function of a known (or deduced) minimum pressure in a
flow field of known geometry. Based upon the mechanics of spherical bubble
growth, Chapter 3, an initial nucleus size and concentration can be inferred.
The detection can be optical or acoustic. Both external flows or internal
flows such as a Venturi tube can be used and it appears that the Venturi
tube has received the most development from its first use by Oldenziel (1979)
and subsequent modification to include a centre body (Lecoffre, 1987). A
more recent variation of a Venturi tube with an adjustable centre body (Gin-
droz, 1990) appears to give the potential of measuring "cavitatable" nuclei
to diameters of less than 5 micrometers 7 in concentrations up to 10 nuclei
per millilitre. Optical detection methods reviewed in the current I.T.T.C.
survey of nuclei measurement methods (Huang, 1990) include light scatter
ing, phase Doppler anemometry (PDA) and direct holography with pulsed
ruby lasers. These holographic measurements can classify microparticulates
(solid or microbubble) from about 10 micrometers and larger. In principal,
light scattering instruments can detect smaller sizes. A direct comparison
of the holographic, light scattering, PDA and Venturi systems (not includ
ing the Gindroz development) was carried out at Penn State University;
these comparisons are not yet fully available, although it does appear that
7
Present holographic methods have a practical limit of about 10 micrometers. If ac
tive nuclei in tests are shown to be less than this value, new schemes capable of exact
calibration, possibly the PDA will have to be developed.
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

36 Cavitation
direct holographic methods and the PDA approach are comparable to mi-
crobubbles in the absence of many solid particulates. There seems to be
some problem for light scattering and the PDA methods when the test fluid
contains both microbubbles and solid microparticulates; clearly further de
velopment is needed and appears to be underway. Based upon these results,
the I.T.T.C. cavitation committee concludes that the cavitation susceptibil
ity devices (the Venturi tubes) be used as a reference device for continu
ously monitoring during cavitation tests; that significant progress has been
made with phase Doppler anemometry, but that holography remains the
best method to discriminate solid and gas microparticulates.
It is of course possible to detect microbubbles acoustically; indeed this is
a standard oceanographic technique (Akulichev et al., 1986; Medwin, 1977).
More recently, the attention of acoustic beams by microbubbles in a water
tunnel test facility has been used to infer microbubble distributions (Taka-
gawa, 1989) which appears to have avoided the difficulties with these kinds
of measurements for internal flows in water tunnels and pumps.
In closing this section we review briefly the number and size distributions
of nuclei found in different circumstances; Fig. 2.17 shows such a "global"
collection from many sources. Note that at a given size, say 20 to 50 microm
eters, there are five orders of magnitude difference in these populations. The
shape of all these distributions is of the form N(R) ~ Rn where n lies be
tween 3 and 4. There must be a cut off size below which N(R) decreases but
this is not known for the present data. One may ask what are the important
sizes for cavitation inception on the bodies and in the machines described
for travelling bubble cavitation (the only form for which bubble mechanics
closely applicable). We may follow Knapp et al. to state that the critical
bubble radius from which vaporous growth may proceed is given by
<7i+ C P m i n < - — — (2.7)

o i l Cj
where S is the surface tension (N/m), R is the radius of the bubble at the
location where the pressure is minimum and q is the free stream dynamic
pressure. For example, from the test data of Fig. 2.14 for a test speed
of 13 m/s , values of Oi + Cp m j n — —0.25 are evident and from Eq. 2.7
we obtain a value of R = 4.6 micrometers (taking 5" = .073 N/m). The
implication of the inception data and this result is that no larger bubbles
were available to cavitate. But this critical formulation is a quasi-static
process; on real bodies the physical space and time available for classical
bubble growth are all limited so that even much larger bubbles will not
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

10''
PETERSON el ol
71975) to-=0.49,
SCATTERING
O
PETERSON elol
(I975)(<7»0.49 1
HOLOGRAPHY) ARNDT 8> KELLER
~ I0I2£ (1976)
AIR C O N T E N T -
12.5 ppm
O
KELLER 8
I- GAVRILOV WEITENOORF
o (1970) (1976)
-z. (FRESH WATER) GASSED WATER
3
10' AIR CONTENT-
3 0 ppm
KATZ0978) .
m BILLET
(1985)
\-
a 10" -BEERS &REID ARNDT a
KELLER (1976)
MICROPLANKTI
AIR CONTENT-
(1973) 6.3 ppm
CO KOOAMA et a I.
z (1964)
MEDWIN (1977)
cc GAVRILOV (1970) (OCEAN, AUGUST)
UJ
CD 10s JAFTER STANDING
5 HOURS) /■ GATES
5 / (1978)
MEDWIN (1977)
(OCEAN, FEBRUARY)
I0»
RADIUS R (micrometers)
Figure 2.17: Summary plot of the distribution of microparticulates and mi-

crobubbles from several investigators measured by a variety of techniques
(Murai, 1986 [2.62])
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

38 Cavitation
be physically observed as shown in the sequence of Fig. 2.6. In a recent

important numerical simulation of travelling bubble inception; Meyer (Meyer
et al., 1990) showed that on a 5.08 cm Schiebe body (cp m , n = —0.75),
to reproduce the observed value of a, = 0.55 at a flow speed of 9 m/s
the critical radius was 11.7 micrometers; full stochastic simulations over a
variety of conditions which were found to compare favourably with the actual
experiments demonstrating the need for microbubbles with a range of 25-50
micrometers diameter, long thought to be the desirable range in cavitation
test facilities. From this finding the nuclei sizes quoted by Gindroz all seem
rather "small".
2.4.3 Concluding Remarks

The importance of the nuclei population in hydrodynamic applications has
been known for some years. It now appears this is also true for hydraulic
machines, although there is still much less documentation than needed of
nuclei or microbubble populations in this field. In part, this is because a
calibrated nuclei measurement method for laboratory use has yet to be ac
cepted although there is a reasonable expectation that this will occur in the
near future. Even in the case where the nuclei distribution is known a pre
diction of the type of cavitation and inception index cannot yet be made
with certainty except for travelling bubble cavitation. The general trends
are known; namely, that with an adequate supply of nuclei, travelling bub
ble cavitation is experienced; with fewer nuclei, attached cavitation is seen
at inception even on bodies without separation. These attached forms are
typically highly three-dimensional wedges or patches and it is as if the liq
uid flow under the tensions sustained prior to inception becomes unstable to
three-dimensional growth due to a random nucleation and forms the attached
region. Even attached forms stabilised in regions of a prior laminar separa
tion or for the attached forms just mentioned have highly three-dimensional
cavitation detachment interfaces. There appears to be evidence that these
real-fluid effects (Avellan, et al., 1988) at the leading cavity edge of a play
role in the high degree of rotationality or vorticity seen in the regions of
the sheet cavity collapse; the vortical nature of these flows is a property
of the liquid flow itself, not the internal cavitation, and thus would ap
pear to be a manifestation again of the real-fluid or viscous property of the
three-dimensional flow in which the cavitation is taking place. Thus, it is
conceivable or even likely that the evolution of individual travelling bubble
cavitation events may have an important viscous flow interaction particu-
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

larly in the last phases of collapse/rebound. It appears to be important to

understand these vortical flows in greater detail, particularly for cloud cavi
tation, because the resulting intensity of collapse leads to erosion. Similarly,
in pumps it is necessary to know the inception of cavitation in its various
forms for the prediction of long life.
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

40 Cavitation
References
2.1 Acosta, A.J., 1973, "Hydrofoils and Hydrofoil Craft", Ann. Rev. Fi.
Mech., Vol. 5, pp. 161-185, Van Dyke, M., Vincenti, W.G. and We-
hausen, J.G. (eds.).
2.2 Acosta, A.J. and Parkin, B.R., 1980, "Report of the ATTC Cav
itation Inception Committee", Ann Arbor Science Pub. Inc., 230
Collingwood, P.O. Box 1425, Ann Arbor, MI 48106.
2.3 Akulichev, V.A. and Ilychev, V.I., 1986, "Acoustic Cavitation in Sea
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
Water", Proc. Int. Symp. on Propellers and Cavitation, Wuxi, China,
pp201-206.
2.4 Anton, I., 1984, Cavitatia, Vol. 1, 2, Editura Academei, Bucuresti

(in Romanian).
2.5 Arakeri, V.H. and Acosta, A.J., 1973, "Viscous Effects in Inception
of Cavitation on Axisymmetric Bodies", J. Fl. Eng., Trans. A.S.M.E.
2.6 Arakeri, V.H., 1975, "Viscous Effects on the Position of Laminar

Separation from Smooth Bodies", J. Fl. Mech. Vol. 68, Part 4, p.
779.
2.7 Arakeri, V.H. and Shanumugathan, V., 1985, "On the Evidence for
the Effect of Bubble Interference on Cavitation Noise", J. FI. Mech.,
Vol. 159, pp. 137-150.
2.8 Arndt, R.E.A. and Daily, J.W., 1969, "Cavitation in Turbulent

Boundary Layers", in Robertson and Wislicenus (1969), pp. 64-87.
2.9 Arndt, R.E.A., 1981, "Cavitation in Fluid Machinery and Hydraulic

Structures", Ann. Rev. Fi. Mech., Vol. 13, pp. 273-328.
2.10 Arndt, R.E.A. and Webb, D.R. (eds.), 1985, Cavitation in Hydraulic
Structures and Turbomachinery, A.S.M.E. FI. Eng. Div., Vol. 25.
2.11 Avellan, F., Gindroz, B., Henry, P., Bachmani, P., Vulliod, G. and
Wegner, M., 1986, "Influence of the Test Head and of the Water Nu
clei Content on the Cavitation Performance of Water Turbine Mod
els", 13th Symp. Sect. Hydraulic Machines, Equipment and Cavita
tion, Int'l. Assoc, Hyd. Res., Montreal.

2.12 Avellan, F., Dupont, P. and Rhyming, I., 1988, "Generation Mech
anism and Dynamics of Cavitation Vortices Downstream of a Fixed
Leading Edge Cavity", 17th Symposium on Naval Hydrodynamics,
The Hague.
2.13 Brennen, C.E., 1978, "The Unsteady, Dynamic Characteristics of Hy

draulic Systems with Emphasis on Cavitation and Turbomachines",
Joint Symp. on Design and Operation of Fluid Machinery, I.A.H.R.,
A.S.M.E., A.S.C.E., Fort Collins.
2.14 Brennen, C.E. and Braisted, D.M., 1980, "Stability of Hydraulic Sys
tems with Focus on Cavitating Pumps", 10th I.A.H.R. Symp. on Hy
draulic Res., Tokyo, pp. 255-269.
2.15 Bolleter, U., Stirnemann, A. and Eberl, J., 1990, "The dynamic
Transfer Matrix of a Pump and its Use in Pumping System Design,"
article in Kim and Yang (1990).
2.16 Ceccio, S.L. and Brennen, C.E., 1990a, "Observations of the Dynam
ics and Acoustics of Attached Cavities", Cavitation and Multiphase
Flow Forum, Furuya, 0. (ed.), A.S.M.E., Vol. 98, pp. 79-84.
2.17 Ceccio, S.L. and Brennen, C.E., 1990b, "The Dynamics and Acoustics
of Travelling Bubble Cavitation", Eighteenth Symposium on Naval
Hydrodynamics, Ann Arbor, MI.
2.18 Chahine, G.L., 1982, "Cloud Cavitation Theory," 14th Symposium

on Naval Hydrodynamics, Ann Arbor, MI.
2.19 Chauvin, J. (ed.), 1973, Recent Progress in Pump Research in Karman

--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
Inst. for FI. Dyn. Lecture Series 61.
2.20 Chebayevskii, V.F. and Petrov, V.J., 1973, Cavitation Character

istics of High Speed Inducer Pumps, U.S. Nat. Trans. Tech. Serv.
AD-A004-520 (Kavitatsionnyye Kharakteristiki Bysokooborotnykh
Shneko Tsentrobezhnykh Nasosov, USSR, 1973).
2.21 Cimbala, J.M. and Billet, M.L., 1989, "Cavitation in Rectangular

Slots", article in Morgan and Parkin (1989).
2.22 d'Agostino, L. and Brennen, C.E., 1983, "On the Acoustical Dy
namics of Bubble Clouds", Cavitation and Multiphase Flow Forum,
Houston.

42 Cavitation
2.23 Edstrand, H., 1984, "Remarks on Ship Model Testing, Facilities and
Test Results", David Taylor Naval Ship Research and Development
Center Rept. DTNSRDC-83/054.
2.24 Falcao de Campos, J.A.C., George, M.F. and Mackay, M., 1989, 'Ex
perimental Investigation of Tip Vortex Cavitation for Elliptical and
Rectangular Wings", Cavitation and Multiphase Flow Forum, Fu-
ruya, O. (ed.), A.S.M.E., La Jolla.
2.25 Franc, J.P. and Michel, J.M., 1985, "Attached Cavitation and the
Boundary Layer and the Boundary Layer ...", J. Fl. Mech., Vol. 154,
pp. 63-90.
2.26 Furuya, O., 1975, "Exact Supercavitating Cascade Theory", J. Fl.

Eng., A.S.M.E. Vol. 97, pp. 419-429.
2.27 Furuya, O. (ed.), 1990, Cavitation and Multiphase Flow Forum -
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
1990, A.S.M.E. Fed. Vol. 98.
2.28 Gates, E.M. and Acosta, A.J., 1978, "Some Effects of Several
Freestream Factors on Cavitation Inception on Axisymmetric Bod
ies", 12th Symposium on Naval Hydrodynamics, Washington, D.C.
2.29 Gates, E.M. and Billet, M.L., 1980, "Cavitation Nuclei and Incep
tion", 10th IAHR Symp. on Hyd. Res., Tokyo, pp. 3-24.
2.30 Gindroz, B., 1990, "A New Nuclei Counter", Report from Inst. de
Mach. Hydronlique et de Mechanic des Fluides, Ecole Polytechnique
Federal de Lausanne, 1015 Lausanne, Switzerland.
2.31 Gopalakrishnan, S., 1985, "Modern Cavitation Criteria for Centrifu

gal Pumps", 2nd International Pump Symposium, Turbomachinery
Lab., Texas A&M Univ., College Station, TX, Childs, D. and Bailey,
J.C. (eds.), pp. 3-11.
2.32 Grein, H., 1973, "Cavitation", article in Chauvin (1973).
2.33 Giilich, J.F., 1989, "Guidelines for the Prevention of Cavita

tion in Centrifugal Feedpumps", EPRI Report GS-6398. See also,
1986, "Quantitative Prediction of Cavitation Erosion in Centrifugal
Pumps", 13th Symposium, Section on Hydraulic Machinery, Equip
ment and Cavitation. Int'l. Assoc. Hyd. Res., Montreal.

2.34 Hart, D.P., Brennen, C.E. and Acosta, A.J., 1990, "Observations of
Cavitation on a Three-dimensional Oscillating Hydrofoil", article in
Furuya (1990).
2.35 Henry, P., 1978, "Influence of the Amount of Bubble Nuclei on Cav
itation Tests of a Francis Turbine," Cavitation and Polyphase Flow
Forum, A.S.M.E., Fort Collins, pp. 23-28.
2.36 Holl, J.W., 1969, "Limited Cavitation", Robertson and Wislicenus
(1969), pp. 26-64.
2.37 Huang, T.T., 1964, "The Effects of Turbulent Stimulation on Cavi
tation of Axisymmetric Headforms", in Morgan and Parkin (1964).
2.38 Huang, T.T. and Shen, Y.T., 1986, "The Application of Turbulence
Stimulator to Reduce Scale Effect on Cavitation", in Anon. (1986),
pp. 182-190.
2.39 Huang, T.T., 1990, Report of the Cavitation Committee, Interna
tional Towing Tank Conference, Madrid.
2.40 Janigro, A. and Ferrini, F., 1973, "Inducer Pumps", article in Chau-
vin (1973).
2.41 Johnsson, C.A., 1969, "Cavitation Inception on Headforms, Further
Tests", 12th I.T.T.C, Rome, pp. 381-392.
2.42 Kanijo, K., Suzuki, A., Shimura, T., Hashimoto, R., Watanabe, M.,
Watanabe, Y., Iwabuchi, T. and Mori, Y., 1975, "Experimental In
vestigation of Small, High-Speed, High-Head Liquid Oxygen Pump",
N.A.L. TR-415 (in Japanese).
2.43 Kato, H., 1985, "On the Structure of Cavity - New Insight into the
Cavity Flows", Int'l. Symp. Jets and Cavities, Kim, J., Furuya, O.
and Parkin, B.R. (eds.), A.S.M.E., F.E.D., Vol. 31.
2.44 Keller, A.P., 1979, "Cavitation Inception Measurement and Flow Vi
sualisation on Axisymmetric Bodies at Two Different Free Stream
Turbulence Levels and Test Procedures", article in Margan and
Parkin (1979).
2.45 Kim, J.H. and Yang, W.J. (eds.), 1990, "Third International Sympo
sium on Transport Phenomena and Dynamics of Rotating Machin
ery", ISROMAC-3, Hemisphere.
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

44 Cavitation
2.46 Knapp, R.T., 1952, "Cavitation Mechanics and its Relation to the
Design of Hydraulic Equipment," Proc. (a) Inst. Mech. Engr., Vol.
166, pp. 150-163.
2.47 Knapp, R.T., Daily, J.W. and Hammitt, F.G., 1979, Cavitation, Inst.
Hydraulic Research, U. Iowa, Iowa USA.
2.48 Kubota, A., 1988, "Numerical Studies of Unsteady Cavitation on a

Hydrofoil by a Bubbly Two-Phase Flow Model", Ph.D. Thesis, U.
Tokyo, Dept. Nav. Arch., H. Kato, Dir.
2.49 Kubota, S., Kato, H., Yamaguchi, H. and Maeda, M., 1987, "Un
steady Structure Measurement of Cloud Cavitation on a Foil Section
Using Conditional Sampling Technique", Int'l. Symp. on Cavitation
Research Facilities, A.S.M.E., F.E.D. Vol. 57, Holl, J.W. and Billet,
M.L. (eds.), pp. 161-169.
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
2.50 Kuiper, G., 1979, "Some Experiments with Distinguished Types of
Cavitation on Ship Propellers", Morgan and Parkin (1979), pp. 171-
207.
2.51 Lecoffre, Y., 1987, 'Procedures and Instrumentation for Monitoring

Gas Content in Cavitation Test Loops", Int. Symp. on Cavitation
Research Facilities, Holl, J.W. and Billet, M.L. (eds.), A.S.M.E.,
Boston.
2.52 Li, S., Ni, A., Huang, J. and Xu, F., 1986, "Estimation of Incipi
ent Cavitation Number by a Probability Method", Int'l. Symp. on
Propeller and Cavitation, Wuxi, China (1986), pp. 281-285.
2.53 McNulty, P.J. and Pearsall, I.S., 1979, "Cavitation Inception in

Pumps", in Morgan and Parkin (1979), pp. 163-170.
2.54 McNulty, P.J., 1985, "Detection of Cavitation Inception by a Single

Point Acoustical Measurement", article in Arndt and Webb (1985).
2.55 Medwin, H., 1977, "Acoustical Determination of Bubble Size Spec

tra", J. Acoust. Soc. Am., Vol. 62, pp. 1041-1044. Also see Medwin,
H. and Breitz, N.G., 1989, "Ambient and Transient Bubble Spectral
Densities in Quiescent Seas and Under Spilling Breakers", J. Geo-
phys. Res., Vol. 94, No. C9, pp. 12,751-12,759, Sept. 15.

2.56 Meyer, R.S., Billet, M.L. and Holl, J.W., 1990, "Free-Stream Nuclei
and Cavitation", in Morgan and Parkin (1990), pp. 55-62.
2.57 Morgan, W.B.and Parkin, B.R. (eds.), 1979, International Sympo

sium on Cavitation Inception, A.S.M.E.
2.58 Morgan, W.B. and Parkin, B.R. (eds.), 1984, Int'l. Symp. Cavitation,
A.S.M.E.
2.59 Morgan, W.B. and Parkin, B.R. (eds.), 1989, Int'l. Symp. Cavitation,
A.S.M.E.
2.60 Murai, H and Kinoe, T., 1968, Rep. Inst. High Speed Mechanics
(Sendai), Vol. 20, pp. 263-285.
2.61 Murai, H. and Itoh,, H., 1985, "Pressure Fluctuations in Axial Flow
Pump Induced by Transient Cavitation", article in Arndt and Webb
(1985), pp. 111-118.
2.62 Murai, H. (ed.), 1986, International Symposium on Cavitation,

Sendai.
2.63 Narasaka, T., Oba, R. and Hirata, Y., 1990, "Cavitation Observations
Around the CY4-Hydrofoil by Means of High Speed Photography",
Rept. Inst. of Fl. Science, Tohoku Univ., Vol. 1, pp. 1-13.
2.64 Nishiyama, H. and Nishiyama, T., 1980, "Dynamic Transfer Charac

teristics of Partially Cavitated Hydrofoil Cascade," 10th IAHR Symp.
on Hyd. Res., Tokyo, pp. 243-254.
2.65 Oldenziel, D.M., 1979, "New Instruments for Cavitation Research",

in Morgan and Parkin (1979), pp. 111-124.
2.66 Ooi, K.K. and Acosta, A.J., 1983, "The Utilization of Specially Tai
lored Air Bubbles as Static Pressure Sensors in a Jet', J. Fl. Eng.,
Vol. 106, No. 4, pp. 459-465.
2.67 Robertson, J.M. and Wislicenus, G.F., 1969, Cavitation State of

--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
Knowledge, A.S.M.E.
2.68 Rood, E.P., 1989, "Mechanisms of Cavitation Inception", in Morgan

and Parkin (1989).

46 Cavitation
2.69 Sacks, L.E. and Nottage, H.B., 1965, "System Oscillations Associated
with a Cavitating Inducer", / . Basic Eng., A.S.M.E. Ser. D., vol. 87.
2.70 Schiebe, F.R., 1972, "Measurements of the Cavitation Susceptibility
of Water Using Standard Bodies", Rept. 118, St. Anthony Falls Hyd.
Lab., Univ. of Minnesota.
2.71 Stirnemann, A., Eberl, J. and Bolleter, U., 1987, "Experimental De
termination of the Dynamic Transfer Matrix for a Pump", J. Fl. Eng.
Trans., A.S.M.E., Vol. 109, No. 3, pp. 218-225.
2.72 Stouffel, B. and Ludwig, G., 1990, "The Influence of the Gap Flow
at the Impeller Inlet on Cavitation and Part Load Recirculation of a
Centrifugal Pump", article in Kim and Yang (1990).
2.73 Takagawa, S., 1989, "A New Acoustic Attenuation Method for Profil
ing of Cavitation Nuclei Distribution", in Morgan and Parkin (1989),
pp. 73-78.
2.74 Takamatsu, Y., Ishizaka, K., Kishikawa, T. and Kusaka, Y., 1980,
"An Investigation of Tandem-Bladed Helical Inducer for Cavitation-
Resistant Pump", 10th I.A.H.R. Symp. on Hydraulic Res. Sect, for
Hyd. Mech., Equip, and Cavitation, Tokyo, pp. 291-303.
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
2.75 Valensek, B., 1983, "Pump Prestages for Extreme Suction Heads",
Mechanical Engineering Journal, Turboinstitut, Ljubljana.
2.76 van der Muelen, J.H.J., 1986, "On Correlating Erosion and Lumines
cence from Cavitation on a Hydrofoil", article in Anon. (1986), pp.
261-267.
2.77 van der Muelen, J.H.J. and Wijnant, I.L., 1990, "On the Structure
and Intensity of Sheet Cavitation", in Cavitation and Multiphase
Flow Forum, Furuya, O. (ed.) loc cit.
2.78 Wood, G.M. and Whippen, W.G., 1969, "Cavitation Effects in Turbo-
machinery", article in Robertson and Wislicenus (1969), pp. 148-165.
2.79 Wu, T.Y., 1972, "Cavity and Wake Flows", Ann. Rev. Fl. Mech., Vol.
4, pp. 243-285.
2.80 Xu, W., Chen, G. and Pan, S., 1986, "The Influence of Cavitation
Nuclei on Cavitation Inception on the Smooth Axisymmetric Bodies",
article in Anon. (1986).

Chapter 3
Bubble Dynamics
Part 1: Single Bubble
A. Shima
Single bubble behaviour is the basis of bubble dynamics, helping to build
an understanding of the cavitation problems involved in hydraulic machin
ery. In particular, the bubble collapse is responsible for the generation of
impulsive pressures, shock waves, noise, and cavitation damage, etc. These
subjects are well documented 1 . Therefore, only a brief review, with the
emphasis on bubble collapse, is given.
3.1 Rayleigh Analysis

3.1.1 Besant's Problem
The basis for theoretical studies on cavitation bubbles was made by Rayleigh
[3.12]. He revealed the motion of an empty spherical bubble by solving the
Besant's problem [3.2]. As formulated by Besant, the problem is "An infinite
mass of homogeneous incompressible fluid acted upon by no forces is at rest,
and a spherical portion of the fluid is suddenly annihilated; it is required to
find the instantaneous alteration of pressure at any point of the mass, and
the time in which the bubble will be filled up, the pressure at an infinite
distance being supposed to remain constant".
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
'For example, Cavitation (Knapp, Daily and Hammitt, 1970, McGrow-Hill); Cavitation
and Bubble Dynamics (Brennen, 1995, Oxford); and, several review articles.
47
48 Cavitation
3.1.2 Rayleigh's Solution

Considering the bubble-motion in an inviscid and incompressible liquid, if
R and R are the velocity and the radius at the bubble surface at time t,
respectively, and u is the radial component of velocity at any distance r
from the centre, the following equation is derived (Figure 3.1).
Figure 3.1: A spherical bubble.
i =£ (3-D
R r2
Using eqn. (3.1), the kinematic energy of the liquid around a bubble is given
by
1 f°°
-p / u24irr2dr = 2wpR2R3, (3.2)
2 JR
where p is the liquid density.
The work done by the liquid in which the bubble radius changes from
Ro to R is
PooAV = 4-^(Rl-R% (3.3)
where p ^ is the pressure of liquid at an infinite distance from the bubble;
and, the pressure at the bubble surface is assumed to be zero.
Equating equation (3.2) with equation (3.3), we obtain
#=!&[(*,._„. ,3.4)
Then
= 5L/* r-Ro #3/2
(R*-R*y/*dR- (3
-5)
V 2Poo JR
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

Bubble Dynamics: Single Bubble 49
Let f3 = •#■, eqn. (3.5) becomes

03/2
Wfe-i 0(1-/P)!/^- (3-6)

Hence, the time T for completing the collapse is proportional to Roy/p/p0
The integral of eqn. (3.6) in terms of the T function is
^ , - p - r(5/6)r(i/2)
6po<
0.91468iW—. =
(3.7)
V Poo
From eqn. (3.4), we can see that the velocity R becomes infinite as
R approaches zero. Although the actual collapse time is usually greater
than the one predicted by eqn. (3.7), this is often used as an approximate
collapse-time [3.15].
Rayleigh also solved the problem of a bubble filled with a permanent gas.
His result shows that the gas plays a retarding role: preventing the complete
collapse of the bubble, making it rebound and oscillate.
3.2 Vaporous and Gaseous Bubbles

3.2.1 Classification
The onset of cavitation is characterised by the occurrence of tiny vapour
bubbles and/or gas bubbles in the liquid. As regards bubble-growth, the
explosive growth is an attribute of vapour cavitation while gas cavitation
is characterised by a slow diffusion process. Blake [3.3] characterised the
vaporous cavitation by a sudden expansion of bubbles, and the gaseous cav
itation by the growth of nuclei due to gas diffusion from the surrounding
liquid. Actually, vaporous cavitation is always accompanied by gas diffu
sion into the bubbles. However, the diffusion process is usually very slow
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
compared with the rapid expansion of the vapour bubbles. Therefore, it is

practical to ignore the diffusion effect in many cases of vaporous cavitation.
3.2.2 Stability
According to Plesset and Mitchell [3.16], for nucleus-expansion, a bubble
with a spherical shape is stable as long as A/R (ie the ratio of the deforma
tion amplitude A to the initial bubble radius Rg) is small.

50 Cavitation
For the collapse of a vaporous bubble, its deformation is small for 1 >
R/R0 > 0.2. As R -> 0, the deformation increases in proportion to i ? - 1 / 4 ,
and the bubble becomes unstable.
For a collapsing gas-bubble, provided that the initial gas pressure in the
bubble is not extremely low, the bubble keeps a spherical form quite well
almost until its minimum radius.
3.3 Viscosity and Compressibility Effects

3.3.1 Effects of Viscosity and Surface Tension
Poritsky [3.18] examined the effect of viscosity on the motion of a spherical
bubble with constant internal-pressure. Poritsky's treatment modified the
Rayleigh analysis, resulting in a generalised Rayleigh-Plesset equation [3.4].
M+|i?2 = V - P o o ) - ^ - ~ , (3-8)
2 p pit pH
where a is the surface tension of the liquid, pv is the saturated vapour
pressure and \i is the viscosity.
The numerical analysis by Poritsky [6] shows that if the non-dimensional
viscosity,
M- B ,,4/< v, (3-9)
is fairly large, the bubble motion will be retarded significantly, and if a = 0 is

assumed, for a large value of p, (= 0.46) the collapse time becomes infinite.
Shu [3.30] examined the above trend of Poritsky's numerical results and
analytically proved the existence of such a critical value of p0. That is, the
bubble motion without surface tension leads to a infinite collapsing time if
p! > p0.
3.3.2 Compressibility Effect

In many cases of bubble collapse, the local velocities reach an appreciable
fraction of the sound velocity in the liquid so that the liquid compressibility
is no longer negligible.
Herring [3.7] derived a first-order approximation to include liquid com
pressibility. However, it only deals with the velocities and pressures at the
bubble wall. Trilling [3.35] also obtained a first-order approximate equation.
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

Its solution includes the velocity and pressure fields in the liquid. For the
first-order approximation, the assumption is that the liquid velocity u is
small compared with the sound velocity C in the liquid. A brief review of
their important results is given below.
Assuming that the velocity potential satisfies the acoustic equation for
the diverging spherical waves, Trilling [3.35] derived the following equation
for bubble motion,
where V = dR/dt, C ^ is the sound velocity at infinity, pw is the pressure

in liquid at the bubble wall. With this formulation, he calculated a rebound
process of a perfect, nonconducting, nonviscous gas bubble after an adiabatic
collapse. For p ^ = 0.15 MPa, initial gas pressure po = 2.03 kPa, specific
heats ratio of gas 7 = 1.4 and initial bubble radius RQ = 10 mm, he found
the resultant minimum-radius to be 0.64 mm and the maximum pressure to
be about 220 MPa.
In Gilmore's theory [3.6], he assumes that the quantity, r(h + u2/2),
propagates outward at a characteristic velocity of (c + u), where r is the
distance from the bubble centre and h is the enthalpy of the liquid. The
following bubble-motion equation is obtained.
where
Poo + k>
H = !fc^[(^±4)^_il
!
(3.12)
( n - l ) p \>Poo + B J '
{B w 300 MPa, n ~ 7.0
for water at temperature 293.2 K);
Pw = Pt-^-^x- (313)
For IH |<SC C2 (corresponding to |pi — poo |4C 2000 MPa for water), which is
usually the case in bubble collapse, the following solution can be obtained
from equation (3.10),
V,22 _
= 2(Poo-p
^ ° ° - ^t )[ ( % )3
^R> _ 1] (3.14)
l
3P (l--£)
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

52 Cavitation
If this result (ie V oc R~xl2) is compared with that (ie V oc R~2/2) for the
incompressible case, it is evident that the compressibility slows down the
motion of bubble, in particular for the final stage of collapse.
Apart from the above approximation, Benjamin [3.1] obtained the second-
order approximation of velocity potential <j> by using a perturbation method;
and, Jahsman [3.8] obtained a second-order approximation for a gas bubble
by using a similar approach. Their analytical investigations have been ver
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
ified by Lilliston's numerical calculation [3.11]. Tomita-Shima [3.31, 3.22]
obtained the following second-order approximation of bubble motion by the
PLK approach.
o Louterborn
or \ Irrcomp.
10.0
t x 10 sec
Figure 3.2: Comparison with experimental values of the variation with time
of the bubble radius silicone oil. RQ = 2.525 mm, 7 = 1.33 Shima-Fujiwara
[3.23].
. _ , « „ _ * * + » # , + »#(,_<* + -—
10 ci 3Coo 5 CQQ
R
Pr=R
+ Poo - P2r=ft - + ^2- { 2 A R p i r = R
Coo C
+ P
°°~ P l r = *[i? 2 + 3(Poo Plr=R)
)} (3.15)

where
/•R0x3 7 2<T R
Plr=R = M-R) ~R~A^
P2r=fi = Plr=fl ~ (-7- + C) ^,0

3 pC«,
here, p is the initial gas pressure in a bubble, 7 is the polytropic index, and £
is the bulk viscosity of liquid. In Fig. 3.2 the values derived from eqns. (3.15)
and (3.17) were compared with the experimental results by Lauterborn [3.9]
for the case of silicone oil [3.23]. The compressible solution agrees well with
the experiments, except for the results after the second rebound. The liquid
compressibility makes the bubble radius decreases remarkably.
3.4 Bubble Rebound

3.4.1 Introduction
Our knowledge about the final collapse of a bubble and its subsequent re
bound is very limited. However, it is understood that in these processes the
liquid compressibility and viscosity play more important roles and the effect
of heat transfer is no longer negligible.
3.4.2 Studies of Rebound

The nonequilibrium collapse of a bubble containing both the noncondensible
gas and the vapour in an infinite volume of viscous compressible liquid has
been numerically studied by Tomita and Shima [3.32]. The numerical and
experimental results are shown in Fig. 3.3. The experimental data were
produced by a spark induced bubble in water [3.25]. The numerical approach
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
predicted the bubble motion quite well. A remarkable damping after the
rebound and the heat-transfer effect were also demonstrated.
The study by Fujikawa and Akamatsu [3.5] demonstrated the bubble
behaviour from its collapse to the rebound, and showed a strong influence
of the evaporation and condensation.

54 Cavitation
5.0
Insrlio l _ Re-100.7 kPo
Inflrfio + Heof Transfer] y T„= 295.4 K
L—a?
4.0 o Exper\wet\t
Rmai-3.1 m/n
R, =0.7 mm
E
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
-0.295 -0.2
Figure 3.3: Effect of the heat transfer on the variation of the bubble radius
with time: i? ma x — 3.1 mm, L —> oo. Shima-Tomita [3.25]
3.5 Nonspherical Collapse and Micro-jet Forma

tion
3.5.1 Collapse near Solid Wall
Rattray [3.19] investigated a bubble collapsing near a solid wall by using the
perturbation theory, assuming the deviation from the spherical symmetry
to be small. The result shows that an initially spherical bubble elongates in
the normal direction to the solid wall at the early stage of collapse, and the
collapse is retarded by the existence of the solid wall.
Plesset and Chapman [3.17], using the Liebmann iterative method, stud
ied a vapour bubble collapsing near a solid wall in an inviscid incompressible
liquid. They showed the jet formation at rather early stage of the collapse
and the magnitude order of the jet-impact on the wall. Their results for the
variation of the bubble shape for the case of W (= RQ/21) = 1/3 (where / is

o Experiment
— Theory
y////////////A soiid;îrmm^%
Figure 3.4: Comparison of experimentally determined bubble shapes on col
lapse of a spherical bubble near a solid wall with theoretical curves taken
from Plesset & Chapman. Lauterborn-Bolle [3.10].
the distance between the bubble centre and the solid wall, R0 is the initial
bubble-radius) coincides quite well with the experimental results for a laser
induced bubble by Lauterborn and Bolle [3.10] (see Figure 3.4). Nakajima
and Shima [3.13] showed that the jet formation is decelerated by the liquid
viscosity [3.28] (see Figure 3.5).
3.5.2 Collapse in Contact with Solid Wall

Naude and Ellis [3.14] studied, with a perturbation method, a non-
hemispherical bubble attached to a solid wall in an inviscid and
incompressible liquid. Their study demonstrated that during the collapse
a projecting part appeared at the off-side of the bubble; and, it further
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

56 Cavitation
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
0 0.2 0.4 0.6 0.8 1.0 1.2 1.4

t
Figure 3.5: Jet velocity as a function of time Nakajima-Shima [3.13].
developed into a jet and finally struck against the solid wall. This phe
nomenon was also confirmed by the high-speed photograph of a spark
induced bubble. Shima and Nakajima [3.21], using a variational method,
show that the surface tension accelerates the bubble collapse and accents
the deformation of bubble during the final collapsing stages. However, the
deformation of bubble-shape during the process of collapse will be very dif
ferent if the contact angle a is not 90° as for those studies mentioned before.
Particularly, if the contact angle a < 90°, the deformation at the end of col
lapse will be remarkably large. The projecting part appears with a greatly
increased velocity at the late stage of collapse. Therefore, the jet produced by
such a projecting part possesses a destructive power and will cause damage

if it strikes against the solid wall. This has been experimentally confirmed
[3.24, 3.29).
3.6 Pressures Generated at Collapse

3.6.1 Impact Pressure and Their Modes
An important factor relating to cavitation damage is the generation of pres
sures by collapsing bubbles. Among these crucial pressures, one is the im
pulsive pressures or shock waves occurring during the final stage of bubble
collapse, another is the water hammer pressures caused by the microjet-
impingement.
The maximum impulsive pressures p m a x occurring during the bubble col
lapse can be estimated as follows [3.20, 3.31]:
for incompressible liquid,
J W = Po(7£L)Slr-7p-; (3.17)
for compressible liquid,
2(J
Ro 37 2a 4/i(^+Q , Ro .37 n x ,,itt
7
Pmax = P 0 ( p ) ~ p + „ 2 P 2 r2 l P o ( p ) ~ R Poo}- (3-18)
■n-min •"•-min y n m i n oo ■flmin "min
where po is the initial gas pressure inside the bubble, p ^ is the liquid pressure
at infinity, Ro is the initial bubble radius, i? m i n is the minimum bubble
radius, p is the liquid density at infinity, a is the surface tension of the liquid,
\i is the viscosity of the liquid, 7 is the specific heats ratio of gas, and £ is
the bubble viscosity of the liquid. Such estimated values for the maximum
impulsive pressure pmax/Poo a n d the minimum bubble radius i?min &re shown
in Table 3.1. The inclusion of liquid compressibility reduces the value of
Pmax/PooJ a n < i, the difference between the compressible and incompressible
cases increases as the pressure-ratio q(= po/poo) reduces.
The effect of liquid temperature around the bubble on the maximum
impulsive pressure and the minimum bubble radius was numerically studied
[3.32]. One of the results is shown in Figure 3.6.
It is found by the experimental investigation into the collapse of a single
spark generated bubble near a solid wall [3.26] that there are three types of
collapse modes subject to the value of L/Rmax (L is the distance between
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,
the electrodes and the solid wall, i?max is the maximum bubble radius): (a)

58 Cavitation
10s I
Pa, = 10 [.3 kPo
^.o/Ro=O.I
R0 = 1 mm
\
\
\
to'
C? Equilibrium
V
V (aH=oo)
273 293 313 333 353 373

370.27
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
T. [K]
Figure 3.6: Effect of the nonequilibrium parameter aM on the relation be

tween the maximum impulse pressure p max /Poo and the temperature in liquid
Too(Pgfi/Poo = 0.1) Tomita-Shima [3.32].

50 i i i ■ . i ■. i
Rmoxa3.5mm
Ro-101.3 kPa
'£ 10
i
i i
3 5 \\u\
i
1
Q.
i
.
i
0.5 - o T<o =293.2 K
^ =313.2
o =333.2
- o =343 2
" =353.2
0.1 , . i . . .. , . i . . ..
0. 0.5 5 10
L/Rn
Figure 3.7: EfiFect of water temperature parameter Too on t h e p m a x — L/R

max
curve; Rmax = 3.5 mm. Shima et al. [3.27].
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

60 Cavitation
Ro mm 10 1 HT1 W2 W*
Emu lOitmin Pmax £mu 10 Rmin £mu 10 Rmin Emu
■Kmin 10 Rm\n Poo
Poo Poo Poo Poo
q mm mm mm mm mm
0.10 Comp. 2.731 23 2.729 23 2.706 24 2.511 33 1.784 132
0.10 Incomp. 2.649 27 2.647 27 2.623 28 2.422 38 1.658 181
0.05 Comp. 1.712 83 1.711 83 1.698 86 1.588 113 1.209 347
0.05 Incomp. 1.609 108 1.607 108 1.594 112 1.477 153 1.063 602
0.02 Comp. 0.9228 444 0.9222 445 0.9170 456 0.8730 559 0.7364 1130
0.02 Incomp. 0.7923 843 0.7917 845 0.7858 872 0.7353 1151 0.5681 3394
0.01 Comp. 0.5980 1373 0.5978 1376 0.5953 1399 0.5747 1620 0.5179 2497
0.01 Incomp. 0.4538 4377 0.4535 4391 0.4504 4517 0.4242 5805 0.3340 15984
Table 3.1: The calculated values of the maximum impulse pressure Pmax/Poo
and the minimum radius of bubble (Poo = 101.3 kPa, Too= 273 K, 7 = 1.4)
Tomita-Shima [3.31].
For L/R max < 0.3 and > 1.5, the shock wave is dominant in the impact wall
pressure; (b) For 0.6 < L/Rmax < 0.8, the liquid jet is dominant; and, (c)
For 0.3 < L/Rmax < 0.6 and 0.8 < <L/RL/R
max
max fs
< 1-5)
1-5, both
the liquid jet contribute to the wall-impact-pressure the
the shock
[3.29].
both shock wave
wave and
and
the liquid jet contribute to the wall-impact-pressure [3.29].
3.6.2 Temperature Effect
The effect of temperature on the single-bubble collapse and the impulsive
pressure were experimentally studied [3.27] as shown in Figure 3.7. The
Pmax — L/Rmax curve changes similarly with increasing water temperatures
higher than 333 K. For L/Rmax > 1.2, p m a x maintains its inverse-drop
off relationship with increasing L/Rmax. This is presumably because of the
heat transfer through a solid wall and the decrease of shock-wave strength
at higher water temperatures.
The relationship between the maximum impulsive pressures (at both the
first and second collapses of a bubble) with L/Rmax was experimentally
obtained. The impulsive pressure (which is responsible for the plastic defor
mation of wall material) was closely related to the motion of liquid jet [3.33
and 3.34].
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

References
3.1 Benjamin, T. B., (1956) 'Pressure Waves From Collapsing Cavities',
Proc. 2nd. Symp. on Naval Hydrodynamics, Washington. D. C , 1956,
pp. 207-233.
3.2 Besant, W. H., (1859) Hydrostatics and Hydrodynamics, art. 158,
Cambridge Univ. Press, London.
3.3 Blake, F. G., Jr., (1949) The Onset of Cavitation in Liquids, I, Har
vard Acoustics Res. Lab., TM 12, Sept., 1949.
3.4 Brennen, C. E., (1995) Cavitation and Bubble Dynamics, Oxford
Univ. Press, 1995.
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
3.5 Fujikawa, S. and Akamatsu, T., (1980) 'Effects of the Non-

Equilibrium Condensation of Vapour on the Pressure Wave Produced
by the Collapse of a Bubble in a Liquid', J. Fluid Mech., Vol. 97
(1980), pp. 481-512.
3.6 Gilmore, F. R., (1952) The Growth and Collapse of a Spherical Bubble
in a Viscous Compressible Liquid, Colif. Inst. of Tech., Hydrodyn.
Lab., Rep. No. 26-4 (1952), pp. 1-40.
3.7 Herring, C , (1941) Theory of the Pulsations of the Gas Bubble Pro
duced by an Underwater Explosion, OSRD Report 236 (1941).
3.8 Jahsman, W. E., (1968) 'Collapse of a Gas-Filled Spherical Cavity',
J. Appl. Mech., Trans. ASME, Ser.E, Vol. 35 (1968), pp. 579-587.
3.9 Lauterborn, W., (1974) 'Kavitation durch Laserlicht', Acustica, Vol.
31 (1974), pp. 51-78.
3.10 Lauterborn W. and Bolle, H., (1975) 'Experimental Investigations of

Cavitation-Bubble Collapse in the Neighbourhood of a Solid Bound
ary', J. Fluid Mech., Vol. 72 (1975), pp. 391-399.
3.11 Lilliston, R. R., (1966) Calculations on the Collapse of a Spherical
Gas-Filled Cavity in a Compressible Liquid, DTMB Report No. 2223
(SP6-002), 1966.
3.12 Lord Rayleigh (1917) 'On the Pressure Developed in a Liquid during
the Collapse of a Spherical Cavity', Phil. Mag., Vol.34 (1917), pp.
94-98.

62 Cavitation
3.13 Nakajima, K. and Shima, A., (1977) 'Analysis of the Behavior of

a Bubble in a Viscous Incompressible Liquid by Finite Element
Method', Ing.-Arch., Bd. 46 (1977), S. 21-34.
3.14 Naude, C. F. and Ellis, A. T., (1961) 'On the Mechanism of Cavi
tation Damage by Nonhemispherical Cavities Collapsing in Contact
with a Solid Boundary', J. Basic Engng., Trans. ASME, Ser. D, Vol.
8 (1961), pp. 648-656.
3.15 Plesset, M.S., (1949) 'The Dynamics of Cavitation Bubbles', J. Appl.

Mech., Vol. 16, pp.227-282.
3.16 Plesset, M. S., and Mitchell. T. P, (1956) 'On the Stability of the
Spherical Shape of a Vapor Cavity in a Liquid', Quart. Appl. Math.,
Vol. 13 (1956), pp. 419-430.
3.17 Plesset, M. S. and Chapman, R. B., (1971) 'Collapse of an Initially

Spherical Vapour Cavity in the Neighbourhood of a Solid Boundary',
J. Fluid Mech., Vol. 47 (1971), pp. 283-290.
3.18 Poritsky, H., (1952) 'The Collapse or Growth of a Spherical Bubble

or Cavity in a Viscous Fluid', Proc. First U. S. Nat'I. Congr. Appl.
Mech., ASME, 1952, pp. 813-821.
3.19 Rattray, M., (1951) Perturbation Effects in Cavitation Bubble Dy

namics, Ph.D. Thesis, Calif. Inst. of Tech.
3.20 Shima, A. and Tomita, Y., (1975) 'On the Impulse Pressure Accom
panying Spherical Bubble Collapse in Liquids', Rep. Inst. High Speed
Mech., Tohoku Univ. Vol. 31 (1975), pp. 97-135.
3.21 Shima A. and Nakajima K., (1977) 'Collapse of a Non-Hemispherical

Bubble Attached to a Solid Wall', J. Fluid Mech., Vol. 80 (1977), pp.
369-391.
3.22 Shima, A. and Tomita, Y., (1979) 'The Behavior of a Spherical

Bubble in Mercury/Report 2', Rep. Inst. High Speed Mech., Tohoku
Univ., Vol. 39 (1979), pp. 19-45.
3.23 Shima, A. and Fujiwara, T., (1980) 'The Collapse of Bubbles in Com
pressible Hydraulic Oils', J. Acoust. Soc. Am., Vol. 68 (1980), pp.
1509-1515.
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

3.24 Shima, A., Takayama, K., Tomita, Y. and Miura, N., (1981) 'An Ex
perimental Study on Effects of a Solid Wall on the Method of Bubbles
and Shock Waves in Bubble Collapse', Acustica, Vol. 48 (1981), pp.
293-301.
3.25 Shima, A.and Tomita, Y., (1981) 'The Behavior of a Spherical Bub
ble Near a Solid Wall in a Compressible Liquid', Ing.-Arch., Bd. 51
(1981), S. 243-255.
3.26 Shima, A., Takayama, K., Tomita, Y. and Ohsawa, N., (1983) 'Mech
anism of Impact Pressure Generation from Spark-Generated Bubble
Collapse Near a Wall', AIAA Journal, Vol. 21 (1983), pp. 55-59.
3.27 Shima, A., Tomita, Y and Ohno, T., (1988) 'Temperature Effects on
Single Bubble Collapse and Induced Impulsive Pressure', J. Fluids
Engng., Trans. ASME, Vol. 10 (1988), pp. 194-199.
3.28 Shima, A. and Tomita, Y., (1989) 'Some Numerical Aspects of Cav-
itation Bubble Collapse', Annual Rev. Numerical Fluid Mech. and
Heat Trans. Vol. 11 (1989), pp. 198-226.
3.29 Shima, A., (1997) 'Studies on Bubble Dynamics', Shock Waves, Vol.
7 (1997), pp. 33-42.
3.30 Shu, S. S., (1952) 'Note on the Collapse of a Spherical Cavity in a

Viscous Incompressible Fluid', Proc. First U. S. Nat'l. Congr. Appl.
Mech., ASME, 1952, pp. 823-825.
3.31 Tomita, Y. and Shima, A., (1977) 'On the Behavior of a Spherical
Bubble and the Impulse Pressure in a Viscous Compressible Liquid',
Bulletin of the JSME, Vol. 20 (1977), pp. 1453-1460.
3.32 Tomita, Y. and Shima, A., (1979) 'The Effects of Heat Transfer on
the Behavior of a Bubble and the Impulse Pressure in a Viscous
Compressible Liquid', ZAMM, Bd. 59 (1979), S. 297-306.
3.33 Tomita, Y and Shima, A., (1986) 'Mechanisms of Impulsive Pressure

Generation and Damage Pit Formation by Bubble Collapse', J. Fluid
Mech., Vol. 169 (1986), pp. 535-564.
3.34 Tomita, Y. and Shima, A., (1994) 'Interactive Problems in Bubble
Dynamics', Current Topics in Acoust. Res., Vol. 1 (1994), pp. 231-
246.
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

64 Cavitation
3.35 Trilling, L., (1952) 'The Collapse and Rebound of a Gas Bubble', J.
Appl. Phys., Vol. 23 (1952), pp. 14-17.
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

Part 2: Multi-Bubbles (Stochastic
Behaviour)
S.C. Li
3.7 Origins of Bubble Stochasticity

3.7.1 Introduction
The behaviour of cavitation bubbles in real flow situations is a multibub-
ble performance possessing strong stochasticity. These cavitation-bubble
collapse-pulses are almost always superposed on other pressure-fluctuation
components (e.g. Li, Zhang and Hammitt, 1986 [3.70]), as shown by Figure
3.8. The strong randomness of cavitation-pulse events (in terms of their oc
currence and waveform) is evident, for example, from Figure 3.9 which is a
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
cavitating draft-tube flow of a Francis turbine model (Li, 1988 [3.67]).
p (22 mbar/div) p (22 mbar/div)
Mg^
. . t (6.7 ms/div) „. t (6.7 ms/div)
Figure 3.8: Cavitation pulses from a cavitating venturi flow. V —

38.4 m/s; T = 26.7° C (Li, Zhang and Hammitt, 1986 [3.70])
(a) Cavitation pulses superposed on the background noise (ie the white noise
which composes the machine noise and turbulence), a — 0.65; (b) Cavitation
pulses superposed on the cavitation associated low-frequency fluctuations
and the white noise, a = 0.74.
65
66 Cavitation
Pressure Pressure
-4-
Time Time
(a) (b)
Figure 3.9: The randomness of cavitation pulses from a cavitating draft-tube

flow of a Francis turbine model-runner (Li, 1988 [3.67])
(a) Recorded overall pressure signals (sequence) for a cavitation condition
(sampling rate 1 kHz); (b) Extracted cavitation-pulses from the sequence
showing strong randomness in their occurrence and wave-form (sampling
rate 200 kHz).
In general, cavitation bubbles behave randomly in both spatial and tem

poral domains with the following being stochastic variables:
• The size and content of nuclei, the speeds of bubble growth and col
lapse, the nonsphericity of collapse and the intensity of the impulsive
wave;
• The spatial position where a particular event (e.g. collapse) occurs for
different bubbles; and
The occurrence time of a particular event (e.g. collapse) at a given

spatial position.
The stochasticity originates from the randomness of nuclei in the liquid,

in terms of their spatial position and properties (size, content etc) and from
the cavitation mechanism by which the nuclei are cavitated; and is further
enhanced and characterised by three kinds of interactions throughout the
entire life time of bubbles. One is the bubble-boundary interaction. The
second is the interaction between bubbles. The third is between the bub
bles and the unsteady/fluctuating flow field. The following will address the
parameters contributing to the stochasticity through these interactions.
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

Bubble Dynamics: Multi-Bubbles (Stochastic Behaviour) 67
3.7.2 Bubble-Boundary Interaction

The presence of various boundaries will strongly affect the stochastic be
haviour of bubbles, i.e. its sphericity, jet and counter-jet formations, mi
crobubble generation, global migration and collapsing intensity.
For many situations, boundaries can be treated as a rigid one in which
the dominating factor over the migration and jet formation is the (bubble
to wall) relative distance 7,
S
Umax
where Rmax '■ maximum radius of bubble;
S : distance from the bubble centre to the wall when the bub
ble reaches the maximum radius Rmax-
The influence of 7 on bubble collapse can be seen from the systematic
study by Vogel et al, 1989 [3.80] (which is briefly incorporated in "Bubble-
Microbubble Interaction" of §3.7.3 "Bubble-Bubble Interaction"). For some
situations where tough and elastic material coatings (such as epoxies/epoxy
compounds and polyurethane-based plastics) are used to protect against low-
intensity cavitations (referring to §6.5.3 "Non-fused Materials" and §6.6.3
"Main Concerns in Repair"), the properties of coatings, mainly the relative
inertia m* and stiffness k*, are also dominating parameters,
m
m = p R?
where m : equivalent coating mass related to Rmax\

p : liquid density,
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
and
fc* =
(Poo -Pc)Rma
where A; spring constant of coating;

Poo pressure in the liquid at an infinite distance;
Pc saturated vapour pressure of liquid.
The study of Blake (1988 [3.44]) by evaluating Kelvin impulse over the
half-space boundary with the low-order singularity method illustrates how
the global motion and jet formation are affected by the presence of various
boundaries such as rigid boundary, free surface, two-fluid interface, inertial

68 Cavitation
boundary and membrane boundary. This concept may also be extended to a

finite number of bubbles, giving information and insight into the interaction
mechanism of cavitation bubbles with boundaries. The Kelvin impulse is
defined as
I = p I <jmds
Js
here p is fluid density; <j> is velocity potential; s is the surface of the cavitation
bubble and n is the outward normal to the fluid. The Kelvin impulse can
be caused by the presence of nearby boundary, the ambient velocity and
pressure fields.1 As an example, the case for a buoyant cavitation bubble
(i.e. the buoyancy force is included) collapsing near a rigid wall (below the
bubble) is cited. The Kelvin impulse at the completion of collapse for such
a bubble is
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
I(TR)^[2^S2B(^,\)-B(7-,l)}
where TR: life time of Rayleigh bubble,
TR = 1.83Rmax
I: Kelvin impulse associated with the equivalent of bubble's

motion normal to the wall;
'For example, in the case of rigid wall, the presence of a boundary makes the far side
of the bubble obtain a stronger acceleration during the collapsing phase than the near
side does, resulting in a global migration of the bubble towards the boundary. That is,
the bubble achieves a Kelvin impulse due to the Bjerknes Force which is created by the
pressure gradient normal to the boundary. The Kelvin impulse can be considered as linear
momentum if a virtual mass induced by the fluid motion is attributed to the cavity. Kelvin
impulse was first introduced by Benjamin & Ellis, 1966 [3.43], to analyse bubble dynamics.

6: dimensionless parameter for buoyancy force,
ftma
6- pg- Ap
with Ap being the pressure difference between the initial

pressure po in the liquid at the location of the bubble prior
to it being generated and the vapour pressure pv inside the
bubble (Ap = po - p v );
B: Beta function.
For I(TR) > 0, the bubble will migrate away from the boundary and vice
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
versa. Thus, the neutral collapse (spherical collapse without migration) line
corresponds to I{TR) = 0 yielding
7<$ = UML =- 0.442 .

N 2*(tf,i)
This result agrees well with a set of numerical experiments in terms of jet
formation and its direction (Figure 3.10). The Kelvin impulse concept can
i Yo**o**a
•01 -04 -03 -Oi -0-1 01 02 03 04 04

-»-*
-WOW) ■OUHOAHT-
Figure 3.10: The global motion of a buoyant vapour bubble near a rigid
boundary. Comparison with numerical experiments (Blake, 1988 [3.44])
well explain how the parameters such as 7 and S affect the global motion
and jet formation, but it does not provide detailed information on the bubble

70 Cavitation
surface deformation and does not give any clues to the microbubble gener
ation and the bubble-microbubble interaction since the Kelvin impulse is a
global integral value which only describes the gross characteristics of bubble
migration.
In summary, the bubble-boundary interaction increases with decreasing
7. According to the nature of boundaries, there are basically three cases:
(1) For rigid boundary (m* and k* —► 00), the interaction increases bubble
life-time. During growth phase, the boundary retards the expansion of near
side of bubble, forming a nonspherical bubble with a flattened near side;
whereas during collapse phase, it attracts the bubble resulting in a toroidal
cavity moving towards the boundary and breaking into small microbubbles.
The resultant jet velocity, impact pressure on the boundary, and the damage
potential are generally enhanced with the reduction of 7.
(2) For a free surface (which is such an opposite extreme to the rigid bound
ary that m* and k* -> 0, producing a constant-pressure boundary function
ing as a complete compliance), if buoyancy is not considered, the bubble at
a large value of 7 will move away from the surface during the collapse phase
forming a jet in the same direction of bubble migration; whereas the surface
will develop a spike moving away from the bubble. The numerical results
of bubble-wall movement, particle pathlines, and pressure contours during
growth and collapse for two cases of vapour bubble, 7 = 1.0 and 7 = 1.5,
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
are shown in Figure 3.11. The free surface influences are clearly seen: The
Bjerknes effect is always directed away from the surface with a sharper jet
for smaller values of 7; for 7 = 1.5, the surface rises and falls during bubble
growth and collapse, whereas for reduced value (7 = 1.0) the surface rises
significantly and even continues rising in the collapse phase producing a pro
nounced spike. If buoyancy is not negligible, it will reduce the spike strength
and give the bubble a component of movement rising towards the free sur
face during the collapse phase, particularly for the case of larger values of 7
which might lead to a rising movement of the bubble.
(3) For boundaries between the above two extreme cases (i.e. the compliant
boundaries which are made of composite/flexible materials with non-zero
finite values of surface parameters m* and fc*), there is an energy exchange
process involved between the fluid motions and the boundary. The boundary
might extract energy from the liquid motion which could partially dissipate
in the boundary and return the remainder back to the fluid motion. Through
such an energy exchange process, the performance of bubble growth and col
lapse is altered. The manner of this energy exchange is mainly subject to
the relative location 7 and boundary parameters m* and k*. The favourable

■1.0 I 1 -1.5
Figure 3.11: The growth and collapse of a vapour bubble in the proximity
of a free surface, 7 = 1.0 and 1.5 respectively (Blake & Gibson 1987 [3.45])
(a) Bubble shape, 7 = 1.5; (b) Bubble shape, 7 = 1.0; (c) Particle pathlines,
7 = 1.5; (d) Particle pathlines, 7 = 1.0; (e) Pressure contours, 7 = 1.0 and
T = 1.186; (f) Pressure contours, 7 = 1.0 and T = 1.310
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

72 Cavitation
effect of such a compliant boundary is to retard the bubble migration and its
jet. For certain combinations of 7, m* and k* values, the bubble might be
repelled away and the jet might disappear or even be redirected away from
the boundary, resulting in little damage potential to the boundary. These
concepts are based on recent studies (Shima et al 1989 [3.77] and Zhang et
al 1993 [3.82]) and are shown in Figure 3.12. The relation of relative bubble
s
Rmax • =1.14 , filled: experimental \
3 ■ =1.43 open: numerical j
A =1.71 v
Neutral collapse line
1
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
0
0 1 2 3 4 5
Figure 3.12: The effect of boundary properties on the single-static bubble

collapse
Filled symbols: Experimental results, Shima et al 1989 [3.77]
Open symbols: Numerical results, Zhang et al 1993 [3.82]
distance at the final stage of collapse, RSc (Sc is the bubble distance to
the boundary at the final stage of collapse), with m* (ranging from 0 to 00)
for three 7 values (1.14, 1.43 and 1.71) reveals that collapses occur in either
a repelling zone (RZ) or an attracting zone (AZ), separated by a neutral
collapsing line (dotted line) on which no bubble-migration takes place dur
ing collapse process. This neutral collapsing line passes through the points
where Sc/Rmax = 7-

3.7.3 Bubble-Bubble Interaction

There are mainly two types of bubble-bubble interactions taking place
through bubble/shock-wave action. One occurs in the collapse process of a
bubble itself, i.e. the shock wave emitted from the main collapse strikes the
microbubbles (generated by the main collapse) and accelerates the rate of
microbubble collapses, resulting in much higher collapsing peak pressures.
This interaction takes place in a small region and particularly occurs for the
bubble which collapses near a boundary since microbubbles are mainly pro
duced by the instabilities developed during the nonspherical main-collapse
in the presence of a nearby boundary. The other is the interaction between
the bubbles originally growing from different nuclei and interacting within
a larger spatial distance compared with the bubble-microbubble interaction.
Therefore, the bubble-microbubble interaction is often referred to as small-
scale interaction and the other is referred as large-scale interaction.
Bubble-Microbubble (Small-Scale) Interaction

This interaction can be explained by a laser-induced bubble collapsing near
a solid boundary (Vogel et al, 1989 [3.80]). The collapse appearance, i.e.
the bubble elongation 2 , the formations of the main jet and its counter-jet,
the ring vortex, the microbubble generation and the bubble-microbubble
interaction, strongly depend on the geometric configuration particularly the
relative distance 7.
At very large 7, the bottom flow is not inhibited by the boundary and
the elongation does not develop, resulting in a spherical collapse. A counter-
jet may appear at moderate 7-values but only for 7 > 1 when the lower
bubble wall with a high curvature is not attached to the boundary (Figure
3.13(b)). The counter-jet is always weaker than the (main) jet because of the
inhibited bottom flow. The occurrence of a counter-jet and the elongation e
are functions of 7. The strength of the counter-jet increases with decreasing
7 (as long as 7 > 1) and increasing e. These relations are summarised in
Figure 3.14.
For relatively large 7 values (ranged from 1.9 — 2.55), a ring vortex
forms after the first collapse. After taking a toroidal shape, the bubble
disintegrates into various parts which collapse separately (but nearly at the
2
The mechanism of elongation leading to jet/counter-jet formations is detailed by
Lauterborn, 1982 [3.66]. The elongation is denned as e = j ^ — , here Rhor is the hor
izontal bubble radius at the instance of maximum bubble elongation; h is the distance
between the level of Rhor and the top of bubble.
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

74 Cavitation
(a)
(b)
(c)
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
(d)
Figure 3.13: Jet and counter-jet formation at different dimensionless dis

tances 7. Rmax is the maximum bubble radius and e is the bubble elongation
before collapse. The series were taken at 20,000 frames/s (Vogel et al 1989
[3.80])
(a) 7 = 2.30, e = 1.04, Rmax = 2.0 mm; (b) 7 = 1.56, e = 1.13, i W =
3.2 mm; (c) 7 = 0.75, e = 1.41, Rmax = 1.6 mm; (d) 7 = 0.34, e =
1.38, i?max = 2.3 mm

e
Counterjet:
1.40 _ 0
o No
• Week
1.35 A Pronounced
1.30 - A Strong
1.25 - o
A
1.20
1.15 - o
1.10 o •
1.05 -
1.00 1 o ia o io
0.5 1.0 1.5 2.0 2.5 3.0 y
Figure 3.14: Formation and strength of counter-jet as a function of 7 and e

(Vogel et al 1989 [3.80])
same time) and then coalesce again during rebound. It migrates toward
the solid boundary and starts to expand as soon as it reaches the boundary.
After the second collapse, the vortex core takes a polygonal shape due to the
instability waves on the vortex. During first and second bubble collapse, the
migration is accelerated (Figure 3.15) owing to the conservation of Kelvin
impulse.
The second collapse of the bubble with decreasing 7 values is different
from those with 7 values higher than 2. In Figure 3.16 (7 = 1.66), the
bubble touches the boundary during rebound after first collapse so that a
ring vortex can only exist for a very short period during second collapse
when the bubble lower side is again detached from the boundary (the third
and fourth frames in Figure 3.16). The generation of microbubbles from the
first collapse is caused by the the bubble-wall instabilities (nonsphericity)
which grows during collapse and finally breaks up the bubble. The schlieren
technique makes these microbubbles visible3 in Figure 3.16. The second
collapse strikes these microbubbles which collapse quickly and vanish (frame
5 of Figure 3.16).
With further decreasing 7 (7 = 1-37), no ring vortex forms after the first
3
Sincc the main bubble blocks out the light and the cloud of microbubbles (which
surround the cavity) appear bright due to their high scattering efficiency.
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

76 Cavitation
1 First collapse
| Second collapse
First collapse
y =2-55 t V ^| x
Rmax =2.5 mm Second collapse"
y =2.15
° Rmax =2.05 mm
1.2 t(ms)
Figure 3.15: The migration of a bubble collapsing near a rigid boundary. I is

the distance between the bubble centre and the boundary (Vogel et al 1989
[3.80])
f*% 'VM
Figure 3.16: The second collapse of a bubble, 7 = 1.66, Umax — 3.3 mm.
Taken with spatial filtering at 20,000 frames/s. The location of the solid
boundary is marked with the white horizontal bar in the first frame (Vogel
et al 1989 [3.80])

--``,```,,``````,
collapse. A radial outward flow on the solid boundary arises from the jet,
decelerating the second collapse in the horizontal direction as shown in Fig
ure 3.17. The toroidal bubble disintegrates into several separately collapsing
M:-V
' f^' 1*i
\
-■*•> r ^ ^ \ / * * ^ ^ >**■
KA
Figure 3.17: The first and the second collapses of a bubble, 7 = 1.37, Rmax =
4.0 mm. Taken with spatial filtering at 20,000 frames/s. The location of
the solid boundary is marked with the white horizontal bar in the first frame
(Vogel et al 1989 [3.80])
parts which can be observed in frame 15 where two acoustic transients are
emitted from different sides during second collapse. The microbubbles ap
pear on the upper side of the bubble as a cloud, and the protrusion of the
upper side indicates that a counter-jet is formed from the first collapse.
For the case where 7 is about unity, the bubble dynamics are of interest.
During the first collapse phase, a ring vortex forms before the final stage
of collapse and the microbubbles generated from the first collapse attached
on the boundary wall appear as a white dotted line in Figure 3.18. Figure
3.18(a) shows a bubble at the stage of maximum expansion after the first
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
collapse; (b) is at an intermediate stage; and (c) is at about 30 fisec before

final collapse. The dimensionless time T is defined as

78 Cavitation
(a) (b) (c)
Figure 3.18: The Schlieren series of collapse. 7 = 0.96, Rmax — 3.9 mm

(Vogel et al 1989 [3.80]).
(a) T = 1.03; (b) T = 1.80; (c) T = 2.19
where Tc: Rayleigh's collapse time, Tc = 0.5TR = 0.91-ftmax %/p/Poo\

t: time starting from the instance of bubble generation.
As the vortex ring forms before the completion of the first collapse, the fluid
motion directed radially towards the bubble centre during the collapse is
thus partially transformed into the rotational movement of the ring vortex;
consequently the bubble implosion is decelerated; and the bubble remains
relatively larger during collapse compared with the case for 7 > 1. Therefore,
the pressure rise inside the bubble is quite low for 7 being about unity,
resulting in a weak sound emission.
When 7 < 1, there is no counter-jet formation and the main jet becomes
thicker and hits the lower bubble wall at earlier stage of collapse as shown
in Figure 3.13(c). When 7 is further reduced, for say 7 = 0.34 as shown in
Figure 3.13(d), the jet diameter diminishes again. In the range of 0 < 7 <
1, the microbubbles are produced by the jet flow during bubble collapse,
referring to Figure 3.13(c) and (d). In this range of 7 values, with the
reduction of 7 value, the intensity of bubble collapse increases and achieves
the highest pressure amplitudes for solid boundary at 7 -> 0.
From the above observations, it can be concluded that the main bubble
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

Bubble Dynamics: Multi-Bubbles (Stochastic Behaxnour) 79
collapse and the microbubble generation are mainly dominated by the pa
rameter 7; and the (main) bubble-microbubble interaction plays a significant
role in bubble collapse and its damage to boundary materials. Only if the
bubble-microbubble interaction occurs will the cavitation bubble produce
enough power to damage boundary materials. Otherwise (i.e. if the dam
age capacity from the main jet were only accounted for), cavitation bubble
collapses would not cause any damage, not even a tiny pit, to the very soft
99% pure aluminium 4 .
B u b b l e - B u b b l e (Large-Scale) Interaction
Due to the difficulties in acquiring the image of bubble-bubble interaction

in real flows, various experimental and numerical techniques designed to
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
investigate bubble-bubble interactions by simulating multibubbles collapses
in static fluid field instead of in real cavitation flows.
Two effects of bubble-bubble interactions under the impact of shock
waves 5 are revealed by employing 2D water/gelatine technique (Dear &
Field, 1988 [3.53]). One is the Divergence Effect on jet direction when a
row of cavities are struck by a shock wave simultaneously, even when they
are struck at slightly different times, as shown by Figure 3.19. The jets
in the cavities are well-formed by the strike of shock wave, with the jet-
4
For example, the jet velocity from a collapsing bubble at 7 = 1.0 in water under
atmospheric pressure is about 100 m/sec or less, which is only capable of producing water
hammer pressure of 4.5 kbar. This is in the same range as the pressure of 2.5 kbar achieved
inside a collapsed cavity attached to the boundary. Both values are merely close to the
(static) Brinell hardness of 99% pure aluminium (3 kbar), i.e. far below its dynamic
hardness (13 kbar). Hence such collapses are not capable of eroding soft aluminium.
If, however, we consider the interaction of bubble-microbubbles we shall see that the
collapse of the microbubbles has very considerable damage capacity. For instance, let us
analyse such a case that the microbubbles are generated by the impact of the outward-
flowing jet with the contracting cavity surface (Tomita & Shima, 1986 [3.78]). At small
7-values, referring to Figure 3.13(c) (7 = 0.75) and 3.13(d) (7 = 0.34), the cavitation
bubble reaches its minimum size about 50 psec after the jet hits the boundary. That
is, the microbubbles produced by the jet flow shortly after their generation are struck
by a pressure pulse of 2.5 kbar arising from the main cavity collapse (which is much
higher than the ambient driving pressure), causing a much more violent collapse of these
microbubbles and producing much higher pressures locally at the solid boundary well
beyond the dynamic hardness of aluminium or other materials. The ring-shaped damage
pattern located at the periphery of a collapsing bubble commonly observed in damage
tests is the evidence of this type bubble-microbubble interaction.
5
The shocks are produced by a striker and visualised with schlieren optics. These shock
waves, in reality, are the consequence of collapses of neighbouring bubbles.

80 Cavitation
Figure 3.19: The divergence effect on the jet direction. Three cavities with
3 mm of diameter, 6 mm apart, parallel to the shock wave S (=0.26 GPa).
The jet velocities are 400 m/s approximately. Interframe time is 0.96 fis
(Dear & Field 1988 [3.53])
directions in the outer cavities forced away from the centre one. This jet
divergence effect is a general feature of this sort of bubble mutual influence
and becomes stronger with closer cavities, even forming two jets in each cav
ity due to cavity-wall irregularity and shock-front perturbation. The second
is the Chain Reaction, i.e. the collapse and rebound of one cavity (or one
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
row of cavities) generates strong radial shock wave which then drives the
next one (or row) to collapse and so on. Figure 3.20 shows such a chain
reaction in which the first cavity is collapsed by the shock wave S, forming
a jet as marked with J in frame 2. A rebound shock wave S' is formed
as shown in frame 3. The initial shock wave S has little influence on the

Figure 3.20: T h e chain reaction. T h r e e cavities w i t h 3 mm of diameter,

perpendicular to the shock wave S. Interframe time is 4.25 /xs (Dear &
Field 1988 [3.53])
second cavity which is shielded by the first cavity, only causing a slightly
lateral compression (frame 3). Whereas the rebound shock of t h e first cavity
causes t h e second cavity collapsing and forming a jet (frames 4 a n d 5). T h e
t h i r d cavity collapses in a similar way by the collapse a n d r e b o u n d of second
one. T h u s , a chain reaction along a line of cavities is possible subject to t h e
shock strength, cavity diameter a n d spacing. T h e speed of chain reaction
(which differs from the shock-wave velocity) depends on t h e cavity size a n d
j e t velocity.
According to Lesser's model 6 , the jet velocity caused by shock wave
6
The reflection of the shock wave S acting upon on the cavity wall produces a corner
wave C and a reflected tensile wave R, referring to Fig. 3.21. The velocity, V/, imparted
to the free surface is V/ = 2V sin0 (here, V is the particle velocity behind the shock wave).
With a linear assumption, p = pcV (here, c is the speed of sound in water, c = 1500 m/sec),
the resulting shapes of the cavity wall can be obtained for various nondimensional times t
(t = tc/R, here R is the radius of cavity). Figure 3.21 is for V = 150 m/sec and t = 1 — 5.
The maximum velocity of the cavity wall , i.e. the jet velocity Vjet, is achieved at the
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

82 Cavitation
(a) (b)
Figure 3.21: Cavity-shock wave action (Dear & Field 1988 [3.53])
(a) A plane shock wave S acts upon a circular cavity, producing a corner
wave C and a reflected wave R; (b) Resulting shapes of cavity wall for non-
dimensional times i = 1 — 5
impacting on the bubble is much higher than the jet velocity caused by
nonspherical collapse of a bubble near a solid boundary wall 7 . Therefore,
the shock-wave/bubble action plays an amplifying role in producing a high-
velocity jet which causes very high "water hammer" pressure capable of
damaging materials exposed to it or driving neighbouring bubble to collapse
much more violently even forming a chain reaction. Experimental evidence
shows that a high fraction of collapse energy, i.e. 80 — 90% estimated from
the ratio of jet velocity squared, of one collapsing bubble is transmitted to
the next. So, this is why the chain reaction can remain and even strengthen
the subsequent bubble collapsing.
centre of cavity, Vjct = 2V. If the effect of jet tip convergence and nonlinear shock wave
relation are concerned, Vjct « 3V.
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
7
For example, the jet velocity will be approximately 400 m/sec for the shock wave-
bubble action case (bubble diameter 3 mm, shock wave strength 2.6 kPa) which is much
higher than the jet velocity of 100 m/sec produced by the nonspherical collapse of bubble
itself. Some experiment even shows that if the shock wave of 10 kPa produced from the
collapse of vortex cavitation bubble hits a bubble near a solid wall, it could produce a very
high jet velocity of 1000 m/sec and "water hammer" pressure of 20 kbar

For the case of two bubbles interacting through the shock waves emit
ted from their own collapses, their behaviour is affected by bubble size and
sphericity, spacing, and collapse timing. With a linear approximation ap
proach, Fujikawa, 1986 [3.58], numerically studied two initially spherical bub
bles with different radii. 8 Their interaction is shown in Figure 3.22, in which
the bubble radii are normalised by their own initial values i?ioo(= 1 mm) and
#200 (= 0.5 mm) with M = jf00- = 0.5; the time is normalised by Rayleigh
collapse time, T = # i o o ( ^ ) 0 5 = 99.3 fisec. In Figure 3.22(a), the two ve
locities of left (u/ = 0°) and right (a/ = 180°) walls of bubble 2, the relative
velocity of ambient liquid to bubble 2, the actual time and the normalised
time (in brackets) are indicated in each frame. The initial gas pressure in
both bubbles is Pi g 0 = P2go = Pgo = 7091 kPa equivalent to 0.07 P^. The
saturated vapour pressure is Pv = 2.337 kPa. The initial distance between
two bubbles is 5 mm. The interaction, referring to Figure 3.22, is featured as
follows. The larger bubble (here, bubble 1) behaves almost spherically as if
there were no neighbouring bubble. Whereas, the smaller bubble is affected
dramatically. In the phase of first collapse, although its wall shape still re
mains spherical, the whole bubble is attracted by the larger one, causing a
translatory movement towards left. Then in the phase of first rebound, the
smaller bubble becomes a slightly flat due to the influence from larger bub
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
ble. At the late stage of second collapse, bubble 2 is elongated parallel to the
relative flow direction and, at second rebound stage, the left part 9 of wall (at
(j = 0°) becomes abruptly unstable forming a protrusion directed to bubble
1 with the speed exceeding 120 m/sec, referring to the spike of R (u/ = 0°)
in Figure 3.22(c). The right side (u/ = 180°) of the bubble wall is stable
and remains basically spherical in shape. Both parts of wall do not rebound
simultaneously which makes the pressure waves radiated from different parts
of the bubble wall non-spherical. The non-condensible gas content affects
the bubble-wall instability. When the value of gas pressure Pgo is reduced
to 5.065 kPa, much larger instabilities appear at the second collapse and
rebound stages forming a faster protrusion at w' = 0° with the velocity ex
ceeding 750 m/sec. Further reducing Pgo to 1.013 kPa even causes the early
8
He assumed two initially spherical cavitation bubbles in irrotational unbounded liquid
(water) undergoing a collapse process under the step rising of water pressure at infinity. His
model included liquid compressibility, gas content in the bubbles and bubble translatory
motion. For details see [3.57, 3.58 and 3.59].
9
I t is at the downstream side of the relative flow because at this instance bubble 2
is moving away from bubble 1 at a speed much faster than the mean velocity of the
surrounding fluid causing the relative flow velocity towards bubble 1.

84 Cavitation
Pgo=0 07 M=0.5 0=10

-20.50 -21.65
R2i(i)'=0>=-0.06m/s R2((i>'-:l80)=-0.06m/s -2.42
Vn=Om/i 126.90(1. p 8 )
1=0.00 1U (1/ 1=0.00) I I I
0.0
' _J 1_
10 2.0 3.0 -23.94 -26.82
-4.52
-23.76 -22.75
128.53(1.294)
-U. I Ll
-a
3.10
62.74(0.531)
-22.50 -28.52
Ji I )l I I L
-9.73
129.84(1.308)
J LL
7.83
545
7.38 0.0 -Or
, 55.20(0.556) _^ -7.16 -15.85
-16.79
-U—i—U 1——i 1—Qh—L 130.63(1.316)
_U I Li I I I rh I
4.02 4.76 0.0
< % ■
0.15
19.02 2.89
82.93(0.835) -20.45
I I I H 1 L 131.10(1.320)
I I I 1
-1.16
0.09
-0.41
0.0 ' ' <a ■
75.80 22.14
100.19(1.009) -17.62
j—£Dj—I 1 L 131.74(1.327)
_U I LJ I I I ^ti L
0.0 ^
-10.61 -10.51
-0.60 123.20 26.02
-9.90
118.75(1.196)
I I I Lj _L I IZ L If I II "I"1*'?5' , ^ ■
0.0 / \ 5.0/ 0.0
(a)
Velocity m/s
120
100
•
R2( (0
A
^
80
-
^J
L1
60
40
R2 • (0
20 (\ <
20
0
"" \l 1 ^ T
"Zf\r7
-Vr /
40
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 16 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6
(b) Time 1/T (c) Time 1/x
Figure 3.22: Dynamic behaviour of two initially spherical bubbles (Fujikawa

1988 [3.59])
(a) Bubble shape and position; (b) Bubble radius; (c) Wall velocity of bubble
2 and the relative flow velocity (to bubble 2). The relative flow velocity, vr,
is defined as positive if the direction of the relative movement of liquid to
bubble 2 is toward the right
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

appearance of the instabilities (at the stages of first collapse and rebound)
which occur at the right part of wall (w' = 180°) forming a inrushing shape
(i.e. jet), instead of a protrusion shape on the left (a/ = 0°), with the jet
velocity above 500 m/sec. The parameter M affects the dynamics of interac
tion as well. When M reduces to 0.3 (R\oo = 1 mm, i?200 — 0-3 mm), both
bubbles behave more or less like that for M = 0.5 but the instabilities occur
at the right part (u/ = 180°), instead of at the left part (u/ = 0°), of wall
in the phase of second collapse and rebound. The right part of wall firstly
starts to flatten out at the downstream side of the relative flow (a/ = 180°)
in the late phase of second collapse it then indents and eventually forms a
crater in subsequent rebound. The study also shows that once the insta
bilities occurs on the wall it will be further amplified by the interactions.
m
The reciprocal of relative spacing, i.e. i—, is demonstrated to be
a dominant parameter in a similar study by Tomita & Shima, 1986 [3.79].
The interaction increases with the reduction of relative spacing. The wave
form (particularly the rise time) of the pressure wave significantly affects the
bubble behaviour. For the impact pulses with finite energy, e.g. triangular
and exponential waves, there is a bubble size that makes the mostly violent
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
collapsing, which indicates the response time of bubble compatible with the
acting durations of pulse.
For the interaction with a multibubble configuration in a 3-D spatial
domain (which is more close to reality), Chahine's work employing an ana
lytical technique (asymptotic expansions) and a numerical simulation tech
nique (3-D boundary element method - BEM) seems to be the first work on
this subject 10 . However, due to the assumptions of incompressible liquid,
the interactions investigated (e.g. Chahine et al 1992 [3.51]) take place in
deed through the variations of hydro static pressure rather than shock-wave
action. Chihane adopted the following nondimensionalisation,
Poo(0) -Pv.
10
The asymptotic expansion approach requires less computation and provides fairly
precise solutions to the problem in which the characteristic size rto of bubbles (taken as
initial bubble size) is smaller compared with the characteristic inter-bubble distance lo
(taken as initially minimum distance between any two bubbles), which is referred as weak
interactions. Whereas the BEM model developed by Chahine et al is capable of simulating
strong interactions.

86 Cavitation
w = rb0Ap
v = wrbo
(AP) '
where e: measure of void fraction with rjo being the charac
teristic size of the bubbles and /o being the charac
teristic inter-bubble distance;
7: cavitation number;
W: Weber number;
v: ratio of driving force frequency OJ to natural fre
quency of a bubble with radius r&o;
Ap: characteristic pressure variation associated with
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
driving force poo-
The BEM calculation 11 was performed in an infinite and irrotational liquid

domain with the driving pressure field being a step-wise drop at time t = 0.
Two values of initial pressure drop were employed: 7 = 1.004 (large drop)
and 7 = 2.508 (mild drop). The number of bubbles varied from 2 through
8 with varying e. The interaction revealed can be summarised as follows.
Firstly, for symmetric bubble configurations (bubbles being arranged at the
edges of a line/square/cube), the bubble growth is inhibited; whereas the
collapse is enhanced because the interaction reduces the driving-pressure
drop (for growth) as a result of neighbouring bubbles' growth and increases
the driving-pressure (for collapse) as a result of neighbouring bubbles' col
lapses. Owing to this effect, each bubble ends its collapse under a driving
pressure which is an order of magnitude higher than that for the isolated
bubble case. The more bubbles are involved, the higher the resulting pres
sure (Figure 3.23). The time period of bubble oscillation increases with the
number N of interacting bubbles. The jet advancement toward the cloud
centre increases with N as well. These features are more pronounced as
the value of e is increased (Figure 3.24). The pressure at the bubble-cloud
centre is nondimensionalised with respect to the maximum pressure that
would have been induced by a " Rayleigh-Plesset bubble" at a distance ^ .
" T h e calculation assumed: The bubbles contain both vapour and noncondensible gas;
the pressure inside the bubble is the sum of gas pressure pg and vapour pressure p„; p„ is
constant equivalent to equilibrium vapour pressure; the amount of noncondensible gas is
constant; and, a polytropic relation, PgV = const, for the gas. For details see [3.48 and
3.51].

Normalised pressure
20
IU
0 1 2 3 4 5
Normalised time
Figure 3.23: The influence of number N of the interacting bubbles on the

pressure variation at the cloud centre (dotted lines indicate the asymptotic
results), e = 0.047, T = 2.508, W = 6.7 x 105 (Chahine & Duraiswami 1992
[3.51])
Normalised pressure
30
20
10
•10
Normalised time
Figure 3.24: The influence of void fraction e on the pressure variation at

the cloud centre (dotted lines indicate the asymptotic results). N = 4, 7 =
2.508, W = 6.7 x 105 (Chahine & Duraiswami 1992 [3.51])
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

88 Cavitation
Secondly, there is a Phasing Effect when an initial phase difference is im

posed on the bubbles arranged in such a symmetrical way that they would
otherwise behave identically. Figure 3.25 shows a four-bubble case with the
Figure 3.25: Phasing effect of bubble interaction, 4 bubbles initially sym

metrically distributed but with different initial radii (Chahine & Duraiswami
1992 [3.51])
bubbles being centred on the corners of a square with such a time delay
being imposed that at t = 0 the bubbles have relative initial size in the ratio
of 2, 1, 3, 1 counterclockwise starting from the bubble at the origin. The
time periods for the bubbles with larger size at t — 0 are increased whereas
for the "delayed" bubbles (i.e. smaller ones at t = 0) are prevented from
growing very much by the neighbouring bubbles resulting in very early col
lapse. These earlier collapsing bubbles, on the other hand, produce large
pressure rises which make the larger bubbles collapse very strongly. Thirdly,
the growth of the outer bubbles in a cloud shields the inner bubbles from
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
behaving as the outer bubbles do, i.e. Screen or Shielding Effect. Figure 3.26
is a typical case of five bubbles. Initially, the inner bubble grows similarly to
the outer ones but later remains unchanged for a certain time period. Nev
ertheless the outer bubbles have proceeded to the violent collapse stage with
the development of re-entrant jet toward the central bubble. The central
(inner) bubble seems to be shielded by the outer ones, with its time period
being approximately doubled. The study on a 16-bubble cloud also shows a
similar screen effect. Actually, the time delay between the outer-layer and

--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
Figure 3.26: Screen/shielding effect of bubble interaction, 5 bubble configu
ration, e = 0.474 (Chahine & Duraiswami 1992 [3.51])
inner-layer bubbles is related to the nature of pressure-wave propagation in a

bubbly cloud that distinguishes such a cloud from the Compliance Model12.
Apart from behaving as a passive compliance, the bubbly cloud can func
tion actively as a (pressure fluctuation) exciter with its own characteristic
frequencies 13 and stimulate huge low-frequency pressure fluctuations in hy
draulic systems through the double-oscillator (one is the cloud, the other
is the liquid-phase in the system) mechanism. Such a phenomenon was
observed in hydraulic systems (Li, Zhang and Hammitt, 1983 [3.69], 1986
[3.70] and Li, 1992 [3.68]), and is called Cavitation Resonance14. A similar
12
The compliance model is currently used for analysing pressure oscillations of cavitating
hydraulic systems. In this model the cavitation cloud is simply represented by a lumped
parameter, capacitance C, C = —jj9 a" v (here VCOI> is the volume of the cavitation
cloud; a is the cavitation number defined as usual; and H is the pressure head) which
implies that all bubbles oscillate uniformly in compliance with the surrounding pressure
variations.
13
That is, the cloud itself is a complete oscillating system rather than a simply lumped-
capacitive element in an oscillating system. Apart from compliance, it possesses, at least,
the inertial-mass, elasticity and viscous resistance. And, for some cases, these parameters
have rather discrete nature than lumped parameters, subject to the geometric shape and
properties of the cloud.
14
For cavitation resonance, the cavitation cloud and the liquid phase portion in the

90 Cavitation
collective phenomenon of gas bubbles involved in oceanic noise from wave

breaking on the shore was studied 15 by Prosperetti in 1993 [3.75].
For the collective dynamics 16 of bubble-bubble interaction in a bubbly
cloud, the study on the vaporous bubbles collapsing in the forms of hemi
spherical and cylindrical clusters (Hansson & M0ch, 1980 [3.61]) explored
important characteristics of multibubble interaction in such bubble clouds
(or clusters):
• The collapse of a bubble cluster, being driven by the ambient pressure
wave, proceeds from the outer boundary of a cluster towards its centre;
• The hydrostatic pressure at the inward-moving boundary increases

continuously and reaches a significantly higher value at the cluster
centre than the ambient pressure; and therefore,
• The individual bubbles involved in the inward-moving collapse proce
dure collapse under an increasingly high driving pressure, resulting in
increasingly violent collapse.
The hemispherical model [3.61] is cited to give an insight into this inter
action phenomenon which is often observed in practices. For example, the
periodic appearance and collapse of cavitation cloud on the ship propeller
and turbine/pump blades. This model (Figure 3.27) is based on the energy
transfer from the collapsing bubbles to the neighbouring bubbles which col
lapse subsequently. The collapse of the cluster is initiated at its boundary
by the hydrostatic pressure in the ambient liquid, then the collapse proceeds
towards the centre of the cluster. The collapse of the outer bubbles creates a
field of increased pressure around the remaining part of the cluster, and the
inward radiating energy from the collapsing bubbles is thus being transfered
into the collapsing energy of the inner bubbles. Consequently, the inten
sity of individual bubble collapse increases towards the centre of the cluster
where under the pressure rise far above the ambient pressure bubbles col
lapse most violently producing a very high jet velocity and damage potential
system are treated as two separate oscillators, the characteristic frequencies of the cloud
(which are related to the properties of cavitation cloud) are varying with the cavitation
condition, and under certain conditions the characteristic frequencies of both oscillators
(i.e. the cloud and the liquid phase portion) coincide producing huge pressure oscillations
in the whole system.
" T h a t is, a low-frequency (compared with the natural frequencies of individual bubbles)
noise component is always observed which is believed to be generated by the bubble-cloud
oscillating at its characteristic frequency.
16
In a macroscale, i.e. the average bubble spacing Al is taken as reference length.
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

vr(r)
Figure 3.27: Hemispherical model of bubble-cluster collapse (Hansson &

M0ch 1980 [3.61])
to the solid boundary wall. In this model, a homogeneous distribution of

vaporous and spherical (initial radius ao) bubbles is assumed, and the thick
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
ness of the shock front at the cluster boundary is ignored. Therefore, this
model becomes invalid when the cluster radius reduces to the mean bubble
spacing AZ,
where ag: bubble radius;

/?: volume fraction of cluster.
The cluster is assumed to be initially with radius RQ in equilibrium in a

semi-infinite space under the ambient pressure p\ equal to vapour pressure.
Then at t = 0, the ambient pressure step-wise rises to a constant value of
Poo (Poo ^ Pi)- Thus a pressure wave with the character of a shock wave in
a compressible liquid propagates into the cluster at the Mach number M\
(based on the sound speed, Cm, within cluster),
here subscripts 1 and 2 refer to the conditions in front of and behind the
shock front respectively; capital letter is designated for conditions at the

92 Cavitation
cluster boundary. The thickness of the shock front, which is of the order of
ao
[/3(1-Mf 2 )]s '
is neglected. If the value of /? is neither close to zero nor to unity, the
value of Cm, i.e. the sound speed within the bubbly mixture domain, can be
approximated as
r2 Pl
m
PP(1-P)
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
here p is liquid density. When the bubbles are inwardly annihilated by the
shock wave, the cluster boundary moves toward the centre at the velocity
Pi
Vsh = CmMi = R = - (3.19)
00(1 - 0)
The velocity at radius r in the liquid domain is radial, vT. Using the relation
Vr = (3R (Vr is liquid particle velocity at boundary R), the velocity vr can
be related to the inward-moving velocity R of the cluster boundary as
vr = Vr\=0R^-.
From the equation of motion

dvr dvT _ 1 dp
dt dr pdr'
the pressure distribution in the liquid domain can be derived as
, ,, / J1RR +R R „oR R
P(r,t)=Poo+P [P /?2^r
Finally, the collapse equation of the cluster can be obtained with p — Pi at
r = R,
RR+(I + ^P)R2 = - ' ^ (3.20)
or in nondimensional form
R*R* + (l + i / ? ) R*2 = - 1 (3.21)
here R* — ^ and t* — ( ^ ) ( ^ ) 5 • The pressure P2 at the cluster bound

ary which drives the boundary bubbles collapsing being a function of R

P 2 =P2 /Poo
R* = R / R 0
150
100
50 - \
"- 1 — — 1
05 1.0
R*=R/Ro t _ ( \
Figure 3.28: The variations of pressure P2 and cluster radius R during the
collapse with /3 « 1 (Hansson & M0ch 1980 [3.61])
can be determined from eqn (3.20). The numerical results for both the
hemispherical- and cylindrical clusters show: the spatially highest pressure
in the liquid domain is the pressure at the boundary of collapsing clus
ter, i.e. the pressure driving the boundary bubbles collapsing which is
Pr,t |r-»ft= p{R>t) = A(*); a nd, the temporally highest driving pressure
Pi{t) will be achieved when R —> 0, i.e. the driving pressure reaches the
highest level (infinity) at the cluster centre when t* —> 1, forming a focusing
collapse when the inward velocity of cluster boundary also reaches infinity
as shown in Figure 3.28. The total collapse time T of cluster is shown to be
proportional to the value of ( ^ - ) 5 •
It should be noticed that if the cluster shape differs from such focusing
configurations, i.e. hemispherical and cylindrical forms, the inward-moving
increasing pressure will be defocused resulting in less pressure increase at the
end of the cluster collapse. Besides, the bubble distributions (e.g. size- and
spatial distributions etc.) within the cluster are also important parameters
influencing the energy transfer and shock-wave action.
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

94 Cavitation
3.7.4 Bubble-Flow Field Interaction

Cavitation in nature is the interaction result of bubbles with flow field. In
classic theory, cavitation is treated as a static problem related only to the
mean minimum pressure coefficient, C Pim i„, of the flow field and the static
critical radius of the nucleus. Actually, cavitation is a Lagrange problem
rather than an Euler problem, almost always associated with turbulent flows
in which the instant pressure varies significantly and randomly. Therefore,
cavitation (its inception, bubble growth, collapse and rebound) is a rather
stochastic problem. Its randomness is strongly related to the (temporal and
spatial) spectra of flow field. The stochasticity of cavitation bubbles is thus
modified and characterised by the bubble/flow-field interaction.
The information of bubble/flow-field interaction was obtained through:
(1) Single/Individual Bubble Study, which shows the influences of veloc
ity/velocity gradient and pressure gradient on the single/individual bubble
(particularly its collapse); (2) Statistic Study, which shows the statistic re
lation of cavitation (particularly the inception) to the pressure fluctuation
field of flow either in micro (Kolmogorov) scale or in global (coherent) scale.
Single/Individual Bubble Study

The early studies on single bubbles show that the main flow parameters
affecting bubble behaviour are velocity/velocity gradient and pressure gra
dient. For example, the first work on flow influence on bubble elongation
and collapse by Kling & Hammitt, 1972 [3.64]; and the review on boundary
influences by Blake & Gibson, 1987 [3.45]. The main points revealed by
these studies are:
• The flow velocity will elongate the bubble in the flow direction, and
if the elongation is parallel to the boundary nearby, the jet velocity
towards the boundary will increase remarkably. For example [3.64],
a elongation of 15% would be achieved by the flow of 25 m/sec in
a venturi for Umax — 2 mm and 7 ( = Rs ) = 1.14. Its resultant
jet velocity (under 2 bar ambient pressure) was estimated [3.80] as
300 m/sec, which was twice as high as that for a similar bubble in
a static field. This elongation is different in nature from the vertical
elongation caused by the nearby solid boundary's retardation to bubble
collapse.
• The pressure gradient will cause bubble flattening. In some cases, the
flattening will form dimpling or even a liquid jet penetrating into and
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

through the cavity, with the flattened part of bubble on the side of
highest pressure and parallel with the constant-pressure surface in the
flow field [3.65].
However, these early studies were restrained by the simplest flow assump
tions of either static fluid or uniform flow and are not adequate to understand
the situation in real flows.
The advance in the BEM technique facilitates the study of multibubble
dynamics with flow assumptions closer to real flow environment, such as
the bubble interactions with the line vortex flow [3.50] and the boundary
shear flow [3.49]. In Chahine's work, the real flow field 7 (incorporating
the performing bubbles) is decomposed into a basic flow field 3"o in absence
of bubble and a bubble flow 3~b owing to bubble dynamics. The basic flow
GFo and real flow 9" both are incompressible and satisfy the Navier Stokes
equations. The 3& is restricted to incompressible and inviscid flow. There
fore, the bubble flow is a potential flow with velocity potential <pb satisfying
Laplace equation
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
V2<& = 0 .
This main assumption holds as long as the Reynolds number based on the
bubble flow (i.e. maximum bubble size and bubble life-time) is large enough
to allow us to neglect the viscosity and the bubble does not generate addi
tional vorticity.
For bubbles in the boundary flow with linear velocity profiles17, two ef
fects were revealed (see Figure 3.29 where S is the distance from the initial
bubble centre to the wall; Rmax is the maximum radius the bubble would
reach if it grew spherically in an infinite domain, Pgo is the initial gas pres
sure) :
• Bubble Stretching in Growth Phase

The line in the figure crossing bubble contours shows the motion of
node 1 (which is initially on the Y axis) during growth and collapse.
The downstream point move away from the initial bubble centre much
faster than the upstream ones, resulting in a stretched growth form.
• Jet Modification in Collapse Phase

The formation of jet is delayed and its development is weakened even
ir
The velocity varies from a value of V,hcar at the far side of bubble of radius Rmaz to
the value of zero at the rigid wall with the basic pressure (i.e. the pressure in the basic
flow 3*|,) being assumed constant across whole shear layer and designated as Pom6-

96 Cavitation
Collapse <^^Sfe»^ ôw
m
1w 0.0 0.3
0.2 0.2
Growth J^tS ^n^^ Row Collapse Flow
0.0
m
mP
0.0
^HHHIJft
-0.2 -0.2
-0.3 0.0 0.3 -0.3 0.0 0.3
(O
Figure 3.29: The influence of a linear shear flow on the bubble dynamics in
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
the proximity of a rigid boundary (Chahine 1990 [3.49])

(a) e ( = VsHear/Vtoyleigh) = 0-043, 7 (= S/Rma^ = 1.77, Pg0/Pamb = 23
(PgO is the initial partial-pressure of the gas inside the bubble and Pamb is
the pressure at the infinity), boundary at Y = —0.30; (b) e — 0.173, 7 =
1.77, Pg01 Pamb = 23, boundary at Y = -0.30; (c) e = 0.043, 7 =
1.0, Pgo/Pamb = 23, boundary at Y = -0.17

for the case of very small shear, \ = 0.043 (Figure 3.29 (a)), compared
with the case in absence of shear flow. The jet direction is being
reversed while the bubble is undergoing its collapse process, i.e. from
the direction inclining towards upstream (the angle between upstream
and downward) to the direction inclining towards downstream (the
angle between downstream and downward).
The dominant parameters in such bubble dynamics are the relative distance,
7 — SI Rmax, characterising the strength of wall influence, and the nondi-
mensional shear strength \, which is the ratio of Vshear to the mean Rayleigh
velocity V/jayjetjh (VRayieigh is based on the Rayleigh time r and the value
Of Rmax, Vfayleigh = \/AP/p),
'shear 'shear
VRayieigh JÊ
The stretching and jet modification effects are enhanced by increased value
of x while the parameter 7 functions similarly as for the single bubble in
static fluid.
The influence of line vortex on bubble dynamics is demonstrated (Chahine
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
1990 [3.49]) to be different from the commonly assumed mode that the bub
ble grows and collapses spherically while it is being captured by the vortex.
If the bubble is initially located in the centre of vortex, it will be stretched
at the direction along the axis of rotation during its growth and further
disintegrate into two tear-shaped bubbles (Figure 3.30). Here, a nondimen-
sional parameter Q, the ratio of the pressure drop caused by rotation to the
ambient pressure,
2
(£) <" »
r
ft = P (
"amb
is adopted to characterise the vortex strength. The nondimensional pressure
at the vortex axis is valued as (1 — Q). For given value of fi, the parameter
of normalised core radius a^ affects bubble deformation,
cTc=-^- (3.23)
here Rmax is the maximum radius the bubble would grow to in an infinite
medium with the ambient pressure equal to the pressure at the vortex axis.
The bubble elongation depends strongly on fi. The bubble splitting is one
of the sources contributing to the inception noise observed in the inception

98 Cavilation
1 1"
_
I 1
I
Viscous kvtm
■ ■■*"«H5
1 -
1
-2
(a)
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
Figure 3.30: Bubble growth and collapse in the centre of a line vortex,
il — 0.95, PgQ I Pamb = 584 (PgQ is the initial partial-pressure of the gas
inside the bubble and Pamb is the pressure at the infinity) (Chahine 1990
[3.50])
(a) ac = 1.18; (b) ac = 0.5; (c) ac = 0.2

-i 1 •1 -I 1
l=.00l s 1=463 s I=.534 s
Figure 3.31: Growth and collapse of a bubble initially located a distance of

2 from the axis of a vortex, 0, = 0.948, Pgo/Pamb — 584.3 (Pgo is the initial
partial-pressure of the gas inside the bubble and Pamb IS t n e pressure at the
infinity), ac = 4 (Chahine 1990 [3.50]) --``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
(a) XOY plane view (perpendicular to the vortex axis); (b) Side view
stage of vortex cavitation. The bubble capture process has also been demon
strated by positioning the bubble initially at a normalised distance from the
vortex axis located at X — 2, Y ~ 0 as shown in Figure 3.31. From Figure
3.31(a), it is seen that the bubble approaches to the vortex axis in a spi
ralling form around the axis and, due to the presence of pressure gradient,
the bubble surface deforms remarkably, forming a reentrant jet directed to
wards the axis. From Figure 3.31(b), the bubble elongation along the axis as
well as the reentrant jet (side view) are revealed. Chahine [3.50] found that
the deformation of bubble surface during capture process is related to the
ratio of the acceleration force ^gradient (due to the pressure gradient) to the

100 Cavitation
bubble wall acceleration ^growth, and this ratio can be expressed in terms of
a characteristic bubble radius i?(, (for say Rmax here) to the viscous core size
Oc,
^gradient _ 4/tfc , .
a
^growth c
This indicates that Rt,/ac is a dominant parameter in such bubble interaction
with line vortex flow. The Weber number, i.e. the ratio of the variation of
local pressure around the bubble to the surface tension (coefficient as) which
can be expressed as
Wt -A(±)(*) (3.25)
is also a parameter affecting the deformation of bubble shape during the
capture process. The larger the Weber number, the higher the instability of
the bubble surface. Therefore, these four parameters, namely vortex strength
Q, nondimensional viscous core size a^, acceleration ratio -7'gradient/l'growth
and the Weber number We, are the parameters dominating the bubble-flow
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
field interaction in such a line vortex environment.
Statistical S t u d y
The classic static theory states that the microbubbles (nuclei) will cavitate
if the ambient pressure Pamb is equal to or lower than the critical pressure
Pcrit, i e .
*amb S *crit ■
Here
P - P -2-2<Js
and R* is the critical radius,
R
-V2^7'
with N being constant for a fixed mass of gas in the bubble; T being the
temperature; and as being the coefficient of surface tension. Actually, hy-
drodynamic cavitation is a stochastic and dynamic phenomenon, which can
be revealed by examining the dynamic equation for spherical bubbles, i.e.
the Rayleigh-Plesset equation,
2as 4uR / •• 3 A 2 \
*amb — Pg + Pv-^r — + p(RR+-R>) (3.26)
R

where Pamb: ambient pressure;

Pg + Pv — ^ : static force term;
^ + p(RR + | f l ) : dynamic force term.
On the one hand, the value of Pamb in a dynamic flow, particularly in a

turbulent flow, varies randomly in both temporal and spatial domains. On
the other hand, the microbubbles which take part in the cavitation inception
process are randomly distributed in the flow with random properties such
as their size and content etc. Therefore, the values of static and dynamic
terms are scattered widely for individual microbubbles. Owing to such ran
domness of both the driving force (i.e. Pamb) and the microbubbles (i.e.
the static and dynamic terms), the cavitation is thus a highly random and
dynamic problem. Besides, there are more factors, which are not included
in eqn (3.26), contributing to the randomness of cavitation. One is the non-
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
sphericity effect of bubbles. Another is that the level of pressure fluctuation
in that flow field will be further randomly enhanced by these violent bubble
activities involved in the cavitation process. The phenomenon of Cavitation
Hysteresis is just such an evidence. 18 .
Most of investigations examine two types of flow, namely boundary shear
flow and free shear flow. They seek to show how microbubbles (nuclei)
statistically respond to pressure fluctuations; and how the statistic properties
of these bubbles and the consequent cavitation inceptions are altered by
the flow. Although the statistical studies have been conducted since the
early 1950's, the phenomenon still remains obscure in many aspects. Knapp,
Daily & Hammit [3.65] and Arndt [3.38] reviewed the early investigations
up to 1970's. The main knowledge acquired from statistic studies can be
summarised as follows.
For cavitation in the boundary shear layer, the statistical properties of
bubbles are dominated by the pressure fluctuation field in terms of their
intensity and duration 19 . For turbulent flows, the instantaneous pressure
18
The cavitation hysteresis is such a phenomenon that the desinent cavitation number
ad is often higher than the incipient cavitation number o~i. It can be reasoned as follows.
Once the rather intermittent incipient cavitation occurs, the cavitating microbubbles will
further increase the pressure fluctuation level in the flow field which thus provides more
intensive driving force for bubbles to cavitate; and through this interaction mechanism,
the cavitation becomes a rather statistically steady one and can well remain so even for
an increased cavitation number during the desinent cavitation process.
19
Knowledge of turbulent pressure fields in boundary shear layers is much less than that
for free shear layers. This is due to the presence of the wall and the small scales near the
wall which make the problem theoretically and experimentally difficult to solve.

102 Cavitation
droppt is 5.5 to 7.9 times its rms value yp12 [3.52], i.e.
Pt
-- 5.5 ~ 7.9 . (3.27)
The value of sjpa can be related to the turbulence level by
(3.28)
For isotropic turbulence,
Va = ^pK(u'^) . (3.29)
Therefore, the distribution of the turbulence level, which varies across the
boundary layer, alters the statistical characteristics of the microbubbles' in
ception performance across the boundary layer. This phenomenon was firstly
demonstrated by Daily & Johnson, 1956 [3.52]. They applied Prandtl's or
der of magnitude argument to Reynolds' equation for 2D steady flow in x
direction and assumed the gradients of turbulence terms are negligible in the
x direction. A simple relation is obtained [3.65] as follows
Flow
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
P,00
Turbulent boundary layer
P.V
Figure 3.32: Two-dimensional turbulent boundary layer
p + pv12 = p i ( x )

where p\ (x): pressure in the free flow region beyond the boundary layer
where v'2 is negligible (referring to Figure 3.32);
p: mean pressure.
This relation indicates that the lowest mean pressure across the boundary
layer is at the location where the turbulence level is highest. It is obvious
that bubbles at this particular location have the highest probability to cav-
itate owing to the minimum value of mean pressure p and the maximum
value of instant pressure drop pt- The variation of va across the boundary
layer thus shifting the statistical distribution of cavitating microbubbles was
experimentally observed by Daily & Johnson [3.52]. The value of vn usually
increases from zero at y/S = 0 to a maximum at y/S = 0.1 ~ 0.2 then
gradually reduces to a certain small value at y/6 = 1. Daily & Johnson have
shown that in a boundary layer the statistical distribution of the cavitating
bubble population (Figure 3.33(c)) moves away from the statistical distribu
tion of the total bubble population (Figure 3.33(a)) towards the boundary
wall and it peaks at a distance y/6 where vn reaches a maximum. Following
Daily & Johnson's work, Arndt & Ippen [3.40] show a similar turbulence
influence for the boundary wall with distributed roughness. The influences
of magnitude and time scale of turbulence on cavitation inception were also
demonstrated by many other studies on inception scale effect. Arakari &
Acosta, 1981 [3.36], emphasise the importance of time scale by postulating
that the turbulent fluctuations may actually stall the local flow near the
wall, leading to a brief period of separation or a turbulent burst with re
verse flow. Then the nuclei within these regions may be exposed to a low
pressure longer than would otherwise be the case, thereby promoting the
growth of microbubbles. This is supported by the fact that the frequency of
the most unstable Tollmien-Schlichting wave in the laminar boundary layer
just prior to transition (on a 1.5 cal ogive axisymmetric body used in their
experiment) is about 5 kHz equal to a reference time period of 0.2 msec for
growth, which is about the same order as the bubble life-time (0.1 msec) ob
served. Huang [3.62] found that the inception is correlated with the spatial
amplification ratio A through the mechanism of laminar to turbulent tran
sition or laminar separation. For the tunnel with low level of free-stream
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
turbulence, the inception is well correlated with the computed amplification

ratio value of A = e 11 ; and, for higher levels of free-stream turbulence, it
takes place in a region where the value of A is less than e 11 , for say e or even
e 7 . Gates & Acosta 1981 [3.60] show that free-stream turbulence is an im
portant factor for inception scaling; and that the mechanism through which

104 Cavitation
Figure 3.33: The effect of turbulence level on the statistical characteristics

of bubble cavitating (Daily & Johnson 1956 [3.52])
(a) Statistical distribution of total bubble-population; (b) Turbulence-level
variation; (c) Resultant statistical distribution of cavitating-bubble popula
tion
this factor influences inception is the structural alteration of the boundary

layer in terms of its separation and/or transition (refer to §2.4 "Cavitation
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
Nucleation and Inception"). It is also found, e.g. [3.72], that due to the
Off-body Effect*0, the statistical distribution of cavitating bubble population
along the off-body direction (i.e. perpendicular to the surface) will be altered
since the off-body distance, which is a dominating factor in bubble's cavi
tating, has been randomly changed owing to the off-body effect. Therefore,
all these factors possessing strong randomness will thus affect the statistical
That is, the bubbles will deviate off the convex boundary surface when they travel in
the flow.

characteristics of the bubbles involved in the cavitation inception process.

For cavitation in the turbulent region of a free shear layer, such as the
wake surface of a sharp-edged plate and the high-shear region between jet
and surrounding fluid etc, the cavitation mechanism is essentially different
from that which takes place in flows around streamlined bodies. Its statisti
cal characteristics are also different. The turbulent eddies generated in this
type of shear layer show a better organised vortical structure, and cavitation
occurs in the low-pressure cores of these vortices. Early work as summarised
in [3.65] states that cavitation inception in such turbulent flows is not re
stricted to a single locality but occurs over a wide zone of space of high
shear and eddying showing a strong stochasticity since the successive eddy
strengths, instantaneous pressures and their durations all vary in accordance
with the random nature of turbulence. To account for the stochastic effect
of such pressure fluctuations, i.e. their temporal and spatial occurrence fre
quencies, the cavitation inception coefficient 0{, in a jet say, can be modified
by a temporal coefficient Ct and a spatial coefficient Cs as 21
v/P'2
<Ti = CtCs-^r .
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
2
Here, CtCs is about 3 if a, w 0.55 (Vo is the free-stream velocity). The

rms value of pressure fluctuation, y p ' 2 , is related to the turbulent kinetic
energy ^g- by a factor of 3.6 (which is much higher than that for isotropic
turbulence)
whereas the turbulent kinetic energy -£- is often higher than that in bound
ary shear layer and so is the ratio of peak pressure drop pt to the rms value
p' 2 [3-52],
Pt
= 10 ~ 13 .
Pa
Therefore, much higher local pressure drops will randomly occur in such
turbulent shear flows, leading to a much higher inception coefficient values
than in boundary shear flows.
21
For cavitation in a mixing layer, the temporal pressure fluctuations play a more im
portant role [3.38].

106 Cavitation
As to the relationship of cavitation with turbulent shear flow structure,

the study by O'hern (1990 [3.74]) in the wake of a sharp-edged plate shows
that cavitation inception consistently occurred in the streamwise vortices
and more fully developed cavitation is visible in both streamwise and span-
wise vortices, with the streamwise cavities typically confined to the braid
regions between adjacent spanwise vortices. Figure 3.34 shows the cavita
tion development in the turbulent shear layer with cavitation number ai
(ox indicates local cavitation number based on the mean local flow values
of velocity and pressure). The inception occurs in the form of long- and
thin cavities oriented in the streamwise direction when viewed from side,
referring to Figure 3.34(a). A detailed structure for a mild cavitation case
(ox = 0.65, Re,L — 2.06 x 10 6 ) 22 is presented in Figure 3.35, showing the
streamwise vortices are primarily confined to the braid region. With a Rank-
ine vortex model, the vorticity strength for streamwise vortices, Ts — 7ro;a2
(here, ac is the core radius of Rankine vortex, w is the vorticity), has been
found to be always less than 10% vorticity strength of spanwise vortices, T p ,
i.e.
£<0.1.
Nevertheless, the cavitation inception controlled by streamwise vortices

(which is a common feature of free shear layer flows) indicates the lowest
pressures occurring in the cores of these vortices. Owing to the velocity
dependent 3D instability amplification in the shear layer, the streamwise
vortices play a dominant role on cavitation inception and the inception in
dices thus depend on Reynolds number. For the case of Ui = 4.9m/s, and
rms pressure fluctuations of 153% of the local dynamic pressure, with peak
pressure fluctuation between +344% and —366% has been detected. The
pressure fluctuations observed fit a Gaussian distribution. The relation be
low explains that the negative peak of fluctuation, p', can cause a minimum
flow pressure pmin which is significantly lower than the mean free-stream
value used for evaluating ox and thus induces the inception locally at a
relatively high values of ox, that is
Pmin ~ Pv _ n P'
where Cp\, is the base pressure coefficient based on the local pressure and
22
The cavitation coefficiency OL is defined as CL = (J>L — Pv)/(^pUl), here PL is the
minimum local pressure above the shear region; p„ and p are the water vapour pressure
and density respectively; UL is the maximum velocity above the shear layer.
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

(a) (b)
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
(e) (0
Figure 3.34: Cavitation development (side view) with decreasing cavitation

number OL (note: a different lighting scheme was used for (b)) (O'hern 1990
[3-74])
(a) Re<L = 1.5 x 106, aL = 0.87; (b) Re>L = 2.0 x 10 6 , oL = 0.78; (c)
Re,L = 2.1 x 106, aL = 0.65; (d) Re,L = 2.1 x 106, aL = 0.65; (e) Re,L =
2.1 x 10 6 , aL = 0.41; (f) Re>L = 2.1 x 106, aL = 0.40

108 Cavitation
Figure 3.35: The early stage of cavitation in a turbulent shear layer, UL =

11.38 m/s, OL = 0.65, Re L = 2.06 x 10 6 . Top and side views (O'hern 1990
[3.74])
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

velocity. Owing to the random nature of the vortices and pressure fluctua
tions, the cavitation inception takes place at random locations throughout
the shear layer and no clear downstream or upstream location of the most
probable occurrence of inception could be found. Another feature of this
vortex cavitation is its shedding frequency which is characterised by the
Strouhal number
fd
U
where /: vortex shedding frequency;
d: characteristic length, e.g. the jet diameter;
U: characteristic velocity, e.g. the jet velocity.
This is because the coherent structures play a vital role in the determination
of large-scale pressure fluctuations. For example, the cavitation in ring-like
bursts in turbulent jets appear to have a Strouhal number close to the value
(0.5) for the coherent vortex ring structure in the non-cavitating jet [3.38].
This Strouhal frequency often adds a periodic attribution to the randomness
of cavitation (Further discussion can be found in §3.9 "Power Spectrum of
Cavitation Noise"). The occurrence of cavitation will also alter the flow
structure. Young & Holl [3.81] and France & Michel [3.56] reported the
variations of vortex spacing and Strouhal frequency with cavitation number.
(a)
Figure 3.36: Spatial distributions of phase-averaged pressure-fluctuation

peaks in the near field of a submerged jet flow, jet diameter d =■ 2.5 cm,
jet velocity Vj = 17.5 m/s (Ran & Katz 1993 [3.76])
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
(a) Trigger transducer is at x/d — 1.25 and r/d = 1.5; (b) Trigger transducer
is at x/d — 2.5 and r/d — 1.5

110 Cavitation
N |/No Poo =1.7 bar; Vj=17 5 m/s N l/No p„ = |,7bar; Vj=17.5 m/s
050 O.SU
N | -number of pressure peaks with N l -number of pressure peaks with
040 the indicated amplitude 0.40 the indicated amplitude
No -total number of pressure peaks No -total number of pressure peaks
0.30 0.30
0.20
0.10
ooo I
■ ■
HiMM li
_li
0.20
0.10
0.00
■ 11
IIBB,I»I ,■
-10 -0.8 -0.6 -0.4
-0< -0.2 00 -1.0 -0.8 -0.6 -0.4 -02 00
C M P 1 -P ooV(05 pVj 5 Cp=(P 1 -P oo)/(0.5 PVj \
(a) (b)
Figure 3.37: Probability density distribution of pressure-fluctuation peaks

(Ran & Katz 1993 [3.76])
(a) Location A; (b) Location B
The dependence of statistical characteristics of inception (for the cavi

tation occurring in the near field of a submerged jet) on both the coherent
structure of shear layer (in terms of spatial distribution of its pressure fluc
tuation peaks) and the spatial distribution of microbubbles is clearly demon
strated by Ran & Katz, 1993 [3.76], with air-bubble injection and holographic
technique. The spatial distributions of "phase averaged" pressure fluctua
tion peaks in the near field of jet are shown in Figure 3.36 with the maxima
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
of the spatial distributions at x/d — 1.25 (marked as A) and 2.5 (marked as

B) respectively. And, there are also additional maxima at x/d — 0.5 (Figure
3.36(a)) and x/d = 1.5 ~ 2.0 (Figure 3.36(b)) which indeed indicate the
process of vortex pairing. The probability density distributions of the phase
averaged pressure fluctuation peaks for the locations A and B are in Figure
3.37 showing most of the negative peaks ranging between —0.7 ~ —0.9 (the
distributions for x/d = 2.5 are more scattered but still 15% of the negative
peaks falling in the range of —0.9 ~ —1.0). In fact, the highest negative
peaks, —0.97, agree well with the cavitation inception number,CTJ= 0.99.
By assuming that the spatial distribution of bubbles is independent of the
pressure peaks' probability and their spatial distribution 23 , Ran & Katz es-
23
Obviously, this assumption neglects the capture effect of bubbles by the vortices
as demonstrated by Chahine's study (see "Single/Individual Bubble Study" of §3.7.4

24
timate the probability of inception /J.(<J = <Ti,x, r) by
fi{a = aux,r) = n fp < pv - — , x , r ) [i(R,x,r)
Here, fi(p < pv — ^,x,r) is the probability that the local pressure is low
enough to induce inception (which can be evaluated from the probability
density distribution of pressure peaks), and fi{R, x, r) is the probability that
a bubble presents at the site (which can be evaluated from the test as shown
in Figure 3.38). It is noticed that there is a local bubble population den-
Figure 3.38: Phase-averaged spatial-distribution of bubble-number density.

The trigger is at x/d = 2.5 and r/d = 1.5 (Ran & Katz 1993 [3.76])
sity peak at the location of high-negative-pressure peaks x/d = 2.5 and

r/d = 0.7. The calculated probability distributions of cavitation inception
is presented in Figure 3.39(a) for the values of cavitation number CT, rang
ing from 0.93 ~ 0.5. It is convinced that the initial inception (i.e. when
Oi > 0.90) occurs only in the vortex pairing region where the maxima of
both the negative pressure peaks and the bubble population density present
(coincide).
"Bubble-Flow Interaction"). The strong slip velocities of the order of 10% observed in
this flow is an evidence of such capture effect.
24
Here, actually, they assumed that the microbubbles with n > R all have enough time
to cavitate, i.e. the bubble characteristic time -n, is much less than the characteristic time
of pressure fluctuation T P , fi(n, -C T P , I , r) = 1.
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

112 Cavitation
OUT90C UW IX
MCitucin »
Probability of cavilation inception (%)
1 measured bubble distribution

2 uniform bubble density
^ c 3 uniforjn bubbleflux-density varies
100
0.60 0.70 0 80 0.90 100

Cavitation index o~
(b)
Figure 3.39: Probability of cavitation inception within the entire near field,
i.e. 0 < x/d < 3.5 and 0 < r/d < 1.0 (Ran & Katz 1993 [3.76])
(a) Contour plots; (b) Comparison with another two cases which assumed a
uniform bubble-density and a uniform bubble-flux respectively
The influence of bubble distribution on the inception probability can

be demonstrated by evaluating the probability of a single cavitation event
present in the entire field (0 < | < 3.5 and 0 < g < 1.2), which is
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
for three cases of different bubble distributions (i.e. the measured density
distribution, uniform density distribution and uniform bubble flux distri
bution 25 ), referring to Figure 3.39(b). This result is a good example of
bubble-flow interaction: The migration of bubbles into the primary vor
tices increases the bubble density at the locations of high-negative-pressure
peaks, leading to a higher likelihood of cavitation inception. Therefore, the
inception probability for the measured bubble distribution possesses highest
25
In the last case, bubble flux in the shear layer is uniformly distributed but the bubble
density decreases along the x direction as the shear layer grows.

values over the range of a = 0.84 ~ 0.96, i.e. the re-distribution of bubble
density by the bubble-flow interaction modifies the statistical characteristics
of cavitation bubbles in terms of promoting their inception.
In this jet flow, inception occurs in pairing vortices rather than in the
braid region (i.e. the secondary structure as demonstrated by O'hern [3.74]
in the 2D free shear layer flow). This can be reasoned as follows [3.76]. The
circulation of a single vortex can be estimated as VjA (A is the characteristic
wave length). With a Rankine vortex model, the pressure coefficient in the
centre of vortex is
r, Pc-Poo 1_ /_A_
°p \pVJ ~ TTÛC
During early stages of vortex pairing, the radius of vortex ring rv increases
from rv/d — 0.62 to 0.77, causing the vortex stretching and ac reducing. If
the volume of the vortex ring, i.e. a%rv, remains constant, the cavitation
inception number Oi (i.e. —Op) is then proportional to rv. Therefore, the
stretching increases the value of —C'v from 0.78 to 0.97, resulting in the early
inception in the stage of vortex pairing. However, if the jet diameter is large
enough, the high-pressure peaks and corresponding inception would firstly
occur in the braids because the secondary structure of shear layer for larger
jet plays more significant role and so does its stretching.
The attempt, by using the Kolmogorov theory for homogeneous and
isotropic turbulence, to relate the properties of the temporal pressure field
to cavitation inception for both free turbulent shear flow and fully developed
boundary-layer flow has been made by Arndt & George 1979 [3.39]. For
free shear flow, a hypothetical pressure spectrum in Lagrangian frame of
reference (Figure 3.40) was proposed for analysing the bubble responses to
the pressure fluctuations. The spectrum is normalised with respect to the
mean square pressure and the Lagrangian time scale, d,
d ~ -7
u'
here / is the length scale of smallest eddies, u' is its velocity. Based on Figure
3.40(a), by integrating the spectrum from u> = 0 to w = ^~ (TB is the time
scale for bubble growth at inception) and assuming a normal distribution
of nuclei size, the fraction of the mean square pressure contributing to bub
ble growth without appreciable tension can be evaluated since any pressure
fluctuation will persist for a time longer than the time scale of bubble as
long as its frequency w is less than J-. If the bubble frequency is larger
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

114 Cavitation
f Ap^coOdco'
A nn(CO)
Mean square pressurefluctuation(Lagrangian)
i/rL i/rB 0.2 ( E / v ) 0.2( e / v )
(a) (b)
Figure 3.40: Hypothetical pressure spectrum in a Lagrangian frame of refer

ence. Here, (b) is the integration form of (a) (Arndt & George 1979 [3.39])
Relative concentration (%) Relative concentration (%)

All bubbles All bubbles
Cavitating bubbles "~ "*" Cavitating bubbles
30 _ 30 —
-
U
,TB -- II u
, T
B
25 - =0.1 25 =1.5
h
' - 1 I h
20 - 1 _ 20
—
r 1
1
15 - 15 -
' 1 1
10 - 1 10 1
1
— J
|~ 1 ,__, _ 1
5 r. 5 -
1
1 1
1 1 ">- 1 1 i i r~l—
0 .2 .4 .6 .8 1.0 1.2 0 .2 .4 .6 1.0 1.2
y/ 8 y/ 6
Figure 3.41: Statistical characteristics of cavitation in turbulent boundary

layers (Arndt k Ippen 1967 [3.40])
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

than the Kolmogrov frequency (^)i (which is roughly corresponding to the

state such that the turbulence Reynolds number ^ is less than ( ^ - ) 2 ) , the
entire spectrum may contribute to bubble growth. For turbulent boundary-
layer flow, Arndt & George suggest that the spectral peaks in the outer
layer (^ > 0.1) can be evaluated by the Lagrangian scale, 3, and the highest
frequencies in the flow will be
2
^f- for smooth wall;

^ for rough wall,
26
here, u* is wall friction velocity and h is the roughness height. Thus, if
TB < 1 for smooth wall;
Gr TB < 1 for rough wall,
the nuclei in the flow will have enough time to respond to the entire spec
trum. Otherwise only a fraction below the frequency of ?p- is sensed by
the bubbles. The experiment by Arndt & Ippen [3.40] demonstrates the
impact of above time-scale ratio on the statistical characteristics of bubble
cavitation performance: the zone of maximum cavitation is shifted inwards
as ( ^ ) TB is reduced from 1.5 to 0.1 (Figure 3.41).
3.7.5 Remarks
The stochasticities of the original nucleus properties are modified and char
acterised by bubble-boundary, bubble-bubble and bubble-flow interactions
throughout entire cavitation process. The resultant cavitation and the asso
ciated noise is thus a highly stochastic phenomenon.
Cavitation is not only a passive visualisation of flow; it also changes
the flow structure. Therefore, the resultant flow is definitely not the non-
cavitated structure embedded with bubbles.
The introduction of flow-wall interaction by using a compliant bound
ary will lead to a triple interaction between bubble-flow-wall, resulting in
a more complicated stochastic process which is still a subject that has not
been addressed. A Markov chain featured by a random-walk model with an
u' = y/o\Jp, hereCTU,is mean wall shear stress.
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

116 Cavitation
absorption wall on the boundary might be suitable to simulate this stochastic

process of bubbles 27 .
" T h e feasibility study (S C Li and P W Carpenter, 'Effect of Wall Compliance on Sta

tistical Behaviour of Cavitation Bubbles' (detailed 3rd version), Fluid Dynamics Research
Centre, University of Warwick, Jan. 28, 1994; and also S C Li and P W Carpenter 'Note
on an Envisaged Markov Model for Cavitation Bubble(s) near Compliant Walls', FEDSM
2000, Boston USA) for an ongoing EPSRC project (EPSRC GR/L74729) suggests: (a) It
is possible to present a single bubble in an n-dimensional-discrete-phase-space n , which
can be composed, for example, by certain random variables: characteristic bubble size
R(ui) (e.g. a volume-equivalent radius Rvoi), its distance to the compliant wall S(u>) (or
its relative value S/Rmax) and microjet magnitude J(u>) (at least, three values of-1, 0, 1,
representing repelling, no and attracting jets, can be assigned to J(u>)). Thus, the bubble
random-movements £(u)) is: f (w) = (R(v), S(u>), J(u>)); (b) The stochasticity of bub
ble behaviour can then be represented by a Markov chain with non-stationary transition
probability. It might be formed by pairs of (£„, 6.+1) with two-step transition probabil
ity, p(('i*)IO'i*)) =P('|fc> j) p(fc|j, i), if the system has strong momentum effect whilst the
selected space of status Q is not large enough to include necessary information; (c) In the
case of a compliant wall, the absorption function will have to be reduced or even made
negative. These conjecturs are going to be clarified by further experiments.
--``,```,,``````,,,,`

3.8 Stochastic Models of Cavitation Bubbles

3.8.1 Introduction
The pressure wave produced by each cavitation bubble and its occurrence
time at a given spatial position are random variables. There are several
stochastic models proposed to simulate these stochasticities for various cav
itation situations.
3.8.2 Single-Event Model

In this type of model, the bubble events are treated as a random sequence
of single events, in which the pressure waves produced by activities (growth,
collapse and rebound) of each single bubbles are regarded as single events.
And, the sequence of single events is considered as a stationary random
process.
Fitzpatrick &Strasberg, 1956 [3.55], assumed a model that the time in
terval A between successive events is an independent random variable whose
probability density obeys a Poisson distribution. The early stage of trav
elling cavitation in a fairly steady flow environment can be referred to this
model.
Il'ichev & Lesunovskii, 1963 [3.63], proposed a model, in which the events
appear at a periodic time interval with a stochastic distribution of zero mean
value, to simulate the single bubble events possessing periodic feature. That
is, the mth event will appear at the time tm,
Un = mT0 + A m
where TQ: dominant (periodic) time interval, at which the probabil

ity of event appearance reaches maximum;
random variable with mean value of zero, describing the
stochastic feature of event occurrence, and A m < -£.
This model is applicable to the early stage of cavitations in which the bubble
generation is periodically controlled, such as acoustic cavitation, turbine
blades and ship propeller cavitation.
Instead of assuming the identical single events (i.e. each event has same
waveform and amplitudes), Morozov, 1969 [3.73], proposed a model of Gen
eralised Random Pulse Train to simulate the cavitations in which the event
appearance time, amplitude and duration all can be random variables. The
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

118 Cavitation
single event p(t) is approximated as an exponential function
Pit) =VmZ~*
in which a proportionality relation of the amplitude pm and duration pa

rameter T with the maximum bubble radius .Rmax is assumed. The power
spectrum of such a generalised model will incorporate:
• The spectrum of single event from individual bubble (collapse);
• The statistical distribution function for the duration parameter T and

the statistical distribution function of Rmax, due to the linear relation
between pm, r and Rmax28-
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
• The characteristic function u\ (or u^) of time interval A (or time

deviation from periodicity, A), UA = E{e~lu'K}.
Therefore, Morozov's generalised model can simulate a wider range of cavi-

tations.
3.8.3 Multi Event (Cluster) Model

The single event model is not adequate to simulate the stochastic process of
bubble events if they appear in the form of clusters. In each of such clusters,
there are two or more bubble events involved. The cluster formation can be
attributed to the following causes.
• The bubble-bubble interactions do not become negligible. The bubble-

microbubble (small scale) interaction will have to be taken account if
the main bubble disintegrates into small bubbles and, subsequently,
they collapse under the influence of main bubble collapse. In such case,
the cluster is composed of the main event (by the main collapse) and
the subsequent events (by microbubbles). The bubble-bubble (large
scale) interaction is often the case when cavitation develops into a more
severe situation in which the bubbles are no longer to be considered as
isolated from each other.
• The bubble-flow interaction gives rise to the formation of clusters. For

example, due to the capture effect of a vortex, a cavitation cloud often
28
Probability theorem: if a linear relation exists between two random variables, their
statistical distribution functions will be geometrically similar.

forms in the core of a vortex with increased bubble number density.

Therefore, the collapse of bubbles will rather be in cluster form. This
often occurs in the cavitated vortices, e.g. those shed from sharp-
edged bodies for which the time interval of successive clusters is a
random variable with a certain regularity but without an exact period
(for detail see "Regularity and Periodicity" of §3.9.3 "Spectrum for
Sequence of Single Events").
• The cavitation bubbles are originally generated in cluster form. A typ

ical case is sheet cavitation on the blades of ship propellers. When the
blades passing through the wake flow, they undergo periodic pressure
variations and the sheet cavitation (subject to operating conditions)
periodically form, detach and collapse as a cluster or clusters down
stream of the blades. A similar situation occurs in hydraulic turbines
and pumps. The runner's blades pass through the wakes of guide
vanes and undergo periodic incidence variation, resulting in intermit
tent leading-edge sheet cavitation which periodically forms, detaches
and collapses as a cluster or clusters downstream of the leading edges,
e.g. as reported by Brophy et al [3.47]. In such cases, the sequence
of clusters has a periodic attribution with a random deviation from
the exact period. The period either equals the reciprocal of the blade
passing frequency, if all the blades are cavitating similarly, or the shaft-
rate frequency, if only one blade is cavitating or each blade is cavitating
differently.
For the sequence of events within the cluster, several models have been
proposed. Basically there are two trends.
(1) The events in a cluster are rather close to each other, appearing in
the proximity of reference time (which is often referred as the appearance
time of the cluster) with a high probability of the time interval (between
successive events) variable being zero. The cavitation situations referred to
in this model are those which have such a cloud-collapse condition that the
pressure rise (or pressure wave) reaches the bubbles in the cloud without
an obvious time-difference. For example, when a detached sheet cavitation
cloud flows into a higher-pressure region, if the void fraction of cloud is fairly
low, the pressure-rise wave encountered by the cloud will propagate at a high
speed, reaching all bubbles without obvious time delay.
(2) The void fraction of cloud is so high that the driving pressure wave
propagates at a remarkably reduced low speed in the bubbly medium. The
most outside layer of bubbles collapses first then the collapsing develops
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

120 Cavitation
inwards at a very low velocity as simulated by Hansson & M0ch's model

(referring to Figure 3.27 and "Bubble-Bubble(Large Scale) Interaction" of
§3.7.3 "Bubble-Bubble Interaction"). In such a case, the events within the
cluster scatter much more widely with a low possibility of the time-interval
variable being zero.
3.8.4 Comprehensive Model

Cavitation processes often consist of both clusters and single events, for
which neither the cluster model nor the single-event model alone is adequate.
Therefore, the combination of them, i.e. the comprehensive model, is thus
proposed for simulating certain cavitation situations.
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

3.9 Power Spectrum

3.9.1 Introduction
The sound generated by cavitation can be detected underwater over great
distances; and its power spectrum reflects the mechanism by which the cav
itation is generated. Due to this correlation, the spectra from various cav-
itations have been extensively studied since the First World War in order
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
to detect and monitor cavitation activities associated with warships. Nowa
days, it is also being used for various industrial applications, e.g. turbine
cavitation monitoring.
In reality, cavitation signals are mixed with turbulence and machine
noise. Hence, it is vital to distinguish cavitation noise from the background
noise by various detection techniques in the first place. The features of cav
itation signals picked up by hydrophones are also affected by the following
three factors. (1) The mechanism generating the cavitation plays an domi
nant role in creating distinctive features in the spectrum. (2) The distance
from the cavitation source to the hydrophone affects the recorded pressure
waves since the dispersion effect alters the waves with travelling distance.
(3) The containment, i.e. if cavitation does not occur in an effectively un
bounded medium, should be considered. If the hydrophone is placed near
the wall within the distance of wavelengthx(27r) _1 , the reverberant effect
strongly depends on the nature of the wall boundary and the pressure wave
frequency. Otherwise, the pressure sensed by the hydrophone is indepen
dent of wall nature. For flush-mounted hydrophone, if the wave length is
not much smaller than the surface dimension of hydrophone, the pressure
sensed will be influenced by the adjacent boundary subject to its acoustic
impedance. Further detail on the reverberant effect is beyond the scope of
this book. For information on this topic, particular for commonly used test
facilities, reference [3.46] is suggested.
3.9.2 Spectrum of Single-Bubble Collapse

The energy spectrum of a single cavitation bubble is the basis for under
standing the power spectrum of cavitation noise.
Fitzpatrick & Strasberg [3.55] give a conceptual illustration of the energy
spectrum generated by a single cavitation bubble. They assume an empty
vaporous bubble grows, collapses and rebounds spherically in an incompress
ible liquid under certain external pressure variation. The variation of bubble

122 Cavitation
radius and the resultant sound pressure are calculated based on Rayleigh-
Plesset equation. The bubble is also assumed to collapse to an indefinitely
small radius and to rebound with an arbitrarily postulated loss of energy.
The result is shown in Figures 3.42 and 3.43. Here, pe is the environmental
IP* Es.
RP 0 'P 0
Figure 3.42: Radius and sound-pressure variations for a single vaporous

bubble (Fitzpatrick & Strasberg 1956 [3.55])
pressure that would exist at the bubble trajectory in the bubble's absence;
ps is the resultant sound pressure which is evaluated based on linear sound
theory 29 ; R\ is the maximum radius of bubble and Po is a reference pressure
equal to the pressure difference between the external pressure pe at the in
stance of final phase of collapse and the vapour pressure. The dotted lines
in Figure 3.42 indicate the result for the case that the external pressure is
constant and equal to Po- The sound pressure firstly oscillates once due to
the growth and the initial phase of collapse, and then rises rapidly reaching
an extremely high value and forming a narrow "spike" owing to the violent
29
That is, ps(r,t) = pV 4|ry , here V is the instantaneous volume of cavity, r is the
distance to the cavity centre, C is the sound speed.
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
Figure 3.43: Radius and sound-pressure variations for a single vaporous

bubble which rebounds (Fitzpatrick & Strasberg 1956 [3.55])
collapsing at the final phase of collapse. The life time of bubble, i.e. the
time span from initial growth to final collapse, determines the features of
the low-frequency portion of the energy spectrum (Figure 3.44). The peaks
in the spectrum locate at the frequencies of the order of reciprocal of life
time, and the spectral density at lower frequency region varies (increases) as
the fourth power of frequency, i.e. f4 relation 30 . The oscillations appearing
in the spectrum are due to the multi-collapses of cavity. The oscillations
will be smoothed out for the case of multibubble events since the events are
correlated with each other. However, at higher frequencies, the smoothed
spectrum reveals itself even for the single-bubble case. The narrow spike
features in the high frequency portion of spectrum. The spectral density
at high frequencies varies (decreases) as the reciprocal of two-fifth power of
frequency, i.e. / _ 2 / 5 relation. However, it is known that at the final phase
of collapse the liquid compressibility is no longer negligible especially for va
porous cavity. Therefore, the spikes which possess the nature of a shock wave
30
This relation is based on the condition that the bubble is isolated from the interactions
with neighbouring bubbles and/or boundary. Otherwise, it would be modified accordingly.

124 Cavitation
r2S
5.0 \ Sdf=j ifdt
2.0
1.0
0.5
2.0
0.1 0.3 1.0 3.0 10.0
m,[P/R, f
Figure 3.44: Energy spectrum of a single vaporous bubble in an incompress

ible liquid (Fitzpatrick & Strasberg 1956 [3.55])
and the high-frequency portion of the resultant spectrum must be modified

based on the compressible theory. Using Gilmore's equation derived based
on the Kirkwood-Bethe hypothesis 31 , Fitzpatrick & Strasberg demonstrated
such modification at the high-frequency portion as indicated by the dotted
lines in Figure 3.45. Their theoretical modification do resemble the observed
spectra of cavitation noise (e.g. Mellen 1954 [3.71]) in which the spectral
density in the high-frequency region varies (decreases) approximately as the
reciprocal of the square of frequency f~2 (—6 db/octave). Such f~2 variation
of spectral density at high frequency can thus be regarded as the indication
of a sharp pressure rise in the radiated sound pressure, i.e. the shock-wave
formation.
Thus, the low-frequency portion of spectrum which relates mainly to
bubble life-time can be evaluated based on incompressible theory whereas
the high-frequency portion which relates to the shock wave (i.e. the spike)
31
That is, the quantity of r ( / i + | u 2 ) propagates outwards with variable velocity of (c+u),
here r is the radial coordinate; c is the local value of sound speed; h is the enthalpy and u is
the local flow velocity of liquid particle. For detail, see §3.3 "Viscosity and Compressibility
Effects".
--``,```,,``````,,,,``,`,,`,`,``

0.1 1.0 10 100 1000

fR,[P„/Po f
Figure 3.45: Energy spectrum of a single vaporous bubble in a compressible

liquid (Fitzpatrick & Strasberg 1956 [3.55])
needs to be evaluated with compressible theory. Therefore, the pattern of

the resultant spectrum is dominated by the proportionality factor Q of the
shock, defined by Baiter 1986 [3.41] as (referring to Figure 3.46(a))
Q _ P2TJ3
PITB
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
where p\: characteristic pressure preceding main shock pulse;

pV characteristic pressure of main shock pulse;
rg: duration of shock pulse;
Tg: bubble life time.
According to the value of Q, the single noise event p can be categorised as
three types, referring to Figure 3.46(b). When Q -C 0.01, i.e. the pressure
produced by the collapse at its final phase does not possess a shock nature
but has basically the same nature as the pressure produced by the preceding
performance (i.e. the growth and the early phase of collapse), the incom
pressible theory is applicable and the single bubble noise event p (i.e. the
sound pressure radiated from single bubble) can be evaluated with linear

126 Cavitation
Q«0.01
Q= 0{0.01
logf
logE
tB-
IV
Q>1 f* = l/2it-cB
^ 0 \
logf
(b)
Figure 3.46: Categories of cavitation noise events from a single bubble

(Baiter 1986 [3.41])
(a) Definitions of p i , p2, TB and T B ; (b) Three types of single events
sound theory as
« « . , ) - , ^
with its Fourier transform, P(f),
oo
pify-Wdt
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
/ -oo
and energy spectrum density, E,
'P(/)|2
E(f) = pC
This spectrum is characterised by a maximum located approximately at the
frequency corresponding to the reciprocal of bubble life time TB followed by

several maxima descending quickly with frequency / . The spectral density

varies (increases) with fourth power of frequency / , i.e. it is a / 4 relation,
before reaching the first maximum. When Q is of the order of 0.01, i.e. the
shock-wave nature is no longer negligible, the pressure pulse produced from
the more violent final-phase collapse has to be evaluated with compressible
theory. The low-frequency portion of spectrum has the same features as
for Q -C 0.01. But the decay of maxima becomes more gentle, i.e. the
Smoothing Effect, and the spectrum extends to the higher-frequency range,
becoming a flat portion, due to the more violent collapse. It finally rolls
off at the frequency of / * = z£f~- When Q = 0.1, i.e. the shock wave
plays dominant role in the event p whereas the preceding pressure becomes
much less important, the spectral pattern for Q in the order of 0.01 is still
revealed but with the main maximum not being essentially higher than the
flat portion. If the value of Q is further increased to Q > 1, i.e. the shock
has become so important that the preceding portion is negligible, the event
p can be approximated by the shock wave alone which is often described by
an exponential function,
t_
T
p = p2e B .
The spectral density becomes
(PVB)2
E(f) = PC{\ + (27TT B /)2)
as shown in Figure 3.46(b). This spectrum features a flat portion extending

l
up to / a TA—
2TTT
then rolling off at the rate of / 2
B
3.9.3 Spectrum for Sequence of Single Events

Some cavitations can be modeled by a sequence of single events. For the
conciseness of formulae, Baiter's notation [3.41] is adopted throughout the
remainder of this section. That is, the single event is defined as x,
P
x(t)= ■
(pc)1*'
the Fourier transform of x(t) is
oo
i ( i ' ) e - i 2 , r ^ dt';
/ -oo

--``,```,,````
128 Cavitation
and, the energy spectral density of the single event x(t) is
E(f) =| X(f) |2 .
Thus, the cavitation noise, which consists of a sequence of single events
xm (m = 1,2,3, • • • N) in a certain sampling period T, can be expressed as
y(t,T),
N
y(t,T)= £x m (t-0 m )
m=l
here 6m is the occurrence time of the mth event xm. The Fourier transform
of y(t,T) is
Y(f,T)= J2 ^m(/)e-*^m
m=l
here Xm(f) is the Fourier transform of the mth single event xm. The power
spectral density of cavitation noise produced by such a sequence of single
events is
4>{f) = ^ i m ^ i | y(/,T) | 2 }
K m=ln=l )
here X^(f) is the complex conjugate of Xn{f). The parameter T in the

above expression of <j>(f) can be replaced by n (— E{N/T}) if: (a) N/T is
an independent variable; (b) the dependence of 9n on T is negligible. The
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
power spectral density <f>(f) thus becomes

N N
m=ln=l
N
1 /
= n lim — V E ( | Xm | 2 )
\m=l
■N-lN-m iu{e 9 1\
+2fi e Y: J:
.m=l
E{xmxuke~ -~ ^}
fc=l J /
If the following additional assumptions are made: (a) The random variables
of amplitude and shape of the single event x(t) are entirely independent of
the variables related to the features of the sequence such as the occurrence

time 0\ (b) the values of random variables for any single event in the sequence
are independent of the values for other events in the sequence; and (c) the
process is a stationary random process, then the power spectral density can
be written as
r « - l JV-m
^(/)=n||X(/)|2+|Jf(/)|2Alimo-fie Y" y e-iu($n-em+k)
m=l Jfc=l
(3.30)
with the following notations
X{J)? = E{\Xm(f)\2};
Xjf) = E{Xm(f)};
-iu{om-em+k) E fe-M9m-em+k)\
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
Poisson Process
For cavitation where the bubbles are entirely independent of each other, the
cavitation noise process is often assumed to be a Poisson process, i.e. the
events occur in such a way that the probability density of time interval A
between successive single events (referring to Figure 3.47(a)) obeys
WP(A) = A_1e-A/X
here A is the expectation of A. Actually, it is a special case of a Gamma
11 To | To | To | j t
(a) (b)
Figure 3.47: Random sequences of single events x. (a) Time interval A; (b)
Deviation A from the periodicity TQ
distribution with a — 1 (referring to Fig. 3.48). Owing to the relation

130 Cavitation
1.00 I- W(A)
a
a
W(A) = A e A
0.75
0.50-
0.25-
0 1
Figure 3.48: The Gamma (r) distributions and the Poisson distribution
8m = J2AJ,
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
j=i
the following holds
e-Mem-em+k) = uk^ (3 31)
here u\(f) is the characteristic function of Poisson distribution Wp(A),

r+oo
u\(f) = e —iwA
-i: e-^Wp^dA .
Substituting (3.31) into (3.30) and replacing the last term with its limiting
value,
"A
4>(f)=n(] X I2 + 2 | X Y Re
+ (nX(0)) S(f-0)
2
(3.32)
.1 - " A
here 8(f — 0) is the Delta function for / = 0,
1, / = 0 ;
8 =
0, / ? 0 .

Since Wp(A) is a Poisson function,
Re 0.
.1 - « A
Expression (3.32) reduces t o
4(f) = n\ X(f) | 2 + (nX)2S(f - 0) . (3.33)
That is, the power spectral density consists of a continuous component and
a spectral line at / = 0. This result indicates that except for / — 0, the
power spectral density </>(/) is equal to that the mean value of energy spectral
densities of single events, | X(f) | 2 , times the mean event rate n ( = ^ ) , which
is often referred as the Campbell theorem 32 .
Generalised Process
Morozov [3.73] further generalised his model to simulate sequences of single
events with various stochastic natures, allowing (referringv5ch3-9.tex to §3.8
"Stochastic Models of Cavitation Bubbles"): (a) the pattern of event x(t)
to be various functions; (b) the characteristic parameters of event x(t) to be
random variables; and (c) the process to be different random processes other
than the Poisson process. For the conciseness of formulae, here the event
is assumed to be identical, i.e. |A"|2 = |X| 2 = |X| 2 , and Baiter's notation
[3.41, 3.42] is followed, although there are no difficulties in extending to the
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
case where both the event patternv5ch3.9.tex and its parameters are random
variables as in [3.73].
If the time interval A renders a Gamma (r) distribution, that is
Wr(A) = V . ( A°-'e-°AM a ^ i
T(a)
with the relation

1 _ crA
y/a A
where A: expectation of time interval A;
CTA: standard deviation of time interval A,
32
It was Morozov, 1969 [3.73], who borrowed the Campbell theorem from communication
theory and applied it to cavitation noise study.

132 Cavitation
the power spectral density <fi(f) for such a Gamma process can be obtained
following a similar procedure as for the Poisson process. Function V(/) OI"
the power spectral density <f>(f) is often used to characterise the random
process, which is defined as
n | X(f) |2 •
For such a Gamma process (referring to Figure 3.49(a)),
(1 + ( ^ ) j -2(l + (^) j2cosatan-i^ + l
The power spectral density consists of a continuous component and a spectral

line at / = 0 with
iK/) = ^y+nS(f-0) / = 0;
i>U) = 1 /->oo. .
With a decreasing value of 2A- (i.e. the events have more regularity in occur
rence time such as the bubbles associated with shed vortices) several maxima
emerge at / A = 1,2,3, •••. It is noticeable that when o\/h. = 1 (i.e. the
case of a Poisson process, for which a = 1) the value of ip(f) is always unity,
which again indicates the Campbell theorem.
For a Gaussian process, i.e.
_(*-A)2
V27TCTA
the random process is characterised by

2
sinh K )
cosh r"^' cos u> A
as shown in Figure 3.49(b). It has similar features as for the Gamma process
except for some minor deviation of i>{f) (when SA- = 1) from unity in the
region near / = 0. Therefore, it is practically acceptable that the Campbell
theorem is still valid for a Gaussian process as long as the single events have
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

OA/A OA/A
a
|
1 1
b uJT :
c c
uJT
vJT uJT
d 1/2 d 1/2
e i/VT e i/VT
f 1//T0 f 1/ -/To
g 1//20 8 1//20
h 11 Jw h 1//40
~
, c b " /I
d . I /f
e V^^w
f V_y
g
h
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
1 l
1 2 f A 3
(b)
4 f To
4 f To
4 f To
Figure 3.49: The distinctive patterns of random processes characterised by

the function ip(f) (Baiter et al 1982 [3.42])
(a) The Gamma process; (b) The Gaussian process; (c) Periodic process
with A being Gaussian distribution

134 Cavi tation
no obvious regularity in their occurrence, i.e. the value of CTA/A is close or

equal to unity.
Morozov [3.73] has also studied the case where the stochasticity of events
is dominated by periodicity as shown in Figure 3.47(b), that is
0m - mT0 + A m
here A m is a random variable which indicates the deviation of the mth event
from periodicity. Following the same procedure as before, the function ip(f)
is obtained as
1 T= + 0O , v
2 2 s
V(/) = 1 - | U A | + - | U A | £ [f-jr)
here UA is the characteristic function of the random variable A. The power

spectral density consists of a continuous component and a periodically spaced
series of discrete spectral lines located at /To = ± 1 , 2 , 3 , • • •. As an example,
assuming variable A to be a Gaussian distribution,
1 r=+oo . v
which is shown in Figure 3.49(c). As the level of periodicity increases, i.e.

the value of <7A/TQ (here CTA is the standard deviation of A) decreases, the
discrete spectral lines strengthen and the continuous component weakens
particularly for the low-frequency region. --``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
Regularity and Periodicity

The distinction, between the Regularity (as shown in Figures 3.49(a) and (b))
possessed by the processes with a time interval A being a random variable
and the Periodicity (as shown in Figures 3.49(c)) possessed by the process
with the phase deviation A from the exact period To being a random variable,
has been explained by Baiter et al [3.42].
The expectations of the products of two neighbouring time intervals for
these two cases are quite different, compare
o
E{AmAm-i} = A Regularity case ;
E {A m A m _i} = T02 -I- A 2 - A Periodicity case .

For comparison, assuming both the time interval A for the regularity case
and the deviation A for the periodicity case are Gaussian distributions, their
characteristic functions of time intervals are
e_twA—iu-
«A(/) Regularity case ;
2
e-iuToe-(wo-&)
«A(/) Periodicity case ,
hereCTAand a A are standard deviations of time interval A (of the regularity

case) and phase deviation A (of the periodicity case) respectively. In order to
compare these two cases on an equal basis, i.e. equal values of characteristic
functions, u\(f), of their time intervals, let
V
A T0
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
The resultant ip{f) for both cases are shown in Figure 3.50. Since the time
Figure 3.50: Distinction between regularity and periodicity (Baiter et al 1982

[3.42])
intervals of both cases have equal probability densities, W(A) (owing to

the equality of their characteristic functions,UA(/)), so the difference shown
in the comparison solely reflects the different random-natures of these two
sequences. In the regularity case, the peaks reveal a continuously spectral
form at the harmonic frequencies / A = 1,2,3, • • • because the events appear

136 Cavitation
regularly in the sequence, although they are statistically independent of each

other. In the periodicity case, instead of peaks, there are spectral lines at
/ A = 1,2,3, • • • because the events are not statistically independent of each
other (dominated by the distinctive period To) although their appearance is
shifted from the exact period with a random deviation A.
Excess Power
Owing to the overlap of events in the random process, the total power of the
process, y2, differs from that evaluated as if these events were isolated. This
difference is referred_as Excess Power by Baiter et al [3.42].
The total power y2 can be related to the power spectral density 4>{f) by
+oo
v2 = / 4>(f)4f
- 000
J— 0
with
y = nX(0) .
For the sequence of single events statistically independent of each other, after
inserting eqn (3.32), the above reads as
r+oo r+oo UA(f) ^2
y2 = n / | X(f) \2df + 2n | X(f) I Re. df + y
J—oo J-oo 1-«A(/)J
2
Thus, the total power y is composed of two parts. The first term is the
power corresponding to the value computed as if the events were isolated
(i.e. no overlap). The second and third terms are the excess power due to
the event-overlap. Indeed, the excess power comes from the difference that
the square of the sum of the instantaneous amplitudes of overlapped events
is larger than the sum of the squares of amplitudes of individual events
involved.
By assuming a Gamma (r) distribution of time interval A and a expo
nential function of event x(t'), i.e.
(a/A)Q
V
Wr(A) = ^ { A^-^-^/A
T(a)
and
x(t) — ae * ,
the relations of process nature with excess power can be obtained as follows:
--``,```,,``````,,,,``,`,,`,`,

1. For a = 1, i.e. the Poisson distribution which has a finite value of

Wp(A) at A = 0 (see Figure 3.48) indicating the likelihood of overlap,
the excess power reaches maximum, equal to y 2 ;
2. For a > 1, which means a Gamma distribution with Wr(A) = 0 for

A = 0 (Figure 3.48), i.e. 24- < l (since ^ = -4^), the excess power
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
becomes smaller than y2 subject to the value of £ (i.e. the separation
level of events);
3. For j —► oo, i.e. entirely separated events, the excess power equals
zero;
4. For, j -> 0, i.e. the strong overlap case, the excess power becomes
2<7 2
dominant if -—^ > 1.
AT
3.9.4 Spectrum for Sequence of Clusters

It was Argues, 1968 [3.37], who firstly derived the mathematic formula for
the cluster model. Baiter et al [3.41, 3.42] has given a systematic description
of various cluster models.
The mth cluster, r)m, of single events in the cavitation noise, referring to
Figure 3.51(a), is expressed as
Nx(m)
x
Vm(t - 0m) = 2 J m,n(t — 0m — 1?m,/i)
where xmy.
/xth event within the mth cluster;
9m:
occurrence time of mth cluster;
$m,/i-'
relative occurrence time of /xth event to the occurrence
time, 9m, of mth cluster;
Nx(m): number of events, xm, within the mth cluster.
The sequence of clusters of events in time period (0,T) is
y(t,T)= J2vm(t-em)
m=l
here Nc is the number of clusters. Analogously, the power spectral density
<p(f) for the cavitation consisting of clusters can be generally expressed as
</>(/) = UmE ( i £ £ Hm(f)Hn(f)e-ê--eA (3.34)

V m=l n=l )

138 Cavitation
$\l^
& -fyl-
k\J\i\ i\.
(a)
- e fyi
— du-l-H -— X-fi. 8n
2 3....
f|\ fi\ fl\A /P\
ô ô
(M)V
(b) (c)
Figure 3.51: Models for the cluster of single events (Baiter et al 1982 [3.42])
(a) Subcase a: the relative occurrence time d of events is an independent
random variable; (b) Subcase b: the time interval A between the successive
events is an independent random variable; (c) Subcase c: the deviation S of
the event-occurrence time from the periodicity is an independent random
variable
here Hm(f) is the Fourier transform of cluster r)m,

+oo . ,
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
/
Vm(t')e-'ut dt' .
-oo
Basic Models
Baiter applies the same assumptions made for the single events (§3.9.3 "Spec
trum for Sequence of Single Events") to clusters by taking the cluster analo
gously as the quasi-single event. The sequence of clusters can be categorised
as two basic cases: Case A and Case B.
Case A is a sequence of clusters such that the time interval A between
successive clusters is a statistically independent random variable. Its power

spectral density (/>(/) is
4>(/) = ^(fH~F + 2|:HT.Re -^^l)+(nc.tf)2<5(/-0) (3.35)
here n^ is the mean rate of clusters, which can be related to the mean rate of
single events nby n — n^- NQ, if the number of events in clusters is assumed
to take a constant value No.
Case B is a sequence of clusters such that the deviation A of cluster
occurrence from the exact period To is an independent random variable, but
not the time interval. Its power spectral density <f>(f) is
1_ r=+oo
r
4>(f) = 7F{\H I 2 - I uAH \2) + ^ \ uAH | 2 £ S (/ (3.36)
7?
The nature of the spectrum <j)(f) is dominated by the properties of H(f)
and | H(f) | 2 . With some assumptions 33 , the simplified expressions for H(f)
and | H(f) | 2 can be obtained in the forms,
H(f)
NX = A M=l
Nx-lNx-u
H{f)\2 = Nx\X(f)\* + 2\X(f)\2 Yl P(Nx)Re
NX=A 11=1 1/=1
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
where X:
expectation of the Fourier transforms of single events x^
within the cluster;
P(NX): probability of a cluster consisting of Nx single events;
A, B: upper and lower limits of Nx.
The values of exp(—iud^) and exp(—iuj{d^ — ■dll+ll)) in the above expres
sions for H(f) and | H(f) | 2 are determined by the sequence nature of single
events within the cluster. Accordingly, three types of subcases, i.e. Subcases
a, b, c, are further defined by Baiter as follows.
33
They are: (a) the clusters having same number of single events will produce the same
expectation values of the Fourier transforms of single events; (b) the expectation values
of the Fourier transforms for single events within the cluster are independent of Nx; (c)
the random variables related to the event's shape and amplitude are independent of the
variables related to the sequence of events; (d) the values of the shape and amplitude
variables for any single event are independent of those values for other events; and, (e)
as far as the shape and amplitude variables are concerned, the sequence is a stationary
process.

140 Cavitation
Subcase a is for an event sequence in the cluster such that the event
occurrence time t? relative to the cluster occurrence time is an independent
random variable, referring to Figure 3.51(a). This subcase implies that the
noise events are produced by a cavitation cloud such that the bubbles in the
cloud collapse in parallel (collectively) rather than serially. Its expressions
for H(f) and | H(f) | 2 are
H(f) = NxX(f)uM); (3-37)

2 2
I H(f) | = K(\WJ7-\Mf)xU)\ )
+Nl\u^f)X{f)\2 . (3.38)
Subcase b is for an event sequence such that it has independent time
interval A between successive single events, referring to Figure 3.51(b). With
the time relation
*„ = !> i=2
the following expressions are obtained,
B l-uN'(f\
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
H{f) = X(f) £ P(NX) * y ; (3.39)
YHUW = wx ( f W F + \xJT)\2Re [1 "xjfJ{f)])

B '«*(/) ( l - < « ( / ) )
- 2 | X(f) |2 £ P(Nx)Re (3.40)
NX=A (1-"A(/))2
Subcase c is for a periodic event sequence in which the occurrence time

of single event deviates from the exact period To by an independent random
variable S, referring to Figure 3.51(c). With the relation
dfi = (n- 1)T 0 + Sf, ,
the H(f) and | H(/) | 2 expressions can be obtained,
H(f) = us(f)X(f) £ P(NX) _C ; (3.41)

e
NX=A
|ff(/)|2 = Nx (| X(f) I 2 - | u6(f)X(f) |2)
usifWU)?Tvt^-4S W ^ ^sin(wTo/2)
f y ); -(3.42)

Thus, the appropriate combinations of both the basic case for cluster-
sequence and the subcase for event-sequence can be selected, forming a suit
able model to simulate a particular random process of cavitation noise which
consists of a sequence of clusters. For example, Model Ba, which is the com
bination of Case B and Subcase a, is capable of simulating a cavitation noise
process such that its time interval A for the cluster sequence is dominated
by a period 7Q (not an independent variable) whereas the deviation A from
the periodicity is an independent variable; and, within the cluster, the rela
tive occurrence time d for the events is an independent variable (obeying a
Poisson distribution, say).
Properties of Spectra
The properties of spectra for these cluster models are distinguished by the
variations of H(f) and | H(f) | 2 . For investigating such variations, Baiter
et al [3.42] defines a function x(f) by assuming a event-number equality in
clusters (i.e. Nx — NQ = const),
X(f) =
X(f)
which solely reflects the nature of event-sequence within the cluster. For
Subcase a,
X = N0u# ;
2 |„ |2
YF = ô(l-|^|2)+iVo2 U0 I
For Subcase b,
(1-«A) '
"A
= N0(l + 2Re -2Re
.1 ~u\ (1-«A)2
For Subcase c,
{NQUTQ\
sin
X = us- s i n ( ^ )
2
'sin(^p)\
2 2
X| = JV„ ( l - K | ) + | us
sin(^)
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

142 Cavitation
If a Gaussian distribution is assigned to the time variables (i9, A and S), their
characteristic functions become
In order to make the comparison on an equal basis of u#, u\ and 14,5, let
\/2cr,? _CTA_ y/2vs

Aae A ro
here Aae is the equivalent mean time interval for Subcase a,

— _ 2a,,
Aae _
A^ •
The variations of | x | 2 a n d | X I2 f° r Aô = 5 against these three subcases

is shown in Figures 3.52 and 3.53. Here, the nondimensional frequency 3"
is defined as 3" = /A a e for Subcase a, 3" — fX for Subcase b and 3" = fro
for Subcase c. The features of | x | 2 a n d | X I2 f° r these three subcases are
distinctively revealed:
(1) When 3" — 0, there is a maximum, | x | 2 = I X I2 — No > m a ^ t n r e e
subcases;
(2) W h e n 7 ->• 00,
1 _ 12 J O Subcases a and c ;
' X ' "*■ j 1 Subcase b .
(3) When the values of random variables are scattered most widely, (i.e.
Ctf/Aae = 1/V^> CTA/A = 1 and ag/ro = l / \ / 2 ) , the transition process from
the maximum at 7 = 0 to the asymptote is completed well before 3" = 1 for
these three subcases, but the maximum of Subcase a is relatively broader;
(4) When the scattering level reduces by 10 times (referring to the right
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
columns of Figures 3.52 and 3.53), for Subcase a, the maximum at 7 = 0

extends to a higher frequency range owing to the higher probability of event-
overlap, whereas for Subcases b and c, instead of the broadening of the max
imum at 3* — 0, subsequent maxima appear at 3" = 1,2,3, • • • owing to the
higher levels of event-regularity (Subcase b) and event-periodicity (Subcase
c).

l fXae 0 1 2 3 4 fXae
Subcase a
fX 0 l 2 3 4 | fX
Subcase b
25- Ixl 2
20_ 1/J200
o 8 /x 0 =l/J~2 (1
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
10-
1 \„„„/\~ r\
! fx0 0 1 2 3 4 fx0
Subcase c
Figure 3.52: Spectral characteristics (| x | 2 ) of three subcases with N0 = 5

(Baiter et al 1982 [3.42])
The limit frequency / j , (Subcases a and b), which characterises the broad
ening limit of the spectral maximum towards higher frequency, is (Baiter
[3.41])
0.45
fl = Subcase a ;
-l
/, = 0.7 (XN) Subcase b .
The variables a# and A play the similar role in Subcases a and b. There is
an additional factor N for Subcase b. With the increase of N, the maximum*
at / = 0 firstly becomes higher and narrower then eventually reduces to
a spectral line at / = 0 when N —> oo, which actually becomes the case
for the sequence of single events. In other words, it is the formation of

144 Cavitation
l fXae o 1 5 3 4 fX ae
Subcase a
25- Ixl2
20'
1 O g /To=l/»T2
10-
^ No=5»
! fT 0 o l 2 3 4 fx 0
Subcase c
Figure 3.53: Spectral characteristics (| x I2) of three subcases with N0 — 5

(Baiter et al 1982 [3.42])
cluster (i.e. the breakdown of the sequence of single events into such limited
sequences in which the number of single events are finite value) that causes
the transformation of the spectral line at / = 0 into a maximum with a
hump shape and its further broadening towards the higher frequency range
with decreasing N.
The spectral features can be further revealed by examining the spectral
density <j>(f) for models Aa and Ba with the assumption that the random
variables obey a Gaussian distribution 34 . Combining the appropriate formu-
The usage of subcase a and the Gaussian assumption are not necessary, it is only
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`
because the formulae expressions are more concise.

lae derived before, the power spectral density for model Aa is
4>{f) = rHlWW + m\W)\2e^2^f^(^-l-W-Hf)))

+ (nTX(0)) 2 6(f) (3.43)
here, ni = n^N is the total mean rate of single events; and
sinh((?z^Z]i)
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
V>(/) =
cosh
(12*2*JV) _ cos(27rA/)
For model Ba,
4>(f) = wt(\X(f)\H\X(f)\2e-^^2(^-l-Ne-^Â\
+ (mixmife-vra+V £ *(f-f) < 3 - 44 )

here n7 is the total mean rate of single events, n~i = J^-.
For the high-frequency portion, particularly when / —> oo, the spectra
for models Aa and B a have the same asymptote, that is
4>U) -»• iv| x ( / ) |2 / -»■ oo .

Therefore, in general, the high-frequency portion of the spectrum for cluster
cavitation noise can also be approximated with the Campbell theorem.
For the low-frequency portion, the cluster models (here models Aa and
Ba) render distinctive features different from those of single event models.
According to the conventional theorem (i.e. Campbell theorem) the maxima
(if they exist) of the power spectral density <f>(f) originate from the maxima
of the energy spectral density E(f) of single noise events. The first maxi
mum is located approximately at the frequency, fm, equal to the reciprocal
of bubble life-time, Tg, followed by several smaller maxima at higher fre
quencies (refer to §3.9.2 "Spectrum of Single Bubble Collapse" and Figure
3.46). Whereas, the cluster models produce additional maxima in at least
three different ways as specified by the Aa and Ba models.
(1) The formation of a cluster might amplify the maximum which already
exists in the energy spectrum E(f) of a single event at frequency fm. And,

146 Cavitation
the amplified maximum in the power spectral density <j>{f) will locate at the
frequency fmax (fmax ~ Im)- This amplification can be demonstrated by
model Aa. With some assumptions 35 , the <p(f) for model Aa reduces to
AT2
<t>U) * fh I | X(f) |2 + I _ _ i l | X(f) |M . (3.45)
It gives 36 an amplification factor of N2/N. The value of the amplification

factor is gradually reduced with the increase of the frequency range due to
the value-reduction of exp(—(2na^f)2/2). It finally becomes negligible at a
frequency of / « 0.240-J"1. Thus, the high-frequency portion of E(f) for a
single event, such as that featured by the plateau-like extension as shown in
Figure 3.46(b) for Q = 0(0.01}, will not be amplified by the formation of
clusters if 0.24a#l < (2nTB)~1.
(2) The adequate regularity of cluster formation such as rendered by the
model Aa (for say when CTA/A < 0.1) will introduce a maximum at the
frequency of fmax = 1/A. With similar assumptions 37 , the maximum <f>max
at the frequency of fmax — 1/A is approximated as
<Prnax=fHXH^(Â •
Here, niX2(Q) is the value of <j)(f) as if the total single events were Poisson
processes. The amplification factor is N(A/(2TTUA))2-
(3) The periodicity of cluster formation such as rendered by the basic case
B will produce a maximum on the continuous part of the spectrum. This
effect can be demonstrated by model Ba with the single events of Q > 1,
referring to Figure 3.46(c). Similarly, at frequency fmax,
fmax = (27ra A )- 1 f In (l + ( ^ ) J + 1"

N^ _ 1
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
35
T h e assumptions are: (a) <TA/A = 1 so that V'C/) = 1 excluding the influence of
frequency dependence of rp(f) in eqn.(3.43); (b) <T$ is such a small value t h a t a^ < TB
and ezp(—(27rcrtf/) 2 /2) ss 1 for / = / m to exclude the frequency dependence of o#.
36
By taking | X(f) | 2 and | X(f) | 2 to be approximately the same shape and magnitude.
37
They are: (a) the single event is of Q < 0.01, referring to Figure 3.46(b); (b) A > 2TTTB;
and, (c) (70 is so small compared with TB that e~^2",,s^' « 1 to exclude the frequency
dependence.

t h e m a x i m u m of <f>{f) b a s e d o n e q n (3.44) is
4>max=mX2(0)(l+il)
here ft is the amplification factor,
The amplification factor ft and the frequency fmax depend on the ratio 3 8 of
Ctf/Â- For all three maximum-production/amplification cases, the values of
fmax at which the maxima are located are less than the limiting frequencies,
/(. The introduction/amplification of spectral maxima by cluster formation
agrees quite well with the observed phenomenon: The power spectral density
<t>{}) from some cavitations such as the submerged jet and ship propeller
cavitations often has a maximum above the plateau by 10 to 20 dB (as shown
in Figure 3.54) which would be hardly explained by the maximum produced
in the energy spectrum E(f) of single events. Baiter [3.41] further shows
log<)>
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
fmax fj logf
Figure 3.54: Conceptual pattern of some observed cavitation-noise spectra
38
For the low ratio, for say <TO/<7A <C 1, the single events within the clusters are
highly overlapped and the periodicity of clusters is extremely low producing additional
event-overlap probability owing to the increased cluster-overlap. This results in fmax ~
{2iro-&)~1(2\n f ^ - p and ( I K N. For the high ratio, for say o#/o-& > 1, the single events
are fairly scattered within the clusters whereas the high periodicity of cluster formation
further reduces the event-overlap probability brought up by cluster-overlap. In such case,
/ . » . , * 5^7 and n * J V ( S * ) V .

148 Cavitation
that the down slope from the maximum to the plateau obeys an f~b law
where the value of b is governed by the probability distribution, P(N = k),
of event-number N within the cluster,
P(N = k) =
with the following relation,
0 —oo < z < — 3 ;

z+3 - 3 < z < -1 ;
2 - 1 < z < oo .
For some practical cavitation situations, the value of z falls within — 2 < z <
—1, which means a high probability for isolated single events (i.e. TV = 1)
and a low probability for clusters (i.e. N > 1). For such a range of z, b is
valued as 1 ~ 2.
3.9.5 S p e c t r u m for C o m p r e h e n s i v e S e q u e n c e
(Real Cavitation Noise)
Real cavitation noise is nearly always composed of both isolated single events
and clusters. Proper combinations of both models for single-event sequence
and cluster-sequence can simulate various cavitation processes.
The simulation of propeller cavitation noise by a comprehensive sequence
[3.42] is demonstrated. The proposed comprehensive sequence is shown in
Figure 3.55, which is a periodic sequence of clusters with period TQ. The
cluster rj is initiated by a single event xa with time delay A then followed
by a sub-cluster rjt, composed of single events Xf, (Subcase a) with time delay
£. These time parameters are assumed to be random variables. This model
corresponds to a cavitation environment such that at each time when a blade
of a ship's propeller enters the wake flow a sheet of cavitation forms which
produces the single event xa\ then the sheet is detached from the blade and
entrained to downstream in a cloud form, producing the sub-cluster r/b by
its collapse. The use of Subcase a implies a tendency for parallel collapses
of bubbles in the cloud. The period T 0 is the reciprocal of blade-passing
frequency.
Cluster 77 can be written as
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,
v(t -0)= xa(t -6) + m,{t -0-t)

X
-n 1
»■ A -djl-
x
xa b,^
r(~ k$\f$&T '

-To— - To ►— To-
Figure 3.55: A comprehensive sequence model for the cavitation noise from
a ship propeller (Baiter et al 1982 [3.42])
and its Fourier transform H(/) is

H(f)=Xa(f)+e-^Hb(f)
here Xa(f) and Hb(f) are Fourier transforms of xa and r)b respectively; 8 is
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
the occurrence time of cluster. With some assumptions39, the expressions
for | H | 2 and | H | 2 can be obtained,
H\2 | Xa | 2 + | u^Hb | 2 +Xau\Hb* + Xa\Hb ; (3.46)

jTTj* = p f 7 P + r ^ + û?J^+^Cii««6- (3.47)
Here, u^ is the characteristic function of probability density, WX£), of time
delay £,
u{= J—oo
/ °V^W(£K.
Substituting | tf | 2 and | # | 2 into eqn (3.36) (Case B referring to §3.9.4),
the power spectral density, </>(/), for this model can be obtained.
To specify the model, the single event xa is assumed to be the form of
%a(t) — -a—e
2(' M
39
They are the same as those made for Xa(f) and Ht,{f) in §3.9.3 and §3.9.4 plus the
assumption of £ being an independent random variable.

150 Cavitation
with its Fourier transform
y m .v^ 2 -(y) 2
Since the single events xt, in the sub-cluster r)b are simulated by Subcase a,
the expressions of i/j and | i/^ | 2 can be evaluated according to eqns (3.37)
and (3.38). The single events in the sub-cluster are expressed as
T
xb(t') = be >>
with its Fourier transform
Xb(f) = b—^- .
1 + 2a;r(,
With selected parameter values and some simplifications 40 , the power spec
tral density cf>(f) is obtained (Figure 3.56). 4>(f) (which is multiplied by
To/(brt,)2) is composed of both a continuous part indicated by the solid line
and discrete spectral lines.
The pattern of the continuous part of <f>(f) resembles the experimentally
observed pattern quite well, referring to Figure 4.56. It is characterised
by a steep initial rise at low frequency followed by a broad maximum of
hump-form, in which there are several small peaks, then along the downward
slope dropping onto a plateau and finally rolling off at high frequency. The
dashed line indicates the pattern that the spectral density would have if the
cavitation noise were only composed of the sub-cluster events Xb (therefore,
the sub-cluster rn, renders the basic features of <f>{f))- If the event xa were
taken into account but its correlation with sub-cluster neglected, the curve
would be modified to the pattern indicated by the dotted line. The difference
between the dotted line and solid line just represents the correlation between
the event xa with the sub-cluster TJ^1. The roll-off frequency at point B is
40
The simplifications are: (a) Both the single event xa and the events xt, are uniform, i.e.
the amplitudes a, 6 and time parameters T 0 , U are all constant values; (b) The number of
events contained in sub-clusters is constant value ./Vo; (c) The random variables £, A and
d are all Gaussian distributions. The rather arbitrarily selected values of parameters are:
| = 10- 3 , N0 = 10, ( = 2 x 1 0 - \ i = 8 x 10- 2 , f£ = 10- 5 , % = 2 x 10" 2 , ft =
2 x 10~2, £*■ = 2 x 10 - 3 .
41 °
The role of event xa is producing a peak E on the dotted line (the frequency at which
peak E appears depends on ra andCTA).However, this peak E fades away and, instead,
two peaks F and G emerge in the resultant spectrum owing to the correlation between xa
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,
and rj6 subject to the selection of their parameters. The frequencies corresponding to the
peaks F and G are determined by f, T„, T),, <TA and <7{.

<D To F
10 -
7\7
(bXh )
•'Is ' ^\ 1
c 10 log No B
\
A 1 - •'/ 1 *
dB
.7
7 1/
.■/ /
\ - A
i '
i/^ Hump Plateau
10 fTo
1 3 4 10 2 3 4 5
<D To 2 ■
10 10 10 10
fTo=l \ 5
2 10 ^-6
il
(bTb )
No=10
0 - Hight
of line
dB
atf=0
1
' 1 2 3 4
i fTo
10 10 10 10
Figure 3.56: Power spectral density </>(/) of a comprehensive sequence model

for the cavitation noise from a ship propeller (Baiter et al 1982 [3.42])
/ * = 1/(27TT(,) and slope A obeys an f~2 law, which agrees with theoretical
studies and experimental investigations. If the events Xf, in the sub-cluster
were a Poisson process (which belong to Subcase b, i.e. the events Xj, were
more scattered than Subcase a) the plateau would extend all the way down to
zero frequency instead of the hump. The hump (maximum) rises above the
plateau by lOlogNodB. The frequency, fmax, corresponding to the hump
summit D is
ln[l + * M
Jmax - 2?r
N Â+ C T |
and the frequency corresponding to point C which characterises the hump

broadening, i.e. the limiting frequency /;, is given by
1
fl =
\Fi-KO-l,
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

152 Cavitation
The pattern of discrete spectral lines indicates that the cavitation noise
has strong periodicity. Its features depend on the natures of event xa and
sub-cluster rjb and their correlation. The discrete value of <j>(f) at / = 0 is
valued as {brbNo/To)26(f — 0) which corresponds to 102 dB for the line at
/ = 0 (Figure 3.56).
Through this example, it is evident that: (a) it is possible to simulate real
cavitation noise by using a properly configurated comprehensive sequence
model; and reversely, (b) the mechanism by which the cavitation noise is
generated can be investigated by analysing the features of the observed cav
itation noise spectrum.
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

References
3.36 Arakeri, V. H. and Acosta, A. (1981), 'Viscous Effects in the Incep
tion of Cavitation', ASME J. Fluid Eng. 103, pp280-287.
3.37 Argues, P. Y. (1968), 'Sur Certains Probleraes Statistiques lies a

l'effect de Barkhausen', Journal de Physique 29, pp369-384.
3.38 Arndt, R. E. A. (1981), 'Cavitation in Fluid Machinery and Hydraulic

Structures', Ann. Rev. Fluid Mech., 13, pp273-328.
3.39 Arndt R. E. A. and George, W. K. (1979), 'Pressure Fields and Cavi

tation in Turbulent Shear Flows', 12th Symp. Naval Hydrodynamics,
Washington D. C , pp327-339.
3.40 Arndt, R. and Ippen, A. T. (1967), 'Cavitation near Surfaces of Dis

tributed Roughness', MIT Hydrodyn. Lab. Rept. 104.
3.41 Baiter, H. J. (1986), 'On Different Notions of Cavitation Noise and

What They Imply', Int. Symp. on Cavitation and Multiphase Flow,
ASME FED 45, p P 107-118.
3.42 Baiter, H. J., Gruneis, F. and Tilmann, P. (1982), 'An Extended Base
for the Statistical Description of Cavitation Noise', ASME Int. Symp.
on Cavitation Noise, pp93-108.
3.43 Benjamin, T. B. and Ellis, A. T. (1966), 'The Collapse of Cavitation

Bubbles and the Pressure Thereby Produced against Solid Bound
aries', Phi. Trans. R. Soc. London, A260, pp221-240.
3.44 Blake, J. R. (1988), 'The Kelvin Impulse: Application to Cavitation

Bubble Dynamics', J. Austral. Math. Soc, Ser.b.30, ppl27-146.
3.45 Blake, J. R. and Gibson, D. C. (1987), 'Cavitation Bubbles Near

Boundaries', Ann. Rev. Fluid Mech. 19, pp99-123.
3.46 Blake, W. K. and Sevik, M. M. (1982), 'Recent Developments in

Cavitation Noise Research', ASME Int. Symp. on Cavitation Noise,
ppl-10.
3.47 Brophy, M. C , Stinebring, D. R. and Billet, M. L. (1985), 'An Anal

ysis of Pump Cavitation Damage', J. Propulsion, Vol.1, No.6, pp441-
448.
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

154 Cavitation
3.48 Chahine, G. L. (1983), 'Cloud Cavitation: Theory', 14th Symp. on

Naval Hydrodynamics, Ann Arbor, National Academy Press, Wash
ington, D. C. ppl65-195.
3.49 Chahine, G. L. (1990), 'Numerical Modelling of the Dynamic Be

haviour of Bubbles in Nonuniform Flow Fields', ASME FED 91,
pp57-64.
3.50 Chahine, G. L. (1990), 'Nonspherical Bubble Dynamics in a Line

Vortex', ASME Cavitation and Multiphase Flow Forum, Toronto,
Canada, ppl21-126.
3.51 Chahine, G. L. and Duraiswami, R. (1992), 'Dynamical Interactions

in a Multi-bubble Cloud', ASME J. Fluid Eng. 114, pp680-686.
3.52 Daily, J. W. and Johnson, V. E. (1956), "Turbulence and Boundary-

Layer Effects on Cavitation Inception From Gas Nuclei', Transactions
ASME 78, PP 1695-1706.
3.53 Dear, J. P. and Field, J. E. (1988), 'A study of the Collapse of Arrays
of Cavities', J. Fluid Mech. 190, pp409-425.
3.54 Delius, M., Miiller, M., Vogel, A. and Brendel, W. (1988), 'Shock
Waves and Cavitation', in Biliary Lithotripsy, Chicago: Yearbook.
3.55 Fitzpatrick, H. M. and Strasberg, M. (1956), 'Hydrodynamic Source

of Sound', 1st Symp. on Naval Hydrodynamics, Washington D. C ,
pp241-280.
3.56 France, J. P. and Michel, J. M. (1983), 'Two- and Three-dimensional

Rotational Structures in the Cavitating Turbulent Wake of a Wedge',
4th Symp. on Turbulent Shear Flows, Karlsruhe, Germany.
3.57 Fujikawa, S. (1985), 'Interaction between Two Slightly Nonspherical

Bubbles in a Compressible Liquid: Part 1 Mathematical Formula
tion', ASME Symp. Cavitation in Hydraulic Structures and Turbo-
machinery, FED 25, ppl67-172.
3.58 Fujikawa, S. (1986), 'Interactions between Two Slightly Nonspherical

Bubbles in a Compressible Liquid', Int. Symp. on Cavitation, Sendai,
Japan, pp55-60.
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

3.59 Fujikawa, S. (1988), 'Dynamics of Two Nonspherical Cavitation Bub

bles in Liquids', Fluid Dynamics Research 4, ppl79-194.
3.60 Gates, E. M. and Acosta, A. (1979),'Some Effects of Several

Freestream Factors on Cavitation Inception of Axisymmetric Bod
ies', 12th Symp. Naval Hydrodynamics, Washington D. C. pp86-112.
3.61 Hansson, I. and M0ch, A. (1980), 'The Dynamics of Cavity Clusters

in Ultrasonic (vibratory) Cavitation Erosion', J. Appl. Phys. 51(9),
PP 4651-4658.
3.62 Huang, T.T. (1981), 'Cavitation Inception Observations on Six Ax

isymmetric Headforms', ASME J. Fluid Eng. 103, pp273-279.
3.63 Il'ichev, V. I. and Lesunovskii, V. P. (1963), 'On the Noise Spectra

Associated with Hydrodynamic Cavitation', Akust. Zh. 91, pp32-36.
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
3.64 Kling, G. L. and Hammitt, F. G. (1972), 'A Photographic Study

of Spark-induced Cavitation Bubble Collapse', Trans. ASME D: J.
Basic Engineering 94, pp825-833.
3.65 Knapp, R. T., Daily, J. M. and Hammit, F. G. (1970), Cavitation,

McGraw-Hill, New York, ppl34-141.
3.66 Lauternborn, W. (1982), 'Cavitation Bubble Dynamics - New Tools

for an Intricate Problem', Appl. Sci. Res. 38, ppl65-178.
3.67 Li, S. C. (1988), 'Statistical Characteristics of Cavitation Bubbles in

Cavitating Transient Flows', Report to the Ministry of Water Re
sources, China.
3.68 Li, S. C. (1992), 'Pressure Fluctuations in Cavitating Draft-tube

Flows', ASME FED 136, ppl-6.
3.69 Li, S. C , Zhang, Y. J. and Hammitt, F. G. (1983), 'Low-frequency

Fluctuations Associated with Cavitating Venturi Flows', DRDA
Rept. No.UMICH 014571-64-1, University of Michigan, Ann Arbor,
USA.
3.70 Li, S. C , Zhang, Y. J. and Hammitt, F. G. (1986), 'Statistical Con

siderations of Pressure Pulses from a Cavitating Venturi', Proc. Instn.
Mech. Engrs. 200, N0.C6, pp381-387.

156 Cavitation
3.71 Mellen, R. H. (1954), 'Ultrasonic Spectrum of Cavitation Noise in

Water', J. Acoust. Soc. Am. 26, pp356-360.
3.72 Meyer, R. S., Billet, M. L. and Holl, J. W. (1992), 'Freestream Nuclei
and Travelling Bubble Cavitation', ASME J. Fluid Eng. 114, pp672-
679.
3.73 Morozov, V. P. (1969), 'Cavitation Noise as a Train of Sound Pulses
Generated at Random Times', Soviet Physics-Acoustics 14, No.13,
pp361-365.
3.74 O'hern, T. J. (1990), 'An Experimental Investigation of Turbulent
Shear Flow Cavitation', J. Fluid Mech. 215, pp365-391.
3.75 Prosperetti, A. (1993), 'Bubble Mechanics: Luminescence, Noise and
Two-phase Flow', Theoretical and Applied Mechanics, Elsevier Sci
ence Publishers B. V. pp355-369.
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
3.76 Ran, B. and Katz, J. (1993), 'Pressure Fluctuation and their Effect on
Cavitation Inception within Water Jet', ASME FED 153, ppl57-164.
3.77 Shima, A. et al (1989), 'The Growth and Collapse of Cavitation Bub
bles Near Composite Surface', J. Fluid Mech. 203, ppl99-214.
3.78 Tomita, Y. and Shima, A. (1986), 'Mechanism of Impulsive Pressure

Mech. 169, pp535-564.
3.79 Tomita, Y. and Shima, A. (1986), 'Mechanisms of Impulsive Pres

sure Generation and Damage Pit Formation by Bubble-Shock Wave
Interaction', Int. Symp. on Cavitation, Sendai, Japan, pp77-82.
3.80 Vogel, A., Lauterborn, W. and Timm, R. (1989), 'Optical and Acous
tic Investigations of the Dynamics of Laser-produced Cavitation Bub
bles near a Solid Boundary', J. Fluid Mech. 206, pp299-338.
3.81 Young, J. O. and Holl, J. W. (1966), 'Effects of Cavitation on Periodic

Wakes behind Symmetric Wedges', ASME J. Basic Engineering, 88,
ppl63-176.
3.82 Zhang, S. and Duncan, J. H. (1993), 'Numerical Simulations of Ex
periments on the Interaction of a Cavitation Bubble and a Compliant
Wall', ASME FED 153, ppl37-142.

Chapter 4
Cavitating Flow
4.1 General Features
H. Murai
Among the main effects (mainly negative) caused by cavitation occurring
in hydraulic machinery are (i) change in the performance characteristics, (ii)
generation of vibration and noise and (iii) damage. This chapter is concerned
mainly with (i).
Although there are various types of cavitation as tabulated in §2.1, those
having large influences on the flow field are sheet, travelling bubble and
vortical cavitations. The cavitation type depends on the number density of
cavitation nuclei, higher density favouring sheet, lower favouring bubble, a
moderate density sometimes giving rise to both types coexisting ( see Fig.
4.1(b)).
4.1.1 Sheet Cavitation

Sheet cavitation occurs at the region of re-attachment (where the tempo
ral pressure is minimum [4.12]) when a laminar boundary layer separates,
and develops upstream in the separation bubble and fills it when the cavi
tation number is further reduced, becoming the bubble-ring or bubble-line
cavitation.
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
At the initial state of occurrence of fixed sheet cavitation, the cavity is

white and opaque. As the cavity grows, with decreasing cavitation number,
it becomes clear (a glassy part developed in the cavity) from the upstream
end until the downstream end where it becomes white due to the disturbance
of flow.
When the cavity has grown to a moderate length, about half the length
157
158 Cavitation
11.4 m/s
Figure 4.1: Travelling bubble cavitation on hydro-foil Clark Y 11.7 at a = 4°

(a) still photograph (left); (b) moment photograph (right).
of the cavitating body, the velocity of liquid adjacent to the cavity is nearly
(1 + CT)'1/2) times the free stream velocity at the front part of the cavity,
because the pressure in the cavity is nearly the vapour pressure. In the rear
part of the cavity, the surface gradually grows to be a bubbly two-phase
shear layer, enlarging the wake. At the downstream end of the cavity, the
stream adjacent to the cavity surface rejoins the guiding surface, and a part
of this stream flows back underneath the cavity along the guiding surface
[4.21]. Depending upon the magnitude of the momentum of this re-entrant
flow, it reaches part or all the way back to the upstream end of cavity, and
the flow as it was before the onset of cavitation or the flow with the cavity
of cut off length is recovered. The remaining cavity between the cavity
surface and the re-entrant flow is rolled up, forming a vortical cavity which
is transported downstream as cloud cavitation (Table 2.1) and contributes
to cavitation erosion. Immediately after the flow recovery, the new cavity
begins growing again. As the cycle of growth and shrinking is repeated,
the sheet cavitation is unsteady (until it grows so large that the re-entrant
flow has little momentum and has little effect on the cavity and the main
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

Cavitating Flow 159
flow, which is called supercavitation) and causes the oscillating flow field
around the guiding surface and the oscillating force on the cavitating body.
This unsteady state is called transient cavitation. In certain conditions,
depending on the free stream velocity, the size or the configuration of the
obstacle at the downstream end of cavity, the cycle is not clear or almost
non-existent.
The static pressure in the sheet-cavitation cavity is a little higher than
the vapour pressure, because of the partial pressure of un-dissolved gas in the
cavity and the momentum of the re-entrant flow. The larger this difference
is, the larger the cavitation number is. Fig. 4.2 shows the variation of cavity
pressure on the NACA 4412 profile, in the form of cavitation number ak,
by taking the cavity pressure to be the vapour pressure, with the cavitation
number [4.10] .
1.2
NACA 0 a
4412 a P • r
1.0
a •°
•
<7 = cr K .
0.8
/ *£
0.6
ATTACK ANGLE VELOCITY
/ <P • m 16° 30 FPS
0.4 a » 12° 30 FPS
^ Q JO & = 8' 30 FPS
o = 4* 30 FPS
0-2 4' 45 FPS
0
0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0
Figure 4.2: Variation of cavity pressure with cavitation number [4.10]
As a cavity of sheet cavitation occupies a space on the guiding surface (or

the obstacle) in the flow filled with gases or gases mixed with liquid phase
at almost fixed (vapour) pressure, the pressure distribution on the guiding
surface is changed not only within but around the cavity (compared to the
non-cavitating state) because of the change of flow field. This changes the
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

160 Cavitation
lateral force (lift) and the profile drag on the obstacle. In addition, the
presence of the cavity affects the behaviour of the boundary layer on the
surface, not only changing the skin friction but also moving the position of
the separation point to cause a change of flow field around, and forces, on
the obstacle. Furthermore, the cavity transported downstream spreads the
region of flow separation and the thickness of wake, increasing a hydraulic
loss.
4.1.2 Travelling Bubble Cavitation

Travelling bubble cavitation occurs sometimes separately from other types
of cavitation (see Fig. 4.1(b)) or sometimes along the cavity surface of a
sheet cavitation (see Fig. 4.3(b)). In either case, it has little effect on the
flow field, unless it is intensively localised, when the effect will be similar to
that of sheet cavitation of the same intensity.
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
11.4 m/s
Figure 4.3: Fixed sheet cavitation on hydro-foil Og at a = 5°. (a) still

photograph (left); (b) moment photograph (right).
Travelling bubbles transported downstream along the guiding surface, or

Cavitating Flow 161
the cavity surface of a sheet cavitation, contribute to the cavitation noise

and damage, as detailed in §3 and §6.
4.1.3 Vortical Cavitation

Vortical cavitation can occur in a vortex core, in a flow around a sharp-edged
corner, in a trailing vortex street, in a shear layer, in a wake etc. Only the
vortical cavitation which occurs near the rigid boundary, such as the tip
vortex cavitation, will affect the pressure distribution on the surface.
The occurrence of vortical cavitation in a two dimensional trailing vortex
street will change the shedding frequency of the vortex street, depending on
its intensity. Fig. 4.4 shows the variation of Strouhal number of the shed
ding frequency by the normalised cavitation number of vortical cavitations
occurring in the vortex street behind the wedges measured by Young and
Holl [4.54], and Fig. 4.5 behind the circular cylinder by Shalnev [4.47]. The
Strouhal number is at first unchanged, then increases to a maximum and
decreases to approach the state of supercavitation.
On a body with a sharp corner, cavitation occurs at the corner. Whereas,
in a flow around the blunt-nosed body without a sharp corner, cavitation
often occurs in the cores of vortices in the shear layer of its wake. In both
cases, vortical cavitations behave as described above (see Fig. 4.6).
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
4.1.4 Hydraulic Loss Caused by Cavitation

The cavitation caused hydraulic loss will be described by investigating the
variation of the drag and pressure coefficients, ie. CDP and Cp, on a circular
cylinder caused by cavitation. CDP does not initially increase at the incep
tion of cavitation. Its variation depends on whether the Reynolds number is
greater or less than the critical Reynolds number Rec (Figs. 4.7 and 4.8(a)
and (b)) [4.17] .
If Re > Rec, CDP initially decreases and then increases gradually to
reach a maximum, and after a gradual decreasing it finally decreases linearly
as the cavitation further develops. If Re < Rec, in contrast, CDP, after
a slight increase following the cavitation occurrence, jumps abruptly to a
maximum larger than the former one, and then decreases in a similar way to
the former case. The critical cavitation number at which the abrupt rise of
CDP takes place decreases with the increase of the turbulence in the uniform
flow.
Variations of CDP in these two cases can be explained as follows with

162 Cavitation
JJ[>=V2>/v=2.0*10 9
c
0.S
fW/V e = i5
030 ■
02* ■
02*
0» ■
0.22 ■
coo - 0 = 30°
02» *
0.26 - ■
1
024 • ••*. • v
022
* . *«
■
O20
*
1 .1 i i i _, i i i
Olt
«*
e =eo e
0i22
0.20
018 ■
0I«
OW
0.22 0 = 90°
OM
018
01*
014
Of Ox
ft*
0> Q2 OJ 0.4 0.5 OS R7 0« 09 1.0
Figure 4.4: Variation of Strouhal number of shedding frequency with nor

malised cavitation number [4.54].
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

Cavitating Flow 163
0.30 3.0
X
S r
s o X
0.22 \ N. 2.2
\
< > 3
> \. 0
0.14 9 1.4
10 1.4 1.8 2.2 (J 2.6
Figure 4.5: Variation of Strouhal number of shedding frequency with cavi

tation number, circular cylinder [4.47].
(a) (b)
a=1.33
0 = 0.9
Figure 4.6: Cavitation on circular cylinder, (a) still photograph (left); (b)
moment photograph (right).
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

164 Cavitation
orlt
*Ro r It
Figure 4.7: Variation of drag coefficient on circular cylinder with cavitation

number and Reynolds number [4.17].
J?=12*10» ^=2*10° to **
• **j ■ /*«
• /M * <4« • «4 • rn
OS i-4 .Rvr\x.
• r« • /<*#
a 7P» ♦ <*> • m • rtt
KN*2 « I • l I i > i
—%Z£* ^ U ♦
^
&£ rflT " fS « '/it
•en; :<!
«C to 1 iff H& lier
« ivf
C P = (P- P0)/(l/2pV0) a
e
(a) (b)
Figure 4.8: Variation of pressure coefficient on circular cylinder with cavita

tion number [4.17]. (a) Re < Rcrit', (b)i? e > Rcrit-
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

Cavitating Flow 165
reference to the distributions of Cp in Fig. 4.7(a) and (b). For the flow of
Re > Rec, when the cavity is so small that its closure end re-attaches to the
cylinder surface, the frictional loss in the flow adjacent to the cavity surface
is smaller than that when the cavity surface was in contact with the cylinder
surface in the non-cavitating state, and the re-entrant jet at the cavity closure
gives a considerable momentum toward upstream to recover the pressure and
reduce CDP- Even after a cavity has developed beyond the cylinder surface,
the pressure recovery by the re-entrant jet is so maintained in the vicinity
of the tail end to mitigate the increase of CDP until its maximum, when the
pressure on the cylinder surface in the cavity becomes uniform. The flow
of Re < Rec causes a laminar separation and the pressure recovery on the
downstream surface is small in the non-cavitating flow, therefore, the CDP
difference caused by the cavitating flow is not much while the cavity is small
because the pressure recovery by the momentum of the re-entrant jet is also
small (owing to the large cavity pressure). Nevertheless, the cavity pressure
becomes uniform as soon as the cavity has developed beyond the cylinder
surface, resulting in the abrupt increase of CDP to a maximum.
From the above, it can be inferred that the hydraulic loss contributed by
cavitation mainly derives from the enlargement of the region of flow separa
tion and the thickness of wake caused by the transport of the mass of gas
phase at the grown stage of cavitation, although cavitation is also accompa
nied by energy dissipation caused by the phase changes, and the isolation and
dissolution between liquid and gas, and the momentum exchanges between
gas and liquid phases, and the solid wall.
--``,```,,``````,,,,``,`,,`,`,``-

166 Cavitation
4.2 Hydrofoil a n d Hydrofoil Cascade
H. Murai
4.2.1 Cavitation Characteristics of Hydrofoil

Aspect of Occurrence
The position and length of the cavitating zone, and the type of cavitation on
a hydrofoil depend on its profile form and attack angle, and on the cavitation
number. Figs. 4.9 and 4.10 show the aspects of occurrence on two hydrofoils,
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
Clark Y 8 % [4.36] and the ogival 8 % [4.30] , as examples of hydrofoils with
two different profile forms (ie. the airfoil and the thin-front types). In the
left hand parts of the figures, Xo/l and X/l (I is the chord length) show the
chord-wise positions (indicated on the abscissa) of upstream and downstream
(closure) edges of incipient cavitation for the attack angle (indicated on the
ordinate) in Zones I, III and II (ie. the vicinity of leading edge on the upper
surface, the rear portion on the upper surface, and the vicinity of leading
edge on the lower surface, respectively). The right hand parts of figures
show the variation of the chordwise position of the closure edge, X/l, of
cavity for a given attack angle (indicated on the ordinate) with the variation
of the cavitation number (indicated on the abscissa) in Zones I, II and III,
respectively.
On C.Y.8 the cavitating zone transfers from Zone I to Zone III with
the increase of attack angle, and two types of cavitation, bubble and sheet
are mixed though there are more bubbles at a relatively small attack angle.
On 08, however, there is a lower limit to the attack angle, at an angle
smaller than which Zone I never cavitates and Zone III begins cavitating
at a relatively small cavitation number. Once the attack angle exceeds the
limit, Zone I begins cavitating initially with sheet type, and then Zone III
begins cavitating with bubble type, as the cavitation number decreases.
The position of the closure edge of the cavity indicated in Figs. 4.9 and
4.10 is the time-averaged value. The cavity length oscillates in a transient
range of cavitation number, from soon after inception to approaching su-
percavitation, and the oscillating force acts on the hydrofoil in the range,
especially violently while the time-averaged closure edge lies between 60-70%
and 110-120 % chord, when the instantaneous cavity closure edge crosses over
the trailing edge and the re-entrant jet hits the lower surface of the hydro
foil, in which case the circulation is large enough to let the re-entrant jet

Cavitating Flow 167
Figure 4.9: Variation of front and rear edges, and length of cavitation with
cavitation number and attack angle on hydrofoil Clark Y 8 [4.36]. tw: water
temperature; a/a3: air content R = V l / 7 .
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
Figure 4.10: Variation of front and rear edges, and length of cavitation with
cavitation number and attack angle on hydrofoil 0& [4.30].

168 Cavitation
reach there around the trailing edge. The range of the violent oscillation is
indicated in the figures.
Performance of Hydrofoil
The following describes the variation of lift and drag coefficients with cavi
tation. The lift coefficient, CL, is initially unchanged when the inception of
cavitation in Zone I or II takes place (Fig. 4.11).
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
Figure 4.11: Variation of lift coefficient with cavitation number and attack
angle on hydrofoil Clark Y 8 [4.35].
At a relatively large attack angle, as cavitation number decreases further,

the variation of lift coefficient will mostly tend to show a maximum after a
slight increase, and then begin to decrease (gradually at first then more or
less linearly). This increase will be larger if the thickness ratio of hydrofoil
(especially its front part) is smaller. However, an airfoil type with relative
thickness hardly show such an increase.
At a relatively small attack angle, for an airfoil type, there tends to be
no initial increase. On a hydrofoil with a thin front (eg. the ogival or airfoil
type of small thickness ratio), the lift coefficient begins decreasing soon after

Cavitating Flow 169
the cavitation occurs at Zone III.

When the cavity develops beyond 140 % chord the lift coefficient begins
decreasing almost linearly, on a hydrofoil of any profile type.
The increase of lift coefficient by cavitation occurrence has been shown
to be caused by the increase of effective camber or the area increase of the
low pressure corresponding to the cavity (vapour) pressure in the vicinity of
the leading edge on the upper surface, despite the lack of a peak pressure
there (Fig. 4.12).
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
Figure 4.12: Increase of lift by cavitation occurrence.
The drag coefficient, CD, begins increasing with decreasing cavitation

number after the cavity closure edge has reached 10-25% chord, when the
cavitation occurs in Zone I, at almost the same time as the beginning of
increase of the lift coefficient (Fig. 4.13). When the cavitation occurs in
Zone III, the drag coefficient begins increasing shortly after its occurrence.
The drag coefficient of some hydrofoils with a thin front like the ogival profile,
however, at first decreases and then increases with the decrease of cavitation
number, at a suitable attack angle.
The drag coefficient has a maximum of about twice the non-cavitating
value when the cavity closure edge reaches 110-120 % chord. With the fur
ther decrease of cavitation number, the drag coefficient decreases gradually
at first and then linearly after the cavity closure edge has reached about
140% chord, at the same time as the linear decrease of lift coefficient.
The variation of drag coefficient with decreasing cavitation number (in
cluding the drag reduction of a hydrofoil with the thin front) can be explained
by the mechanism described in §4.1.4, apart from the drag reduction caused

170 Cavitation
1 1111 1 I
1
i Profile Clark Y 8 r 0
w c UrlU-lWC — . y
M f*
I
M
/
J/ ■^SrmfV/wm V\
jt s _sL
- - © - • -?*
m f-2m
s -J,
\|
w
—
kl
n a
'0.1 OS ID IS 7.0 IS
Figure 4.13: Variation of drag coefficient with cavitation number and attack
angle on hydrofoil Clark Y 8 [4.35].
by the reduction of friction on the surface in contact with the cavity. The
increased region of low pressure inclined forward to the free stream in the
front part of chord also contributes to the drag reduction.
This drag reduction, together with the lift augmentation described above,
sometimes brings a performance improvement to hydraulic machinery. This
is especially so for a centrifugal impeller with backward blades, as in addition
to the effects described above, the cavity on the suction surface of blade is
pushed towards the surface by the centrifugal force on the water adjacent to
the cavity to make the cavity and its wake so thin that the drag is reduced
further.
4.2.2 Cavitation Characteristics of Conventional Airfoil Sec

tions
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
The hydrodynamic data, the aspects of cavitation occurrence, and the per
formance characteristics under cavitating and non^cavitating conditions of
the airfoil sections of Clark Y, Clark YH, RAF 6 [4.36], Munk 6, NACA
23012 [4.29], NACA 4412 and NACA 661-012 [4.19] are available.
Among them, the thickness ratios of the Clark YH and RAF 6 profiles
include 11.7 % and 6 %, and those of the Clark Y profile include every 2 %

Cavitating Flow 171
from 4 % through 14 % (for details see [4.35]). Figs. 4.14(a), (b) and (c),
and 4.15(a), (b) and (c) show the examples.
On a Clark Y profile with a thickness ratio > 6 %, there exists the
transference of cavitating zone between Zones I and III with the change of
attack angle, but for thinner than 6 % there is a lower limit attack angle for
the Zone I cavitation to develop beyond a small certain limiting length, 5-10
% chord. The range of attack angle between that for the transference (from
Zone I cavitation to Zone III or the above limit) and that at which Zones
II and III begin cavitating simultaneously at a cavitation number 1 will be
wider if the thickness ratio is larger. But the cavitation number at which
Zone III begins cavitating will be smaller if the thickness ratio is smaller.
That is to say that in respect of cavitation occurrence, the thinner profile is
more sensitive to the change of attack angle.
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
The range of attack angle for strong oscillation will be wider if the thick
ness is smaller. The 14 % section does not oscillate strongly in the range of
attack angle between —3° and 5°.
From the comparisons on cavitation occurrence and development as made
above for 11.7 % and 6 % thickness sections of Clark Y, Clark YH and RAF
6, the relative superiority of the Clark Y profile, in spite of its relatively
simple profile, is noticed.
The variations of the lift and drag coefficients of hydrofoil type, including
Clark Y, due to the cavitation occurrence and development are qualitatively
identical to those described in §4.2.1 on the airfoil type. The dependence
on thickness of the variations of the lift and drag coefficients with cavitation
number of Clark Y profile is also qualitatively identical to the dependence
on the thickness of the airfoil type described in §4.2.1. Fig. 4.16 shows the
variation of the polar diagram with the thicknesses of Clark Y profile at
different cavitation numbers. It shows the relative superiority of the Clark
Y 8% profile at a small cavitation number.
4.2.3 Cavitation Characteristics of Ogival Hydrofoils

The hydrodynamic data of ogival and its modified profiles have been pub
lished [4.39].
Because a hydrofoil section in this group has the sharp leading edge and
a thin front part, it has a lower limit attack angle for a cavity in Zone I to
grow as shown in Fig. 4.10. The lower limit attack angle will be larger if the
1
In this range the attack angle for maximum lift-drag ratio is included.

172 Cavitation
Profile Clark Y 14
t,-/U-/ift
-1 1 —i
a Prauis UATK i 14
«-
l\1
n %*m • -* r-
Sf ?j
if
i i * / ^)z/kma * i-wr at - r,r
,' <r
u IriU . .-/,•
n rnfc f
C
-•/•_
rvr
a ~/**t*t r-ifltf1l l-2mrOut
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
i cr
\*f '.' IS t p
1
(b )
w ~I
ro
c S V P file Clark Y 14
/ V tw-m-an — a -r
/ V %<n — • r
- * ■ • /
/ ** N R*(VrU)'Uf
W
,Y V\
sv v
L —
— • -r.
• -/•.
S i \ ■,
N "v
^ \
\
k \ s•s. 5» i ]
1
yj j
<> I
HI?
s N
t a
9S III IS V IS
(c)
Figure 4.14: Cavitation characteristics of hydrofoil Clark Y 14 [4.35]. (a)

position and length of cavitation; (b) CL versus a; (c) Co versus a.

Cavitating Flow 173
Profile Omrk Y 4
/*, i.-iu-ivi
tfU!
*'<u-unr
(a)
1 11 ?
1 1 1
L-W-W'C
PrafUa Clark Y 4
w a - !•
r if"((H(M/,J <-
u 1 1 1
r
u /£r .i/l-'lmtM
*]Mn L-i—'
. f
12
1
r
•/•
— j tlM . Ja*aor «" AAr dtf'
^ 3bj»««•r' f.
r™
^d^" J1 M W _i 2\
(b )
I I I I I I T"
Profile Clark Y 4
Co. tr-IVHWC
— • ;•
(c)
Figure 4.15: Cavitation characteristics of hydrofoil Clark Y 4 [4.35]. (a)

position and length of cavitation; (b) CL versus a; (c) Co versus a.
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

174 Cavitation
Figure 4.16: Variation of polar diagram with cavitation number and thick
ness on Clark Y hydrofoils [4.35].
thickness ratio is larger. For medium and large thickness sections, there is
a cavitation-free range of attack angle between the Zone I lower limit angle
and the the Zone II upper limit angle. The cavitation occurrence number at
Zone III on a hydrofoil section of this group is generally considerably smaller
than that of a conventional airfoil section.
The maximum lift-drag ratio of a section of this group under the non-
cavitating state is considerably smaller, and the lift coefficient at the opti
mum attack angle is also smaller, except for a thin section, compared with
those of a conventional airfoil section.
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

Cavitating Flow 175
4.2.4 Cavitation Characteristics of Hydrofoil Cascade

The flow through a blade row in turbomachinery can be simulated by the
flow through a hydrofoil cascade, and knowledge of this flow is essential in the
design and flow analysis of turbomachinery, despite discrepancies, notably
that the former is a three dimensional and the latter is a two dimensional
flow.
Figure 4.17: Cascade arrangements, and velocities before and behind cas
cades. (a) decelerating cascade; (b) accelerating cascade.
Numachi established the experimental method and apparatus for the

hydrofoil cascade [4.33], and published the most comprehensive data on the
cavitating cascade [4.38]. The following description will be centred on these
results.
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`--
The researches were conducted on the decelerating and accelerating cas

cades, the arrangements of which are shown in Table 4.1 and Fig. 4.17(a)

176 Cavitation
decelerating accelerating
cascade angle at a = 0 25.73° 27.29°
pitch chord ratio 1.237 1.0
Table 4.1: Experimental arrangements for decelerating and accelerating cas

cades
and (b). In the figures, the velocities in front and at the rear of the cascade
Vi, and V2, and their mean velocity V, to which the attack angle, the lift
and drag coefficients, and the cavitation number are referred, are also shown.
The cavitation number is defined as
P-Py
° pV2/2
where P = Pi + pV?/2 - pV22/2
4.2.5 Cavitation Characteristics of Conventional Airfoil and

Ogival Profiles (in Decelerating and Accelerating Cas
cades)
The hydrodynamic data of the airfoil sections of Clark Y, Clark YH and
the ogival profile of every 6 % thickness ratio arranged in a decelerating
cascade [4.33], and Clark Y and the ogival profile, each of 8 % thickness
ratio, arranged in a accelerating cascade have been published [4.30].
Decelerating Cascade
The attack angle of cavitation occurrence in each zone at a given cavitation
number shifts larger from that in the isolated condition by almost a fixed
value, c*i — a, as can be seen in Fig. 4.18, for the 06 profile for example
[4.33]. The lift coefficients at the occurrence of the Zone I cavitation in the
cascade and isolated conditions, however, are almost the same for the OQ
profile (the value for cascade is only a few per cent larger).
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
The comparison of the lift coefficient at the Zone I cavitation occurrence

among the three profiles shows that Oe exhibits the largest value through
out the cavitation number range from larger than 2.0 through 0.45, except
that Clark YH 6 exhibits the largest for the range of the cavitation number

Cavitating Flow 177
Profile 0 o
9,'25.13' faWt
ID Stifle tmÛ-lltC %?M R*>W-/JHl*
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
Figure 4.18: Differences of attack angles between decelerating cascade and

isolated hydrofoils under cavitating conditions [4.33].
between 1.4 and 1.2. Apart from this range, Clark Y 6 has a greater Ci
than Clark YH 6 for a range up to a cavitation number of 2.0, above which
the Ci of two profiles are more or less the same.
The performance characteristics of the Clark Y 6 section are superior in

the cascade compared to the isolated blade at a lift coefficient smaller than
around 0.6 under the non-cavitating condition and become more so under
the cavitating conditions, but become inferior at a lift coefficient larger than
around 0.6 and the border value reduces with the growth of cavitation due
to the cascade effect, as shown in Fig. 4.19(a). In regard to the 06, section,
however, those in the cascade become inferior to those when isolated over al
most the whole range of lift coefficient under non- and cavitating conditions,
especially at a smaller cavitation number, as shown in Fig.4.19(b). Those of
Clark YH 6 in the cascade and when isolated are comparable.

178 Cavitation
ft vi ui ui
m u* u* w
(b)
Figure 4.19: Comparison of polar diagrams between decelerating cascade

and isolated hydrofoils [4.33]. (a) Clark Y 6; (b) O e .
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

Cavitating Flow 179
Accelerating Cascade
The attack angle of the Zone II cavitation occurrence at a given cavitation

number shifts smaller by almost a fixed value from that in the isolated con
dition, in the opposite direction to the decelerating cascade, i.e. the Zone
II cavitation becomes less likely at a given attack angle in the accelerating
cascade. In contrast, the Zone III cavitation becomes more likely. Zone I
cavitation does not occur in the range of attack angle smaller than 6° on
Clark Y profile but does occur in the small range of attack angle around 0°
at a cavitation number larger than 0.77 on the ogival profile. The lift coeffi
cient at the occurrence of the Zone II cavitation in the accelerating cascade
is several percent smaller than that when isolated [4.30].
In the accelerating cascade, violent vibration is not induced by the oc
currence of cavitation on either profile.
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
Figure 4.20: Comparison of polar diagrams between Clark Y 8 and 0% in

accelerating cascade [4.30].
The comparisons of the drag coefficient at a given lift coefficient in the

cascade with that when isolated show that the Clark Y profile is superior
generally, especially in the ranges of the large lift coefficient and the small
cavitation number, except being inferior in the small range of the lift coeffi
cient of 0.1 through 0.4 by only a small amount. Whereas the 0 8 profile is
inferior in the range of the lift coefficient of 0.1 through 0.6 and generally at
a cavitation number smaller than 0.6, except being superior in the range of

180 Cavitation
the lift coefficient larger than 0.6 at a cavitation number larger than 0.6.
The comparison of the drag coefficient at a given lift coefficient between
the two profiles show that Clark Y profile is superior generally, both in
the accelerating and decelerating cascade, especially in the range of small
cavitation number, except in the range of the lift coefficient larger than
around 1.0 at a cavitation number larger than 1.0, as is shown in Fig. 4.20.
4.2.6 Hydrofoil Profile Suitable for Decelerating and Accel

erating Cascades
The "cascade profile", the hydrofoil profile which exhibits superior perfor
mance under cavitating and non-cavitating conditions when arranged in a
cascade was proposed and investigated by Numach [4.37].
With thickness ratio 8 % and lift coefficient 0.9 or 0.6, by the potential
flow approach (ascertained by measurement [4.37]) combined with exper
iments conducted in the cavitation tunnel for the decelerating [4.34] and
accelerating [4.32] cascades (Table 4.1), the pressure distribution was classi
fied into three types according to the trend of pressure change on the upper
surface between x\ and xi (Fig. 4.21). They are the 1st, 2nd and 3rd
types, being the falling, constant and rising downstream-wise distributions,
respectively. The dependence of profile performance on the type of pres
sure distribution, together with the curvatures of the curves between x\ and
X2, and £3 and £4 (especially the curvature near xi), the positions of the
four points x\, X2, £3 and X4, and the value of the minimum pressure were
investigated.
Fig. 4.21 shows the profile forms with the third type of pressure distri
bution as examples of the profile forms [4.34]. Among them, the profile form
NAS 13368 exhibited superior performance at larger and smaller cavitation
numbers. Fig. 4.22 shows the polar diagrams of the profiles at various cav
itation numbers compared with those of the profiles forms, NAS 11268 and
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
NAS 10168 which exhibited the best performance.

Fig. 4.23 shows the profile forms and the pressure distributions of the
1st type for the accelerating cascade. Among them the profile form TNA
16168 exhibited the best performance. Fig. 4.24 shows its polar diagrams
compared with those of TNA 15268 and TNA 10368, the representatives of
the profile forms with types 2 and 3 respectively. TNA 16168 is best except
at a cavitation number smaller than 0.6 when the lift coefficient is equal
to 0.6 and at the cavitation number of 0.4 when the coefficient is between
0.3 and 0.5, as can be seen in Fig. 4.24, although the incipient cavitation

Cavitating Flow 181
1 1
IV.Ai.MJOlf
/ NAS./MtS
b /^ /
A' _'
1 _L
■
fej / •*"*
^ —— X
26 90 (0 iff , ' /Of
y MS./2366> i
1 1 */ y
\*y
JX&* ^~
AiASiOM
Figure 4.21: Types of pressure distribution, 3rd type [4.34].
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

182 Cavitation
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
Figure 4.22: Suitability of pressure distribution [4.28].
number on the upper surface and the cavitation number at which the lift
coefficient begins reducing noticeably are larger than the other two profiles.
The main conclusions are: (i) the 3rd and 1st type pressure distributions
are suitable for decelerating and accelerating cascades, respectively, (ii) the
profile with the 1st type distribution is superior at a smaller cavitation num
ber and at a lift coefficient larger than the given value, (iii) the aggravation
of performance at a smaller cavitation number is caused by the occurrence of
the Zone III cavitation and the minimum pressure near x\ should be as high
as possible, and (iv) the profile form with type 2 pressure distribution could
never exhibit superior performances for either a decelerating or accelerating
cascade.
These results of researches were used for the development of the thin and
thick hydrofoils for high-head Kaplan turbines [4.31].
4.2.7 Computer Simulation of Partially Cavitating Foil

The established methods for analysing cavitating flow around a hydrofoil are
the free-streamline theory and the linearised theories (e.g. [4.51).
In order to compensate for the inconsistencies of assuming the incom
pressible and in-viscid flow in the free-streamline theory for analysing a par-

Cavitating Flow 183
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
T.NAJ5/S8
** T.MJS/S*
TNA. IVf8
Figure 4.23: Pressure distributions and profile forms, accelerating cascade

[4.32].

184 Cavitation
1.2 7*
a 7P^TNA
, „„ ! ,'VT'NA xl°
IO
Efl* 15268 tt^JlSBS' T'l0368pl e :/;T=F"
f/r 10368 "1:^10368
t k
?■ "16168
0.8 W l 6 B 8 - j y # I6B8 " 4 ?
~ TNA—I
.15268
16168 s^ -f-
JTNAI6I68
**T cr=3,0 ?^cr=u4 " t ^ l . 0 4+ggQ.8 15268 I
15268
0.4
n
02 .|a=-3;J_j-3«. sf-pZ-a-Lv-^,
Jv-4* 1 ^ * .10368.
li."vi0368 îTOi .-ffltTV--fiC
0 0.02 0X14 0.02 a04 _a I 0-04
OJ02 L 002 004 0.02 0O4 002 OJ04
Figure 4.24: Comparison of polar diagrams among hydrofoils with three

--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
types of pressure distribution [4.32].
tially cavitating flow around the hydrofoil (which is unsteady and affected
by compressibility and viscosity of gas-liquid two phase flow), flow models
divide the cavity into the front and rear parts, the former being the free-
streamline and the latter being the wake connected smoothly to the front
part, have been proposed [4.53,4.23,4.20].
Alternatively, a flow model has been proposed which treats the inside
and outside of a cavity as one continuum, and analyses the compressible and
viscous flow through the Navier-Stokes' equation and a modified Rayleigh's
equation for a bubble cluster simultaneously [4.22]. Simpler models also
based on the Navier-Stokes' equation but using assumed relation between
pressure and density, have also been proposed [4.45,4.8].
4.2.8 Supercavitating Hydrofoil

Under a supercavitating condition, the hydrofoil neither suffers from the un
steady hydrodynamic force or the cavitation erosion, and has lower frictional
drag because only the lower surface is in contact with the water. The poten
tial this offers for improvement of propellers, hydrofoil boats and hydraulic

Cavitating Flow 185
machines has stimulated extensive research into the use of supercavitation

to reduce profile drag [4.52]. Table 4.2 gives examples of supercavitating
y/c a* ax
Flat Plate 0 0 0
Circular Arc Profile Center Angle: y 6/19•r 7/4
T-B Profile A/2 • [ x'+8/3 • y a ^ e - 4 JC*2] 5/4-A A/4
3 Term Profile A/10- [&x'-20jr*3''s+80jc'a-64jt'0''2] 3/2-A A/2
5 Term Profile A/315 • [210jr'-2240.x'3"'e+12600jr,a 5/3-A A/2
-30912^ B '^+35840je 3 -15360jt ,7X2 ]
CL. = 7r/2-(a+er<.), C D P = TT/2-(ct+ct,) 2 , jr" = x/ c

a a : design attack angle
Table 4.2: Profile forms, and lift and profile drag coefficients of supercavi
tating hydrofoils [4.31].
hydrofoils developed by Johnson Jr. [4.16]. References [4.10] or [4.4] may

be helpful on other hydrofoils. In Table 4.2, the flat plate, the circular arc
profile and the profile proposed by Tulin and Burkart [4.49] are also shown.
The lift coefficient, CL, and the profile drag coefficient, Cpp, in the table
were calculated using the linearised theory at cavitation number 0 (cavity
length I = oo). The values of C'L and C'DP for finite cavitation numbers {a)
can be calculated using the following equations [4.41, 4.7].
C'L = CL/(l + CL/2), C'DP = CDP/(l + CL)

CL = sjClo + 0.85CT2 + oCLO, l/c = l + S/n.CDP/a2
where, / is cavity length, c is chord length and CLO is CL at a — 0.

Supercavitating hydrofoils have the following defects: (i) the sharp lead
ing edge and the thin front part have insufficient strength and rigidity for
supporting the hydrodynamic force concentrated to the leading edge, (ii)
the lift coefficient at the design attack angle is relatively small, and the
profile drag coefficient increases sharply with slight variation from design
attack angle. In order to cope with these defects the following theoretical
and experimental investigations have been conducted.
For the defect (i), the method to give the wedge angle (in degrees) of
--``,```,,``````,,,,``,`,,`,`,``-`-`,,
0.060 times as much as the velocity (in m/s) to the leading edge [4.10], and

186 Cavitation
the methods to give the thickness of parabolic or quasi-parabolic distribution

[4.4] or the roundness near the leading edge [4.25, 4.9] without increasing
the profile drag have been proposed. In addition, the hydro-elastic charac
teristics of flexible supercavitating hydrofoils have been investigated theo
retically [4.26] and experimentally [4.6,4.27]. For the defect (ii), theoretical
and experimental [4.3] suggest that, by attaching a adjustable or jet flap to
the hydrofoil, both a high lift coefficient [4.49,4.24] and a low profile drag
coefficient over a wider range of attack angle can be achieved [4.49,4.11].
4.2.9 Supercavitating Hydrofoil Cascade

The free streamline theory on the choked flow around a cascade of flat plates
[4.5] is well known. Linearised theories on the supercavitating flow around
the cascade of flat plates and circular arc profiles [4.1] and arbitrary profile
hydrofoils [4.40] have also been reported as has the importance of selecting
a suitable stagger angle [4.40].
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

Cavitating Flow 187
4.3 CONTROL VALVES

E. Outa
4.3.1 General Features of Control Valve Cavitation

Various types of control valves are used in various industrial plants, such as
electricity power stations, steel-work plants and chemical plants to regulate
the process pressure and flow rate of the fluid systems. Many of the valves are
operated at extreme levels of fluid pressure. Hence, development of control
valve technology is significantly responsible for smooth and high quality
control operations of modern and future high energy plants.
One of the most serious fluid mechanics problem in control valve tech
nology is cavitation, and the phenomena can occur in all kinds of liquids and
over wide ranges of pressure and temperature. Cavitation, which is a two-
stage phenomena consisting of formation and collapse of vapour bubbles,
induces heavy noise, mechanical vibration, flow fluctuation, and material
damage through cavitation erosion. The control loop is seriously disturbed
by the vibration and the flow oscillation. In a stage of heavily developed
cavitation and flashing of liquid to vapour phase, the flow is choked and un
responsive to the downstream pressure, and the downstream effect becomes
serious because of unsteady column collapse owing to the vapour condensa
tion.
The features of the cavitation are significantly complicated and vary
depending on the shapes of the valve trims. In Fig. 4.25, typical geometries
of standard trims inserted in globe valves are schematically indicated. In a
contoured-plug valve (a), the throttled flow takes a form of an annular jet
issuing from the gap between the plug and the seat-ring. Cavitation bubbles
may be formed in the boundary layer on the plug surface, and also in the
outer shear layer of the jet. In case of a cage guided variable-area trim (b),
several jets issue from the port of the cage and collide with each other at the
centre of the cage cavity. Cavitation bubbles may be formed along the shear
layer surrounding the jet, in a similar way to that seen in an orifice flow, and
also in the mixing zone of the colliding jets. The cavitation configuration will
change with the plug lift, because the cross-section of the jet changes as the
plug moves. In case of a multi-hole cage guided valve (c), many cylindrical
jets issue from the holes of a fine diameter. Ring-shaped cavitation bubbles
as usually seen in small diameter nozzle flows may be generated on each hole.
As the same pattern of cavitation is repeated on each hole, opening the plug
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

188 Cavitation
Cage
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
(a) Contoured-plug valve (b) Cage-guided valve

- variable port area
*«ffil InUTia!! IStg3!
Pu j xdV-x
'-—-, f'APp . AP
. [ 1
APo [ r~~
Cage \l * (l-r)APx
I>d
(c) Multi-hole valve

- variable number
(d) Pressure change in valve
Figure 4.25: Schematics of pressure reducing configuration in typical globe

valves.

Cavitating Flow 189
will not change the basic configuration of cavitation, only the number of
cavitating jets. Cavitation bubbles in the cage-guided and multi-hole valves
collapse in the flow away from the wall, and the vibration and the material
damage are thought to be weaker than those induced by contoured-plug
cavitation.
The complicated flow configuration tends to preclude any physically
based descriptions of cavitation in a valve. Criteria to predict the limits
of various cavitation stages have not yet been established and a one dimen
sional model is useful for initial understanding of the general features of the
valve cavitation. The valve flow is usually replaced by a flow through a sim
ple contraction with a pressure change as indicated in Fig. 4.25(d). When
pressure Pyc of the vena contracta reaches the vapour pressure, then ideal
cavitation is generated. However, high kinematic energy of the jet causes
dynamic phenomena such as vortices and flow separations at the constricted
parts of the flow. The least pressure in the flow process is not necessarily
attained at the vena contracta, but temporarily attained within a vortex
and a separation bubble. Then, the cavitation could occur before the pres
sure Pvc reaches the vapour pressure. To express a pressure margin for the
ideal cavitation inception, either cavitation number a or cavitation index k
is used, which are defined as,
Pd-Pv /AU
CT (41)
= 7wV
k
= ^Tv <«>
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
where P4 is valve outlet pressure, Pu is inlet pressure and Py is vapour

pressure of the liquid.
Tullis (1971) had made numerous number of cavitation experiments for
various types of valves, and the typical data are shown in Fig. 4.26. The
flow rate Q increases monotonously as the differential pressure AP increases
by keeping the inlet pressure and opening of the valve unchanged.The rela
tionship between Q and A P can be expressed as,
Q = CdAy/2AP/P + V2 (4.3)
in which Cd is discharge coefficient, A is valve inlet area, V(= Q/A) is

averaged flow velocity, and p is density of liquid. The discharge coefficient
is defined as,

190 Cavitation
20
a 10
* 5
o
2
1
11 valve
0,5 ■(3 in.) 11 L™
. "NX
N
Ball valve
L
0.2 = 0,22/1Î (2 in.) I
0.05 0.1 0.2 0.5 I
Differential Pressure AP MPa
N ; Cavitatfon-free> I : Incipient
L : Light Cavitation» H : Heavy
V : Very Heavy, S : Super Cavilation
Figure 4.26: Effect of cavitation on discharge flow rate for various types of
valves [4.48].
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

Cavitating Flow 191
(4 4)
""yfitP/p+V* '
The Cd value remains constant for a specified valve opening and valve ge
ometry, so that eqn.(4.3) represents almost a straight line with a slope of
0.5 when Q and AP are plotted on a log-log plane. It is seen in Fig.4.26
that each curve follows the relation until the cavitation has reached a very
heavy stage as indicated by letters V or S. At this stage is reached, the
discharge curve drops off to approach a horizontal line of a constant value
of Q and a further decrease of the outlet pressure does not further increase
the flow rate. This condition with supercavitation developing in the down
stream side of the valve throat is usually called choking cavitation, and the
intersection of the straight line expressed by eqn.(4.3) and the horizontal
line of the choked flow defines the cavitation stage of flow choke.
The choking cavitation is the extreme stage in the cavitation develop
ment. This condition should be avoided in normal control operations, be
cause heavy vibration and cavitation erosion from the collapse of the super
cavitation column are inevitable in the downstream piping.
As indicated by alphabetic letters on each curve in Fig.4.26, cavitation
begins and develops at a considerably lower value of the differential pressure
than the choking value. It should be noted that the flow characteristic
itself can hardly sense the cavitation in the early stage of the development,
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
because volume change of the liquid is only slight. Dynamic data such as
wall vibration, sound pressure and pressure fluctuation are more sensitive to
the cavitation development. In Fig. 4.27, typical data obtained in a test of
a multi-hole cage valve are plotted versusCT,where the inlet pressure is kept
at 5 MPa [Outa, et al, 1986].
As the outlet pressure is decreased, the rms levels of these data increase
monotonously due to development of turbulence. Inception of cavitation
is first found on the acceleration data at a plot denoted by "inc", where
the increasing rate suddenly grows up. In Fourier spectra analysed with
the frequency range up to 100 kHz, a significant growth appears in the
component distributing over a range higher than 25 kHz. This feature
implies that a larger number of fine cavitation bubbles form and collapse
irregularly at these high frequencies. A similar growth in such high frequency
components was also notified in a vibration measurement of butterfly valve
cavitation [Riveland, 1983].
The next stage of the development is the critical stage which is defined
by Tullis and Ball (1974) as the stage that the cavitation development has

192 Cavitation
Pi«B6ure
fluctuation
F l o w discharge Q
80 .!.._!
%
M 70
4Vt -, I. Qck I
a. Qcti Is
60 4 cy o.i
Sound pmcure -o<xx>-
levcl VP/.
J L
O.I 1 JO IO2 1U 4
a P.. - P* kP^«
Figure 4.27: Typical development of cavitation expressed by several mea

sured variables; multi-port cage valve at Cy = 8, 3-inch, Pu = 5MPa[4A4].
become steady, and the ascending rate of the vibration level becomes gradual
as seen at the plot denoted as "cri". It is found that the high frequency
spectral component almost reaches a limiting magnitude, and responds only
slightly to a further reduction of the outlet pressure. The birth and collapse
of small size bubbles in the shear layer has presumably attained a steady
state. This limit is suggested by Tullis and Ball for applications where loud
noise, erosion and heavy vibration are to be avoided. However, the cavitation
at this stage would not be objectionable and would not decrease the valve
life.
The third stage is the initiation of the flow choke, which will be defined
as a limit where pressure fluctuation imposed on the outlet section does not
completely be transmitted into the inlet section. This stage is recognised on
a pressure fluctuation plot indicated as "ch". Such a situation of pressure
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

Cavitating Flow 193
wave transmission is a particular characteristic which arises in a compressible

two-phase flow through a contraction as the velocity reaches the sound speed
of the mixture. The final stage is the flow choke as described before.
It is again noted that the stage of "inc" and "cri" can hardly be deter
mined on the flow rate data, since the time and space averaged value could
scarcely respond to a very slight change of the liquid volume by the birth and
growth of bubbles. Similarly, the far field noise data is not suitable for an
accurate diagnosis, because the data are rather smooth due to attenuation
in the transmission path and because of the widely distributed nature of the
noise sources.
10
3 -
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
.§
5
3 °-3 ]*-flora nh>e
OV-5MP.)
dove n)ve
rrObtBdl)
2*-(love rah*
(BK»M.ft-0.4MPrt
*2«p«tcife 0 12-kKb a 2-pon
0.1 ■ 4-poncafc » 10-inch o 4-portc*c
y"- [Flowcfaofcel • CoMowwJ-pJug O CoMurcd-phif
T MdH-hottcm
0.03 'I I \ i
0.003 0.01 0.03 0.1 0.3 0.003 0.01 0.03 0.1 0.3 0.003 0.01 0.03 0.1
Discharge coefficient Cd
Figure 4.28: Cavitation limits measured for various types of valves [4.44].
The cavitation data of the respective stages are plotted in Fig.4.28 for
several types of the globe valve trims. The cavitation number at the each
stage is seen to increase with the discharge coefficient, described by a simple
power-law relationship with an exponent of 0.5 for trims of contoured-plug
and variable-area cage, as indicated by lines I-I. The cavitation number of
the multi-hole cage trim is less sensitive to the valve opening (lines II-II)
because lifting the plug merely alters the number of the holes through which
high speed flow issues, and the cavitation development is almost entirely
determined by the geometry of the hole and the spacing between the holes.

194 Cavitation
The frictional pressure loss along the passage upstream of the vena con-
tractas of the contoured-plug and the multi-hole cage, which is denoted as
r A P x on the diagram in Fig. 4.25, has a considerable role in delaying the
cavitation inception. However the cavitation development is very rapid, and
the flow choke is initiated immediately after that the critical cavitation is
attained. In contrast, in the variable-area cages, the loss is very slight be
cause of the shape of a sudden contraction, so that the inception takes place
at a considerably higher cavitation number and the cavitation development
is rather gradual. The stage of the flow choke is seen to be reached almost
at an identical value of a irrespective to the trim geometries.
4.3.2 Cavitation Pictures of Globe Valve Flows

Cavitation in valves will appear in various manners, depending on the ge
ometry and flow conditions of the valves. Instantaneous pictures of typical
cavitating flows in a contoured-plug valve and a variable-area cage guided
valve are shown in Fig. 4.29 and 4.30 respectively.
Figure 4.29: Instantaneous pictures showing cavitation development around

a contoured-plug at a high pressure operation. Pu — 5MPa, Pd — 2.4 —
0.2MPa; Cd = 0.012 [4.43].
The series of pictures in Fig.4.29 were taken at a quite high inlet pressure
of 5 MPa, through a side window attached to a three-inch valve body. The
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

Cavitating Flow 195
Figure 4.30: Cavitation bubbles and cloud developing within a 2-port cage-
guided valve [4.44].
outlet pressure ranged between 2.4 and 0.2 MPa. With such a large pressure
difference, the flow velocity at the throat between the plug and the seat- ring
is more than 150 m/s. The shear rates in the boundary layer on the plug
surface, as well as in the surface layer of the annular flow issuing from the
throat, is thought to be extremely high.
The inception is found at a = 0.94 with fine cavitation bubbles forming
around the leading part of the seating-rim, see picture B. The r.m.s level of
the valve vibration increases rapidly to a level more than twenty times the
cavitation-free level. This location of the cavitation region is usual for low
opening cases, and closely related to the presence of a vortex generated by
a separation of the boundary layer flow. The pressure in the vortex may
possibly be reduced to reach vapour pressure.
As the outlet pressure decreases, the cavitation around the seating-rim
further develops, with an additional appearance of bubbles generating and
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

196 Cavitation
collapsing along the outer layer of the annular flow, as seen in pictures B and
C. At the stage of picture D, where Pd = 0.6 MPa, the plug is almost covered
by ring-shaped clouds of cavitation. With bubbles shed downstream of the
seating-rim, a ring-shaped cloud is formed. This implies that the shedding
of bubbles is almost periodic and synchronous on the circumference, giving
rise to serious noise and vibration. Beyond this stage, the downstream side
of the throat is filled with vaporous flow as seen in picture E. The levels of
noise and vibration are reduced considerably low due to the absorption of
pressure waves by the vapour volume.
It is found in these typical cases of high pressure cavitation that the

one-dimensional flow consideration such as indicated in Fig. 4.25(d) is by no
means useful to describe the mechanism of cavitation generation. The fluid
pressure which is related to the vapour pressure is not the pressure at the
vena contracta but presumably a local and temporal pressure inside vortices
produced by the shear layer of the throttled flow.
The picture presented in Fig.4.30 shows the cavitation within a 2-port

cage valve, in which a couple of counter-directing jets issued from the two
throttling port collides with each other at the cage centre. In this case,
the inlet pressure was kept at 0.4 MPa, and the outlet pressure was varied
between 0.29 and 0.07 MPa. The incipient bubble appears, taking a form of
a spiral string, either in the throttling port or in the cage centre depending
on the valve opening as seen in pictures A and B.
In the high opening case, a longitudinal vortex is generated at the side

corner in the port cross-section. In the low opening case, an alternative
longitudinal vortex is generated by the collision of the two jets at the cage
centre. The pressure along the core of a vortex can easily drop to the vapour
pressure, so that in either case, the string-like bubbles appears first as the
outlet pressure is reduced. This inception takes place quite intermittently
with a very short duration, accompanied by pulse-like noise and vibration.
As the outlet pressure is further reduced, cavitation bubbles appear

around the outer surface of the jets, see picture C. The bubble generation
at the cage centre has already been steady, and both types of the bubbles
form a cloud of cavitation as seen in the picture D. Since the collision of
the counter-directing jets is unstable, the cavitation cloud oscillates back
and forth causing the valve to vibrate at a low frequency. The picture E
indicates a heavily oscillating cavitation cloud.
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

Cavitating Flow 197
4.3.3 Cavitation Inception due to Vortex Growth

As described in the previous section, the features of cavitation in control
valves are complicated. However, if the argument is restricted to the in
ception in a practical sense, some physical models for prediction may be
derived from flow visualisation studies and numerical analyses of cavitation-
free Navier-Stckes equations. One of the essential flow structures by which
the field pressure is directly related to the vapour pressure may be the vor
tices arising in the valve flow.
Q2I ! I I I . I 1 I I L
0.002 0.005 0.0! 0.02 0.05 0.1 0.2 0.5
Discharge Coefficient Cd
Figure 4.31: Cavitation number of inception for contoured-plug valves with

flow-to-open installation. Inlet pressure ranges between 0.4 and 5 MPa
[4.13, 4.14, 4.43].
In Fig. 4.31, the cavitation number at inception in contoured-plug valves

is plotted against the discharge coefficient. As the limit denoted by S- is
reached in a high d operation, a string-like bubble is generated in an inter
mittent manner as shown in the attached picture S. By further reducing the
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

198 Cavitation
Figure 4.32: Generation of a string-like cavitation bubble by a longitudinal

Rankine vortex.
outlet pressure, the limit denoted by C- is reached, where a dense cloud of

spherical bubbles is formed around the convex shoulder on the characteristic
surface of the plug, see picture C. In a low Cd operation, cavitation bubbles
first appear around the plug seating-rim as seen in the picture R. This limit
is indicated by R-.
The string-like bubble S is presumed to break out along the axis of a
longitudinal vortex as schematically shown in Fig.4.32. Such a vortex is
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
usually generated by a swirl disturbance imposed at the valve inlet. In the

schlieren picture taken at the moment of the outbreak, the bubble which
traverses the upstream chamber into the throat is found to break down into
a spiral bubble in the downstream side of the throat, generating a spherical
shock wave. In response to the shock initiated blast wave, a pulse-like and
relaxing fluctuation is recorded on time-traces of the pressure.
A semi-empirical formula correlating the incipient cavitation number
Oinc to the discharge coefficient C4 can be obtained by replacing the cross-
sectional configuration of the longitudinal vortex with a Rankine vortex. The
Rankine vortex is a combined vortex composed of an outer irrotational part
and an inner rotational part. The irrotational pressure in the valve throat
section is related to the valve inlet pressure by Bernoulli's theorem, which is
applied along the outer irrotational part of the longitudinal vortex. In the
throat section, the reduction from the irrotational pressure to the pressure
at the vortex core is described in terms of circulation of the vortex. If the

Cavitating Flow 199
radius of the rotational part is small at a high level of the circulation, the
pressure of the core easily drops to the vapour pressure.
By assuming that the outer radius of the vortex at the throat is equal
to the clearance between the plug and the seat-ring, and that the axial flow
in the rotational part is absent, the cavitation number Oinc, at which the
pressure of the vortex core reaches the vapour pressure in the throat section,
can be expressed as:
1 + 4(1 -$)_
= K0 Cd-l, (4.5)
1 - 1% + 2I%logIT
4 _tCd,x II
K K ;
1 - /£ + 2/2l 0 gI T K0> 1 - /2 + 2lZlogIu '
where I — rT/ri is the ratio of the inner-part radius rr to the outer-part
radius r* , and subscripts T and U refer to values at the throat and at the
inlet section respectively.
The coefficient KQ depends on the diameter ratio of the vortex to the
valve passage at the inlet, in addition to the area ratio between the throat
and the valve inlet. If we assume a similarity in the flow structures within
the valve chamber, the values of KQ for various valve sizes can be determined
from the value obtained for a standard-size valve. Then the only parameter
required for the prediction is Iu which expresses the intensity of the swirl
disturbance applied to the valve from the upstream piping.
In Fig. 4.33(a), the cavitation number predicted by the present model is
compared with experimental data for contoured-plug valves. The KQ value
assigned to 3- and 6-inch valves are deduced from the value determined by
a 2-inch valve experiment. The prediction is found satisfactory for practical
application, with the value of Iu in an acceptable range. It is noted that
valve-sizes exceeding three inches slightly affect the inception, whereas the
inception for smaller size valves is significantly delayed by reducing the size.
A string bubble will not be generated, unless the disturbance imposed at
the valve inlet is intense. The KQ value required for the cavitation inception
increases as the inlet pressure becomes high. Detailed descriptions of this
longitudinal vortex model were presented by Inoue, et al (1991) and by Outa,
et al (1993).
The inception of the seating-rim cavitation denoted by R in Fig. 4.31 is
closely related to shear-layer vortices which grow and detach on the seating-
rim as schematically shown in Fig.4.34. A computer generated model shown
in the figure reveals that the flow issued from the valve throat streams along
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

200 Cavitation
12 6'-valvc 0.5 0.3 0.11 1

8l ~K„-3l - \ 1 / 3*-y«lv» _ 7 <; - 0 - 7
b ° / J67 K. - 27
I•? 86 . - ' . -
0.7\
>#t'.-°-vfV0-5
8 ar^X&PTl âi—
2
a 2
4
wv<&^
. A®5-J^ s%!i2
*0-M
—i ■ i
0.02 0.04 0.06 0.08 0.10 0.12 0.01 0.02 0.03
Discharge CoefTiciait Cd Discharge Coefficient Ci
IflhtmwwiIMPll InJttJwwrrlMPil
r-y&n i'-vtrt 6"-rahw 2"-vil*e 3*-v»lve 6"-vmlv»
<vO 2-5 G 5
a 0.8 0 0.8 6 3
± 0.4 a 0.4 • 0.4 <7 0.8
a 0.2 D 0.2 0.4 O 04 • 0.4
{«) Inception of string-like bubble (b) Inception of seating-rim bubble
Figure 4.33: Comparison of predicted and measured limits of cavitation

inception [4.14, 4.15, 4.43].
Distance
Figure 4.34: Computed pattern of shear layer flow on plug surface, and shear
layer vortex model [4.15, 4.43].
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

Cavitating Flow 201
the plug surface, generating a large number of vortices in the boundary

layer. It is further noted that the instantaneous pressure temporarily takes
the lowest value at the core of a vortex K, which locates close to the comer
S where the time-averaged pressure is the lowest. The instantaneous core
pressure of the vortex K is very close to the vapour pressure.
By applying a concept of vortex-induced cavitation [Arndt, 1976], we
consider a Rankine vortex which is generated by the accumulation of vortic-
ity produced in the surface boundary layer. The circulation and the inner
radius of the Rankine vortex are related to the Strouhal number of the vor
tex shedding and to the boundary layer profile. It is again assumed that
the pressure of the vortex core is equal to the vapour pressure at the in
ception. Then, by accounting for both the steady-state pressure reduction
along the valve passage and the temporal pressure reduction by the vortex
at the moment of the shedding, the cavitation number of the inception can
be expressed as:
atnc-(CpS-l) + -.^F^-r.-J-.r-^, (4.7)
where Cps = (F u — P s )/(P U —Fd) denotes the steady-state pressure coefficient

at the location S on the plug.
The coefficient Cps is a monotonously increasing function of the dis
charge coefficient C j , and the coefficient Co which determines the growth
rate of the circulation is found to be 0.46 by the computation. The velocities
qs, qr and Vu denote the values at the location S, the throat T and the valve
inlet respectively. In high Reynolds number flow, the product q^liqrVu) 1S
almost proportional to the inverse square of C<j, and almost equal to 1.2/Cj
in the present case of an equal-percent contoured plug. By expressing the
boundary layer profile with a 1/n -th power law, the boundary layer thick
ness 5S is obtained from conventional flat-plate boundary-layer theory with
a value of n = 6.1. As the valve inlet pressure becomes high, the Reynolds
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
number of the flow on the plug surface increases, and then the reduced value
of 5S contributes to increase the value of <7jnc. Finally, the Strouhal number
5t, which is defined in terms of the throat width h, the throat velocity qx and
the frequency of the vortex shedding, is almost unchanged by the Reynolds
number, and found to be equal to 0.075 (Li/L2), where L\, and L2 are the
plug lift and thickness of the seating-rim respectively.
The value of Oinc calculated by eqn.(4.7) are compared in Fig. 4.33(b)
with the experimental data. The predictions is quite satisfactory, and the
scale-effects of the valve size and the pressure level are also made clear by the

202 Cavitation
present model. The second term in the right-hand side of eqn.(4.7) expresses
the pressure reduction within the vortex, and the value is almost the same in
the magnitude as the value of the first term expressing the pressure reduction
along the valve passage. As C<j decreases, the second term increases to exceed
the first term due to a decrease of the Strouhal number. Further details were
discussed by Inoue, et al (1993) and Outa, et al (1993).
4.3.4 Cavitation Erosion and Anti-Cavitation Valves

Where erosion of the valve trim becomes significant the cavitation number
has been thought to be a fundamental parameter to establish a practical
criterion required for selecting valve type, material and life time. However,
the erosion depends not only on this simple parameter but also on various
local conditions of the valve.
20 hours 100 hours 150 hours
Figure 4.35: Cavitation erosion developed after various test duration at a

very high inlet pressure, 20 MPa. The plug of 11 mm diameter is made of
stainless steel "SUS 316" [4.42].
Among the examples of cavitation discussed above, that generated by

shear layer vortices is thought to be the cause of damage on the frontal slant
of the seating-rim. Typical erosion pictures obtained for a stainless steel
plug during a high pressure operation are presented in Fig. 4.35. After a
working duration of 20 hours, it is found that the circumference of the plug
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

Cavitating Flow 203
LA Di»lvtd02
x —100% 7ppm
4-90* im
* - 90% 50-lOOpfb
D - 75* 7pjm
0 - 6 5 * 7ppn
Working Dnatkw t[hr]
Figure 4.36: Integrated weight loss of a "SUS 316" plug by erosion at an

inlet pressure of 20 MPa [4.42].
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
root is eroded significantly, and that the shut-off capability of the valve is
almost failed. At the end of a 100 hours operation, the eroded portion has
developed in a concentrated area of the plug to a considerable depth, and
the flow characteristic changed significantly. The working limit, in this case,
beyond which repair becomes impossible and fracture at the stem is likely, is
150 hours. The life time can be significantly prolonged by coating the plug
surface with Stellite alloy.
In Fig. 4.36, the weight of the lost part is plotted versus the working
duration. The history is affected seriously by the initial roughness of the
surface around the plug root. Nevertheless, it can be seen that the weight
loss is related by a power law to the working duration. For example, in the
case of 75% lift, in the early stage of erosion up to 50 hours, the eroded
portion develops rather gradual with an exponent of roughly 1/3. In this
situation, the erosion is induced mainly by the shear layer cavitation. This
stage is followed by a stage of rapid development with an exponent of roughly
4. The eroded portion is grooved heavily by a direct impingement of the high
speed throttled jet. As the valve lift is increased to the full open, the stage of

204 Cavitation
Figure 4.37: Tortuous path trims designed for the most severe applications.
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
the rapid development appears immediately, and the valve life time becomes
to almost one tenth of the life time of 75% lift operation. Oxygen dissolved
in the test water has also an effect on the erosion development, and the
exponent in the time dependent relationship is reduced almost by half when
the dissolved oxygen is reduced to less than 100 ppb from a normal condition
of 7 ppm, i.e. 85 % of the saturated amount.
In closing this section, it is noted that multi-layered trims with tortuous
paths are frequently utilised for extremely heavy duty applications. Typical
trims are shown in Fig. 4.37, cited from manufacturers catalogues. In re
ferring to the pressure diagram indicated in Fig. 4.25(d), such design has a
basis on the concept that:
• APp/APo should be increased so that the flow velocity at the outlet

of the each path is reduced,
• APX/APQ should be increased in order that the frictional pressure

reduction is increased, and
• Multi-hole configuration is favourable, because the location of the bub

bles collapse is far from the metal surface.

Cavitating Flow 205
References
4.1 Acosta, A. J. (1960), 'Cavitating Flow Past a Cascade of Circular
Arc Hydrofoils, Calif. Inst. Techl. Rep. E 79, No. 2.
4.2 Arndt, R. E. A. (1976), 'Semiempirical Analysis of Cavitation in

the Wake of a Sharp-Edged Disk', Trans. ASME, Journal of Fluid
Engineering, Vol. 98, No.3, pp. 560-562.
4.3 Auslaender, J. (1964), 'Supereavitating Foils With Flaps Beneath a

Free Surface', J. Basic Engng., Trans. ASME, Vol. 86, Ser. D, pp.
197-204.
4.4 Ausaender, J. (1962), 'Low Drag Supereavitating Hydrofoil Sections',

Hydronautics Inc., TR 001-7.
4.5 Betz, A. and Petersohn, E. (1931), Ing.-Arch., Vol. 2, p. 190.
4.6 Brennen, C , et al. (1980), 'Leading-Edge Flutter of Supereavitating

Hydrofoils', J. Ship Res., Vol. 24, pp. 135-146.
4.7 Brown, P. W. (1963), Davidson Lab. Rep., 971.
4.8 Chen, Y. and Heister, S. D. (1994), 'Two-Phase Modeling of Cavi-

tated Flows', ASME FED Vol. 190, pp. 299-307.
4.9 Furuya, O. (1974), 'Supereavitating Linear Cascades with Rounded

Noses', J. Basic Engng., Trans. ASME, Vol. 97, Ser. D, pp. 35-42.
4.10 Grumman Aircraft Engng. Corp. (1958), Super Cavitating Hydrofoil.
4.11 Harrison, Z. L. and Wang, D. (1967), 'Evaluation of Pressure Distri

bution on a Cavitating Hydrofoil With Flap', J. Ship Res., Vol. 11,
pp. 93-108.
4.12 lhara, A. and Murai, H. (1984), 'Cavitation Inception on a Circular

Cylinder at Critical and Supercritical Flow Range', Intnl. Symp. on
Cavitation Inception, ASME FED, Vol. 16.
4.13 Inoue, F., Outa, E., Matsuoka, H., and Machiyama, T. (1991), 'Flow
Oscillation in Contoured Plug Valve with Multi-Hole Retainer and
the Related String Cavitation Generation', Proceedings, 3rd Interna
tional Symposium on Fluid Control, Measurement, and Visualisation,
ASME, Book No.100315, pp.77-85.
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

206 Cavitation
4.14 Inoue, F., Outa, E., Tajima, K., and Machiyama, T. (1991), 'An
Experimental Study on Control Valve Cavitation (2nd Report, String
Bubble Cavitation In Contoured-plug Valve)', Transactions of JSME,
Vol. 57, No. 544, pp. 7 - 1 5 (in Japanese).
4.15 Inoue, F., Outa, E., Tajima, K., and Machiyama, T. (1993), 'An Ex
perimental Study on Control Valve Cavitation (3rd Report, Vortex
Shedding and Cavitation Inception on a Contoured Plug)', Transac
tions of JSME, Vol. 59, No. 562, pp. 131 - 138 (in Japanese).
4.16 Johnson, V. E., Jr. (1958), 'The Influence of Depth of Submersion,
Aspect Ratio, and Thickness on Supercavitating Hydrofoils Operat
ing at Zero Cavitation Number', Proc. 2nd Symp. Naval Hydrodyn.,
ONR-38, Washington, D. C.
4.17 Kamimoto, G. et al. (1953), 'Cavitation Tests of the Cylinder in Tur

bulent Flow', Trans JSME, Vol. 19, No. 85, pp. 32-37 (in Japanese).
4.18 Kermeen, R. W. et al. (1955), 'Mechanism of Cavitation Inception
and the Related Scale-Effects Problem', Trans. ASME, Vol. 77, pp.
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
533-540.
4.19 Kermeen, R. W. (1956), 'Water Tunnel Tests of NACA 4412 and
Walchner Profile 7 Hydrofoils in Noncavitating and Cavitating
Flows', Calif. Inst. of Tech. Hydrodyn. Lab. Rep. 47-5; Kermeen,
R. W. (1956), 'Water Tunnel Tests of NACA 661-012 Hydrofoil in
Noncavitating and Cavitating Flows', Ibid. 47-7.
4.20 Kinnas, S. A. and Fine, N. E. (1993), 'A Numerical Nonlinear Anal

ysis of the Flow around Two- and Three-dimensional Partially Cavi
tating Hydrofoils', J. Fluid Mech., Vol. 254, pp. 151-181.
4.21 Knapp, R. T., Daily, J. W. and Hammitt, F. G. (1979), Cavitation,
Inst. Hydraulic Res., Univ. Iowa, Iowa, USA.
4.22 Kubota, A. et al. (1992), 'A New Modelling of Cavitating Flows: a
Numerical Study of Unsteady Cavitation on a Hydrofoil Section', J.
Fluid Mech., Vol. 240, pp. 59-96.
4.23 Lemonnier, H. and Rowe, A. (1988), 'Another Approach in Modelling
Cavitating Flows', J. Fluid Mech., Vol. 195, pp. 557-580; Rowe, A.
and Blottiaux, O. (1993), 'Aspects of Modeling Partially Cavitating
Flows', J. Ship Res., Vol. 37, pp. 34-48.

Cavitating Flow 207
4.24 Lin, J. D. (1961), 'A Free Streamline Theory About a Flat Plate with
Flap at Zero Cavitation Number', Hydronautics Inc. TR 119-3.
4.25 Murai, H. and Kinoe, T. (1968/1969), 'Theoretical Research on

Blunt-Nosed Hydrofoil in Fully Cavitating Flow', RIHSM, Vol. 20,
No. 205, pp. 263-297; Murai, H. et al. (1973), 'Theoretical Research
on Blunt Nosed Supercavitating Hydrofoils at Finite Cavitation Num
ber', Proc. 2nd Intnl. JSME Symp. Fluid Machinery and Fluidics,
Tokyo, pp. 247-254.
4.26 Murai, H. and Shimizu, S. (1981), 'Hydroelastic Characteristics

of Flexible Super-cavitating Hydrofoils, Report 1. Two-dimensional
Theory at Zero Cavitation Number When Clamped at Trailing Edge',
Bull. JSME, Vol. 24, No. 194, pp. 1379-1387; Murai, H. and Shimizu,
S. (1981), The Same Tittle Report 2. 'Two-dimensional Theory at
Zero Cavitation Number When Supported Elastically', Ibid. Vol. 24,
No. 194, pp. 1388-96; Murai, H. and Shimizu, S. (1985), 'Steady and
Unsteady Characteristics of Chordwise Flexible Supereavitating Hy
drofoil at Non-Zero Cavitation Number', RIHSM, Vol. 50, No. 372,
pp. 1-24.
4.27 Murai, H. et al. (1983), 'Experimental Research on Hydroelastic

Characteristics of Flexible Supercavitating Hydrofoils', Proc. 2nd
Intnl. Conf. on Cavitation, London, pp. 205-210.
4.28 Numachi, F. and Kawashima, O. (1960/1961), 'Theoretical Consid

erations Concerning the Boundary Layer and its Separation on Hy
drofoils Suitable for Arrangement in Cascade (Report 1)', RIHSM,
Vol. 12, No. 114, pp. 89-107.
4.29 Numachi, F. (1941), 'Cavitation Tests on Four Hydrufoil Profiles',

Trans. JSME, Part 3, Vol. 7, No. 28. pp. III-l - III-9 (in Japanese).
4.30 Numachi, F. (1961/1962), 'Cavitation Tests on Hydrofoils Arranged

in Accelerating Cascade', RIHSM, Vol. 13, No. 121, pp. 1-25.
4.31 Numachi, F. (1969/1970), 'Cavitation Tests on Hydrofoils Designed

for Accelerating Flow Cascade/Report 4 (Three Profiles Designed for
High Head Kaplan Turbine)', RIHSM, Vol. 21, No. 210, pp. 27-55.
4.32 Numachi, F. (1964/1965), 'Cavitation Tests on Hydrofoils Designed

for Accelerating Flow Cascade/Report 3 (Five Profiles Generated
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

208 Cavitation
From Prescribed Pressure Configurations of Types 1 and 3)', RIHSM,

Vol. 16, No. 151, pp. 1-33.
4.33 Numachi, F. (1954), 'Cavitation Tests on Hydrofoils in Cascade and

its Theoretical Basis of Experiment', RIHSM, Vol. 4, No. 39, pp.
125-158.
4.34 Numachi, F. et al.(1953), 'Cavitation Tests on Hydrofoil Profiles Suit

able for Arrangement in Cascade (3rd Report)', RIHSM, Vol. 3, No.
28, pp.99-138; Numachi, F., et al. (1953), the Same Tittle (4th Re
port) 'Tests on Three Cascade Profiles of Type 3 with Thickness
Ratio of 8 Percent', Ibid. Vol. 3, No. 29, pp. 139-162.
4.35 Numachi, F. et al. (1967), 'Cavitation Tests on Clark Y Profiles of

Several Thickness Ratio', RIHSM, Vol. 8, No. 74, pp. 47-56.
4.36 Numachi, F. et al. (1957), 'Cavitation Tests on Six Profiles for Blade
Elements', RIHSM, Vol. 8, No. 73, pp. 25-46.
4.37 Numachi, F. (1971), 'Surface Pressure Distribution on Hydrofoil Pro

files in Cascade, Report 1/Decelerating Cascade Flow', RIHSM, Vol.
24, No. 240, pp.143-179.
4.38 Numachi, F. (1962/1963), 'Summary Report on the Research of Cav

itation Phenomena Obtained Hitherto in the Institute of High Speed
Mechanics (Report 2)', RIHSM, Vol. 14, No. 135, pp. 73-87.
4.39 Numachi, F. (1954), 'Summary Report on the Research of Cavitation
Phenomena Obtained Hitherto by Our Institute', RIHSM, Vol. 4, No.
40, pp. 159-190.
4.40 Oba, R. (1965), 'Theory on Supercavitating Cascade Flow for Arbi

trary Form Hydrofoil', Proc. Cavitation in Fluid Machinery, ASME
Bound Volume, pp. 70-90.
4.41 Oba, R. (1963/1964), 'Theory on Supercavitating Hydrofoils at Ar

bitrary Cavitation Coefficient', RIHSM, Vol. 15, No. 141, pp. 1-20.
4.42 Okutsu, R., Yuzawa, S., Hashizume, T., Outa, E. and Machiyama,
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
T. (1997), 'Cavitation Erosion Features in Industrial Control Valve

at the Inlet Pressure of 20 MPa\ Proceedings, International Confer
ence on Power Engineering 97, JSME/ASME/CFTE, Tokyo, Vol 1,
pp.339-344.

Cavitating Flow 209
4.43 Outa, E., Inoue, F., Tajima, K., and Machiyama, T. (1993), 'Incep
tion of Vortex-Generated Cavitation in Industrial Contoured-Plug
Valve', Proceedings, 4th International Symposium on Cavitation In
ception, ASME Winter Annual Meeting, New Orleans, FED Vol.177,
pp.143-151.
4.44 Outa, E., Tajima, K., Machiyama, T., and Inoue, F. (1986), 'High
Frequency Vibration of Control Valves at Low Openings Due to Cav
itation Development', Proceedings, 2nd International Symposium on
Cavitation and Multiphase Flow Noise', FED, Vol. 45, NCA-Vol. 2,
ASME, pp. 31-40.
4.45 Reboud, J. L. and Delannoy, Y. (1994), 'Two-Phase Flow Modelling

of Unsteady Cavitation', 2nd Intnl. Symp. on Cavitation, Tokyo,
Japan, pp. 39-49.
4.46 Riveland, M. L. (1983), 'The Industrial Detection and Evaluation of

Control Valve Cavitation', ISA Transactions, Vol.22, No.3, pp. 71 -
80.
4.47 Shalnev, K. K. (1956), Symp. NPL, 1955, London, 22, p. 1.
4.48 Tullis, J. P. (1971), 'Choking And Supercavitating Valves', Proceed

ings of ASCE, J. Hydraulic Division, Vol.97, HY 12, pp. 1931-1945.
4.49 Tulin, M. P. and Burkart, M. P. (1955), 'Linearized Theory for Flows

about Lifting Foils at Zero Cavitation Number', David W. Model
Basin Rep. C-638.
4.50 Tullis, J. P. and Ball, J. W. (1974), 'Cavitation Data for Valves and its
Application', Proceedings, Conference on Cavitation, IME, Heriot-
Watt University, C 153, pp. 55-63.
4.51 Wade, R. B. (1967), 'Linearized Theory of a Partially Cavitating

Plane-Convex Hydrofoil Including the Effects of Camber and Thick
ness', J. Ship Res., Vol. 11, pp. 20-27.
4.52 Wu, T. Y., Ann. Rev. Fluid Mech. Vol. 4, pp. 243-285.
4.53 Yamaguchi, H. and Kato, H. (1983), 'On Application of Nonlinear

Cavity Flow Theory to Thick Foil Sections', 2nd Intnl. Conf. on Cav
itation, IMechE, pp. 167-174.
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

210 Cavitation
4.54 Young, J. O. and HoU, J. W. (1966), 'Effects of Cavitation on Periodic

Wakes Behind Symmetric Wedges', Trans. ASME, Vol. 88, Ser. D,
pp. 163-176.
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

Chapter 5
Cavitation Phenomena in
Hydraulic Machinery
5.1 General Features of Turbine Cavitation

H. Tanaka
5.1.1 Cavitation in Francis Turbines

A Francis turbine is designed for a specific operating condition which is de
fined by specific hydraulic energy, turbine output and speed. The profiles of
water passages and runner blades are designed to give a possible best per
formance at this design condition. The runner does not have any adjustable
part to improve the performance at off-design conditions like a Kaplan tur
bine. Therefore, the flow at off-design conditions has a great tendency to
promote cavitations and these cavitations are thus related to the off-design
conditions. Figure 5.1 shows various off-design operating ranges of a Fran
cis turbine where cavitations are likely to occur. The ordinate represents
discharge factor, QED, and the abscissa represents speed factor, TIED,
nED =
7I~; QED =
Wf
The range indicated by Q is the overload condition, where runner
blades develop cavitation at their downstream areas as shown in Figure 5.2.
If the Thoma number (cavitation number) is reduced until the cavitating
area is about a half length of the blade pitch, the turbine efficiency begins to
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`--- 211
212 Cavitation
120 140
Figure 5.1: OfF-design ranges and related cavitations
A
®t>S vi«w
Figure 5.2: Operating-range related cavitations
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

Cavitation Phenomena in Hydraulic Machinery 213
decline. Figure 5.4 is an example of the cavitation tests of a model Francis

turbine for five different guide vane openings.
The break point of turbine efficiency is denned as "critical Thoma num
ber" or "critical sigma", acl. If the Thoma number is reduced below the
critical sigma, cavitation grows rapidly. At an efficiency drop of 2 - 3 %,
cavitation may develop over most of the runner exit area, producing strong
cavitation noise.
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
o, oc
Thoma number o
Figure 5.3: Characteristic Thoma numbers.
The value of critical sigma increases with the increase of load, as shown in
Figure 5.5. It shows the variation of the critical sigma with turbine discharge
(this relationship is obtained from the test results shown in Figure 5.4). The
performance characteristics of turbine (top) is shown for reference, where
' i n IEC code, two characteristic Thoma numbers, Co and a\ ,are defined. oo (Thoma
number zero) is the lowest value of the Thoma number for which efficiency remains un
changed. cri (Thoma number one) is the value of Thoma number for which a drop of 1 %
efficiency is obtained. The critical Thoma number,<xc, is a characteristic Thoma number
which has been used widely in practice. It is the value below which the efficiency will be
come a value lower than its base value which corresponds to the non-cavitating condition
as croshown in Figure 5.3.

214 Cavitation
-1 1 1—
92
90
T^7 GVO = 30 mm
92 27.5mm
„ 88, 90- 25
£ 92*
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
88
■ s 2 2 . 5 Run
92
M 88_
90
92'
88 17.5
90
88 A QVO = 30 &..
120
• a 0 o-
27.5mm
-o o- a
25
2 2 . 5 mm 100 _
.O*°Ô"-0—
e o>
17.5 mil
80
f ovo s 30 on
27 5
1.0
25 ma
° °
-
"°lT^w^ n n
5 22.S an
■
^4/TTTl O 0
°
0.8
1 7 . 5 .«
-O a—O-
*<*r
0.6
0.1 0.2 0.3 0.4
a
Figure 5.4: Cavitation test results at different guide vane openings

140
Hs = 201 m i n - ' k y - V - "
D? = 350 mn nip - 215 Bin-
120
GVO 3 30 mil
100
±H-
v.
'v T ' > /
4-> V^-Qio = 97.0 liiy.
80
60
40
b mm
20
150 200 250 300 350

n. (nln')
0.20
D? - 350 nan
, m = 220 win 1
0.15
0.10
0.05
60 80 100 120
Qt (lit/a)
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
Figure 5.5: Efficiency and cavitation characteristics of a Francis turbine

216 Cavitation
n\ = -j^rx, Qi = jjte and Pi = TJTT- ^ t n e cavitation is limited only to

the vicinity of blade trailing edge and the bubbles collapse downstream, no
serious erosion will occur. On the contrary, should the cavitation appear in
the upstream area and the bubbles collapse on blade surface, heavy erosion
will result.
In the partial load operating range designated as (2) , a large vortex
cavitation appears in the draft tube just below the runner cone as shown
in Figure 5.6. This vortex cavitation, which is usually called "draft tube
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
Figure 5.6: Draft tube cavitation observed in partial load operation
whirl", is caused by the residual circumferential velocity component of the

flow discharged from the runner. The helical vortex core usually revolves
at a frequency of 0.25 - 0.35 times the rotating speed of the runner. This
cavitation does not cause erosion but often produces large low frequency
pressure pulsations and power fluctuation (power swing) due to the revolu
tion of the cavitation core. In some load ranges (usually about 50 - 60 % of
the full load), it becomes so powerful and rough that it develops large burst-

Camtation Phenomena in Hydraulic Machinery 217
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
Figure 5.7: Example measures to eliminate the vibration and noise caused
by draft tube cavitation

218 Cavitation
ing noise in the draft tube or causes strong vibration to the turbine and the
power-house. To mitigate such undesirable effects, two measures are usually
employed. One widely used measure is the admission of air. The air is usu
ally admitted through the runner cone or the diametrical transverse air pipe
installed in the upper section of the draft tube as shown in Figure 5.7(a).
Air inlets on the draft tube wall (Figure 5.7(a)) are also used. The air inflow
can stabilise the vortex core and suppress the vibration. The amount of air
admitted should be optimised to suppress the vibration effectively but with
out any adverse effect on efficiency. The other measure is the installation of
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
fins in the upper part of the draft tube, as shown in Figure 5.7(b), to reduce
the vortex intensity. Since too large or too many fins will impair efficiency,
a suitable number of fins with reasonable size should be selected.
In the operating range (3) , corresponding to high head operation, local
cavitation is developed on the suction side of the blade inlet edge near the
band as shown in Figure 5.2. This cavitation is localised to a small area but
with high intensity, likely to cause deep erosion. To avoid this cavitation,
proper selection of the operating range of the model is vital. Precautions
must be taken to exclude this operating range as far as possible from the
actual operating range expected for the prototype. Various factors, such as
the blade profile near the area, the angle between the suction side of the blade
and the runner band, the distance between the guide vanes and the inlet
of runner blades, all affect the cavitation intensity in this area (3) . In the
operating range (3) , corresponding to partial load operation, flow separation
takes place along the inlet edge as shown in Figure 5.2. The upper end of
the vortex cavitation touches the underside of the crown and may cause
light erosion there. The lower end of the vortex usually does not contact
the runner band but it flows away along the passage between the runner
blades near the band. Therefore, this vortex cavitation is sometimes called
"inter-blade vortices". This type of cavitation does not cause any erosion
or vibration in most of the operating range. However, for an extremely
high head operating range, these vortices become unstable and may cause
hazardous vibration.
In the operating range © , corresponding to extreme low head opera
tion, the inlet flow has a large negative incidence angle which may cause
flow separation, accompanied by an unstable vortex cavitation. It causes
heavy vibration in extremely low head operating range (below 70 % of the
optimum head).
Apart from the major cavitations described above, minor local cavita-
tions may also be observed sometimes. One is the cavitation downstream

of the runner seal clearance. The erosion caused by this cavitation is usu
ally minor but it becomes heavier when the runner seal clearance is larger.
Typical damage location is shown in Figure 5.8
Figure 5.8: Example of cavitation erosion on the wearing ring
"Required submergence" for a turbine to prevent cavitation damage is

usually determined by the critical sigma as shown in Figure 5.5. In general,
the "required Thoma number" or "plant sigma" (<JP) is determined by adding
a certain margin to (or multiplying a safety factor) the value of the critical
sigma at the maximum turbine discharge. The required submergence is
then calculated from this plant sigma. It should be noted, however, that
the critical sigma represents the Thoma number where the cavitation at the
suction side of the blades grows large enough to impair turbine efficiency
but it does not say any thing about the degree of cavitation in other local
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

220 Cavitation
areas such as the inlet area of the blade, which is scarcely related to the
deterioration of efficiency. Furthermore, the plant sigma determined in such
a way does not guarantee a cavitation-free operation.
For low head machines, the cavitation intensity (power density relating
to cavitation erosion damage) is low and the presence of some cavitation
may be tolerated. On the contrary, for very high head machines, the cavi
tation intensity is very high because of high flow velocities. The presence of
even very localised cavitation may cause severe erosion. To ensure no cav
itation erosion in these machines, a plant sigma larger than the "incipient
Thoma number",CTJ,should be employed (whereCTJis the Thoma number
corresponding to the inception of cavitation).
5.1.2 Cavitation in Propeller Turbines

Like Francis turbines, propeller turbines with fixed blades show various types
of cavitation relating to the operating ranges as shown in Figure 5.9. In
the operating range ® , cavitation occurs on the suction side of the blade
due to heavy blade-loading, the location of which is indicated by (T) in
Figure 5.10. In the operating range (2) , vortex cavitation in the draft tube
takes place in a similar way to Francis turbines (see Figure 5.6). In the
operating range (3) , the incidence angle of the relative inflow at the blade
inlet becomes larger, causing the cavitation along the inlet edge indicated
by (3) in Figure 5.10. On the other hand, in the range (§) , the incidence
angle becomes negative and cavitation appears on the pressure side of the
blade inlet-edge as marked by @ in Figure 5.10.
In the case of Kaplan turbines (ie with adjustable runner blades), the
machines usually operate at so-called "on cam" condition, ie the optimum
combination of runner blade angle and guide vane opening. Therefore, the
angle of incidence of the runner blade is maintained close to the optimum
value over a wide operating range. Consequently, if the profile of the runner
blade is properly designed, cavitation is unlikely to occur in areas (3) or (3) .
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
Also, in partial load operation, the regulation of blade angle minimises the
residual circumferential component of the absolute discharge flow velocity
from the runner, making the vortex cavitation in the draft tube much less
severe compared with Kaplan turbines. Therefore, for Kaplan turbines, only
the cavitation near the full load range shall be considered.
Additionally, for propeller or Kaplan turbines, clearance cavitation may
develop at the blade tip as shown in Figure 5.11(a). This often causes
erosion on the suction side of blades (as shown in Figure 5.11(b)) where the

nm• 422 Bin- kW*B y>

200 B| 350 an
/fli/*
ISO
-x'fly^-. lUST***^
*» 'J~ 'SfM /*"
w 160
<5 / y v /
fyyi 'k/-'
§r*
140 1
w 9
120
-^^ /
« t i
400 500 600
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
n, ( » l n ' )
Figure 5.9: Operating ranges of a propeller turbine where cavitation takes

place
Rt«ctlon »ld«
Figure 5.10: Typical cavitation on the runner blade of a propeller turbine

specific to each operating range

222 Cavitation
Figure 5.11: Clearance cavitation in propeller turbines
tip vortex cavitation touches the blade surface.

Cavitation on the blade surface of high head Kaplan turbines can be re
duced by increasing the solidity (chord/pitch ratio) of the runner. However,
the tip cavitation can not be reduced significantly because the shape of the
blade tip is common to both high head and low head machines. Neverthe
less, the erosion can be reduced by providing a fin at the blade tip as shown
in Figure 5.12(b). The fin itself does not reduce the intensity or amount of
the clearance cavitation, but it keeps away the vortex cavitation from the
blade surface, preventing the possible damage.
The runaway speed of both propeller and high specific speed Francis
turbines is affected by cavitation. However, the effect is so insignificant
that it may be practically ignored. On the other hand, for Kaplan turbines,
the effect is obvious and should be taken into account when estimating the
runaway speed of a prototype Kaplan turbine. The highest runaway speed of
a Kaplan turbine is usually observed at a runner blade angle between 9° ~
12°. Cavitation also affects the runaway speed significantly in this range
of blade angle. Diagrams of runaway speed versus a at the blade angle of
12° of various Kaplan turbines with the number of blades, ZR, from 5 to 8
are shown in Figure 5.13. Since the value of Thoma number to cause the
increase of runaway speed is usually close to the plant sigma, it is desirable
to check the runaway speed at plant sigma by a model test.
Runaway speed varies in two different modes (Figure 5.14) when Thoma
number is reduced [5.17], subject to the appearance of cavitation (Figure
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

B)ade t i p Discharge r i n g z*"

Fillet/
Iview A |
Fillets
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
Figure 5.12: Fin at blade tip to prevent tip cavitation erosion
1600
0 0.2 0.4 0.6 0.6 0 0.2 0.4 0.6 0.» 0 0.2 0.4 0.6 0.8 0 0-2 0.4 0.6 0.8
g a a <>
&, » 5 U = 6 U = 7 U ° 8
Figure 5.13: Runaway speed vs a at blade angle of 12° for Kaplan turbines
with blade number from 5 to 8 (model runner diameter = 200 mm)

224 Cavitation
B.A.=12*
Mode B - GVO =18mm|
V • Transition
1400 p
\Wo± Mode
c
■H
E
1200 h
Figure 5.14: Two modes of runaway speed variation with cavitation
b* = 81
Mod« B
IaT^-71
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
Hoda B
Figure 5.15: Different appearance of cavitation observed in each mode of

runaway condition under cavitation

5.15). In Mode A, the same appearance of cavitation is observed on all

blades but, in Mode B, cavitation patterns repeat cyclically on every two
blades. Mode A is more stable than Mode B at a larger cavitation number
and vice versa at a smaller cavitation number. Transition from one mode to
another takes place suddenly with some hysteresis at an arbitrary Thoma
number as shown in Figure 5.14.
5.1.3 Cavitation in Pelton Turbines

The pressure in Pelton turbines would not be expected to drop below at
mospheric. Nevertheless, erosion damage is often observed on the bucket
Figure 5.16: Erosion observed on the bucket of Pelton turbines
of Pelton turbines. It is proposed that such erosion damage is caused by

two mechanisms. One is due to the cavitation induced by very high velocity
flow passing over an undulating surface, or high intensity vortices in flow
separation. The other is due to the high velocity impingement of droplets,
which causes drop erosion (sometimes called liquid impact erosion or im
pingement erosion). Cavitation erosion is a kind of fatigue failure caused
by high pressure impacts from cavitation bubble collapses and drop ero
sion is also a fatigue failure but caused by the high pressure impacts from
the droplet impingements. As their failure nature is similar (although their
causes are different), they show very similar erosive appearances which are
hard to distinguish by visual observation. The locations (Figure 5.16) where
erosion often takes place are described by Grein [5.15] and the cause of ero
sion observed at each location is classified (refer to §6.2.4).
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

226 Cavitation
5.1.4 Cavitation in Francis Pump-Turbines

In the case of reversible pump-turbines, cavitation generally appears at a
larger Thoma number in pump mode than in turbine mode. Therefore, the
submergence of a pump-turbine is usually determined based on the cavitation
performance in pump mode.
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
i i i d I i i I o
0 40 80 120 1(0
?unp dlschars* (%)
Figure 5.17: An example of cavitation performance in pumping operation of

a Francis type pump-turbine
A typical example of the cavitation characteristics of a pump-turbine in

pump mode is shown in Figure 5.17. In the partial discharge range below
the optimum discharge rate, cavitation appears on the suction side of the
blade leading edge (refer to pump mode). Efficiency deteriorates when cav
itation grows over a wide area (about two thirds of the blade pitch) on the
suction side of the blade entrance. Therefore, the critical Thoma number,
ac, is considerably lower than the cavitation inception Thoma number, CTJ,

as shown by Figure 5.17.

When the discharge is larger than the optimum discharge rate, cavi
tation appears on the pressure side of the blade entrance. In this case,
even slight cavitation may cause efficiency deterioration. Therefore, critical
Thoma number, ac, is relatively close to the incipient Thoma number, cr*.
v i : Turbine efficiency ( r e l a t i v e )
if,: Prop e f f i c i e n c y ( r e l a t i v e )
K~Q c u r v mm p a p
< I '
I I I
», « 99 %
l»-9» «
Discbarge Q (■*/»)
Figure 5.18: Characteristics of a Francis pump-turbine (Bajina Basta P.S.)

--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
When the discharge is extremely low, partial reverse flow occurs in the
impeller/runner as illustrated in Figure 5.17. The transition from uniform
parallel flow (in normal operating range) to the partial reverse flow pattern
takes place abruptly at the boundary flow rate. The flow pattern change is
very distinct and the cavitation characteristics also change discontinuously
across the boundary flow rate. In the range of partial reverse flow, high

228 Cavitation
intensity unstable vortices are induced before entering the impeller/runner,

causing vortex cavitation accompanying strong noise and vibration. There
fore, it is practically impossible for large pump-turbines to operate in this
range.
In general, cavitation in turbine mode is relatively less important than
in pump mode. However, cavitation along the blade leading edge (refer to
turbine mode) at a partial load sometimes causes erosion, especially for high
head pump-turbines.
if: inlat cavitation aroaioo
Figure 5.19: Cavitation erosion (in turbine mode) of a Francis pump-turbine
Cavitation performance in pump mode is inferior to turbine mode and it

also declines rapidly at off-design operating conditions. The optimum oper
ating head in turbine mode is usually higher than in pump mode. Therefore,
the selection of operating head-range tends to be a compromise between
these two optimum heads closer to the optimum head for pump mode and
somewhat off-design for turbine mode. Consequently, in turbine mode, the
machine always operates in a head range lower than its optimum as shown
in Figure 5.18. This means the machine operating at a higher rotating speed
than the optimum speed in terms of the speed factor, TIED- Therefore, at
such operating condition, the incidence angle becomes negative (by several
degrees), causing cavitation on the pressure side of the blade inlet edge as
shown in Figure 5.19. For high head pump-turbines, this cavitation may
cause serious damage (refer to §6.3.4).
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

5.2 General Features of P u m p Cavitation

R. K. Turton
5.2.1 Introduction
The zones where cavitation is likely to occur are described briefly, the effect
on pump normal hydraulic performance is discussed, and generally accepted
criteria for deciding acceptable cavitation behaviour are discussed, and the
role of recirculation is described. A discussion of material selection and
relevant aspects of pump detail design concludes the section.
5.2.2 General Effect on Pumps

As liquid approaches the impeller of a pump, Figure 5.20, the tangential
TOTAL ENERGY LME

--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
Figure 5.20: The variation of local static pressure in a pump suction, and
its relation to vapour pressure

230 Cavitation
component of velocity begins to increase, till at the leading edge the periph
eral velocity is approximately that of the impeller at that point. The fluid
absolute velocity thus increases, and the static pressure falls. The diagram
Figure 5.21: Areas at risk from cavitation damage
illustrates further pressure reduction on the suction side, possibly due to

wrong blade inlet angle, or flow giving lower fluid inlet angles, thus promot
ing separation of flow; if fluid angles increase above design value separation
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
will occur on the pressure side. As the pressure nears vapour pressure level
bubbles form and are swept into the blade passages where they collapse.
Figure 5.21 illustrates zones where low pressure is likely due to local velocity
peaks and on occasions in the flow range when separation can occur. Dam
age occurs downstream from the zone of bubble inception where the cavities
collapse. An early sign of damage is a "polished" appearance where bubbles
are collapsing, and later erosion and corrosion creates cavitation damage,
such as that shown in Figure 5.22. Figure 5.23 illustrates the hydraulic ef
fect of the cavitation bubbles, and it can be seen that as suction pressure
falls the flow range of the pump is progressively reduced.
The phenomenon of re-circulation in the suction zone of a pump and
between the volute casing and impeller is well-known. It can occur at
high flows, but appears to be mostly a problem at partial flow, giving rise

Figure 5.22: Severe cavitation damage on a small centrifugal pump impeller.

Reprinted by permission of the Council of the Institution of Mechanical
Engineers from Knapp R.T. "Cavitation Mechanics and its Relation to the
Design of Hydraulic Equipment". Proc. IMechE 166: 150-163, 1952.
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
PUMP
ENERGY
RISE
" ^
V
KTIC*
PRESSURE
FLOW RATE
Figure 5.23: The effect on a pump characteristic of reducing suction pressure

232 Cavitation
Figure 5.24: Recirculation in the suction zone of a centrifugal pump impeller
to flow separation. Figures 5.24 and 5.25, based on material in an article

by Bush, Warren and Karassik [5.10] illustrate the flow patterns and the
increase in NPSHTeqUirec[ associated with re-circulation, and this gives rise
to damage. It has been shown in both conventional and pumped-storage that
re-circulation between impeller and casing can cause considerable damage to
impeller blade tips, and to the casing.
5.2.3 Net Positive Suction Head

The term, NPSH or NPSE, has been used for many years to define the
difference between the local total pressure head and the local vapour pressure
head of the liquid being pumped. Two definitions are used, the system
NPSH or NPSH available (NPSHA) calculated as illustrated in Figure
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
5.26, and the required NPSH {NPSHR).

NPSHA is related to the suction system, and falls as flow rate rises, as
sketched in Figure 5.25. NPSHu is related to the dynamic action of the

MCRCASED NPSH HEMMED

AT CENTRE OF VWTCX OUNNG
SUCTION REORCUUTION
INCREASED NPSH REQURCO

AT CENTRE OF VORTEX OURMG
SUCTION REORCUUTION
lb) STANDARD PUMP

--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
Figure 5.25: The relation between NPSH, suction recirculation and cavita
tion problems

234 Cavitation
vapour
praaaura
f
Doodad
auction
loim k
pnwn
•uctton
Nit
atmoapharic
praaaura
vacuum
vapour
praaaura
loiwi
rauction
1
suction
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
NPSH.
abaoluta vacuum
Figure 5.26: A method of calculating system NPSH (based on material

published by Girdlestone pumps).

pump impeller. It may be calculated, but generally the criteria used to de
termine the "critical" value are described in the following section. A typical
curve of NPSHR is sketched in Figure 5.25. If NPSHR < NPSHA cavi
tation should not be a problem, but when NPSHR > NPSHA cavitation
becomes a problem.
5.2.4 Definition of Critical NPSHR

Figure 5.27 illustrates the way the critical NPSHR is established by test, to
satisfy such codes as the American Hydraulic Institute Standards [5.2]. The
pump is subjected to reducing NPSHA while delivering design flow rate at
the design speed. The value of NPSHA at which the pump head rise has
fallen by x% is defined as the critical level; and x is in many codes 3%.
PUMP
SPECIFIC
ENERGY
RISE
TEST PERFORMED
AT CONSTANT SPEED
ANO CONSTANT 'DESIGN' FLOW
NPSE NPSE
( CRITICAL)
Figure 5.27: The conventional method of defining critical NPSHR when

testing at design flow rate.
Cavitation and consequent damage has been found to occur at much

higher values of NPSHR than the critical value, as work published by NEL
[5.11] illustrates. The study illustrated a link between metal removal due to
cavitation and noise level, Figure 5.28. A plot of noise level against NPSHA,
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

236 Cavitation
NOISE
CAPTATION NUMBER
Figure 5.28: The relation between NPSH level and material loss due to
cavitation (based on Nel studies [5.11])
Figure 5.29, indicates that metal removal is occurring long before the critical
NPSHR is reached. Experience supports this conclusion.
5.2.5 Implications for P u m p Design

Since most cavitation problems occur in the suction zone it is the most
critical area when designing the pump. One approach to determining the
eye diameter and hub size to give a good NPSHR will be outlined. Gongwer
[5.14], Lewis [5.19] and Pearsall [5.24] among others have examined this, and
an equation which gives an acceptable NPSHR is:
V2 . _ u ?1
NPSHR = 1.8— + 0.23 - (5.1)
2g 2g
Where V0 is the meridional velocity (V0 = ^ J I ) and u is the peripheral tip
speed (u = i£^£L).
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
If eqn (5.1) is differentiated with respect to DE and equated to zero to

give optimum value

--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
f NPSE
PUMP gH
NPSE
Figure 5.29: The relation between noise and NPSER for a centrifugal pump
(based on NEL studies [5.11])

238 Cavitation
DE = K% (5.2)
K is suggested to be 4.66, so DE can be calculated. This gives a value for

U of 4.16 which agrees with Anderson's [5.3] design rule of
^ = 4
Vo
4 „ 2
NPSHRminimum = 0.001046JV3Q3 (5.3)
and the optimum non-dimensional suction specific speed will be given by

Kss = — 3 2 v vQ—j. ~ 3.25 This applies to an end suction centrifugal
60gi(NPSHR)i
pump with zero inlet whirl. Gongwer assumes a 5 % acceleration from the
eye to the vane inlet.
Two relations much used are the Thoma Cavitation number originally
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
used for Water Turbines. This relates the incipient NPSHR to the pump
head rise,
NPSHR (g NPSHR)
(5.4)
gH
The other is the Suction Specific Speed Nss or S based on 3% head drop
S = ^ V ^ (5.5)
(NPSHR)i
or
UVQ
kss = / 3 (5-6)
(g NPSHR)*
it can be shown that

s= ^ (5.7)
(74
Typical plots are shown as Figure 5.30 and 5.31 and for a centrifugal pump S
is in the range 9,500 to 10,500 (ft, gpm, rpm), with 12,500 for a condensate
machine.

1.1
o "
«■■
? ot ,'J *■
.-
1 u
§ " ,,^ : > *
„-"
V r" - * - • > - sfcfirt sutno*
I"
& 0.K -
/
/< >
J «*
" ^ - O - DOUBLE SUCTION
- -1 it
O.Ot: / '
0.03:
it
ft
0 02 -
0 10 20 3.0 I
Figure 5.30: A plot ofCTagainst characteristic number
,.*> W*
* •
<>• *-+
-**
1.0 y\ ^ >
0.1 ^yS f <*^
,_ 0.6
i
z
S 0.2
<
>
^ 0.10
0.M
0.06
0.0 V
0.02
0 1 2 3 4 5 6 7
Figure 5.31: A plot ofCTand Kss against fcss.
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

240 Cavitation
5.2.6 The Role of the Inducer

A way of improving NPSHR for a pump is to fit an inducer, Figure 5.32.
This is an axial machine fitted instead of the impeller securing nut. The
vanes follow a spiral form, with a shallow helix angle, and may have 2, 3 or 4
blades which are quite thin. The flow passages are long, and bubbles form in
the early portion of the flow path, and normally collapse before leaving the
inducer. As Figure 5.33 shows, an inducer can reduce the NPSH required
for the pump unit by at least 1 m of liquid, but away from the duty flow the
NPSH required rises rapidly. The flow from the inducer must approach the
leading edges of the impeller blades with an acceptable angle of incidence,
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
which limits the range of flow rates over which the combination gives good
NPSH required values, indicating a need to match inducer and impeller
carefully.
Inducer design is a specialised area, as experimental data is mostly from
high rotational speed rocket pumps, making the design of inducers for con
ventional water and process pumps. Reference may be made to Lakshmi-
narayana [5.18], Turton [5.30], [5.28], [5.29] , Tillner [5.27] and Oschner
[5.23].
Figure 5.32: An inducer fitted to a centrifugal pump

£ 140
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
50 100 150
ROW ( AMERICAN 9pm )
Figure 5.33: A typical inducer/pump characteristics

242 Cavitation
5.3 P u m p Cavitation Similarity

V. Chebaevsky and V. Petrov
5.3.1 P r o b l e m Nature and Similarity Conditions
Vane pumps are widely used for delivering various liquids which have differ
ent thermo-physical properties. They can be low-temperature liquid, chem
ically active liquid or poisonous liquid etc. Obviously, these liquids are not
suitable for model test. Therefore, a reliable similarity law is needed for
scaling the model cavitation data, which are obtained from the test using a
different liquid (usually water), to the prototype (actual) machine
In general, the pressure, Pc, in a cavitating volume (ie. cavity) of a
pump differs from the liquid saturation vapour pressure Pv. However, for
most actual circumstances, the pressure Pc is approximately equal to the
saturation vapour pressure Pv. This applies to normal deaerated water up
to 50° C. For such liquid, the NPSH expression can be approximated as
Ps Pv
NPSH = NPSHV
+QL + Z, (m) = (5.8) ~
2
99 9
where: Ps : Absolute static pressure of the liquid in the upstream of
the pump (Pa),
Cs: Section velocity at the point where the pressure Ps(m/s)
is measured,
Z: Vertical distance between the measure point of pressure
Ps and the highest point on the edge of vane inlet (m)
(This value of Z is often small and negligible),
p: Density of the liquid (Kg/m3),
g: Gravity acceleration (m/s 2 ).
However, pressure Pc may be much smaller than Pv for some liquids such
as low-temperature liquid, while much larger than Pv for some other liquids
such as saturated steam solution. For these cases, Pv in eqn (5.8) should be
substituted with Pc, but it is extraordinarily difficult to measure this cavity
pressure in a rotating impeller. The NPSH value, for these liquid cases,
can be evaluated experimentally. But the NPSH value obtained by the
experiment will differ if the liquid used for the test is a different one. The
relationship of the NPSH results for Pc ^ Pv with those for Pc = Pv is
NPSH = NPSH° - Aht + Ahg (5.9)

--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

Where NPSH:
Critical net positive suction head for Pc ^ Pv,
NPSHS
Critical net positive suction head obtained by using
a liquid with Pc ss Pv such as deaerated 'cold' water
(m),
A/i t = Correction value for the influence of thermodynamic
P9
property in liquid (thermodynamics correction value)
(m),
Correction value for the gas soluble in liquid (gas cor
*h* = &••
rection value) (m),
Pg: Pressure of soluble gas in the cavity {Pa).
For axial pump, these correction values are related to the Uquid proper
ties, pump geometric and state parameters as follows2.
Aht = 1.17(3.2 3Ha)io-6K°-5Re°rf w°D2b/0-25 B°™Wsmaw)0-5
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
1-A/it
(5.10)
and
0.5
A/i„ f,K (5.11)
1 - Ahg 1.17(3.2 - 3Ha)10- Re°65m°li25A0-25B0-25(^sinaw)0-5
6
. _ 27T + 7/J!,,
where: / 1 _
27r-3/3ifc
D _ 7rD m sin/?ii,
A
Relative thermodynamic correction value,
Pv
— P
Relative gas correction value,
P Gas saturation pressure in the liquid,
H = ^ Relative head of the axial pump (inducer-
type centrifugal pump),
Hai and Ha Current head and the head of axial pump
at non-cavitation regime (refer to Figure
5.34),
Number of phase transform criterion.
2
See Chebaevsky, V. and Petrov, V. (1982), 'Cavitation in High-performance Rotordy-
namic P u m p ' , Moscow, Machinostroenne (in Russian).

244 Cavitation
n — fS^y Comprehensive coefficient,

Cp(p-pv)p.T.g-
Density of saturated steam,
Pv-
Latent heat of vaporisation,
Lv:
Specific heat of the liquid,
Cp:
Absolute temperature of the liquid,
T:
Reynolds number,
Relative velocity of the liquid at the mean
diameter Dm of vane inlet,
kinematic viscosity of the liquid,
m D = £ w ^ ,
Weber number,
surface tension factor of liquid,
01b Inlet angle of vane at the mean diameter
h
V> = 1 - -sin a Blockage coefficient of the cavity to vane,
w
Thickness of vane at the end of the cavity
(if the cavity-length is not known, maxi
mum thickness of vane at the mean diam
eter Dm can be used as ($&),
Vane pitch at the mean diameter Dm,
ot„ Attack angle at the mean diameter at the
vane inlet,
f3 = C(T) RgT, Specific volume of soluble gas,
C(T) Henry's factor,
Ra Gas constant.
The pressure, Pc, in the cavity can be determined, by using eqns (5.10) and
(5.11), as follows.
Pc = PV{1 - Aht) + PgsAhg (5.12)
Therefore, the modified NPSH is
+ +z
NPSH =
99 - l (5.13)
Thus, by using above modified expression, the pump cavitation performance

has become independent of the liquid physical properties, relating only to
the vane profile, the pump geometric parameters and state parameters.
For vane pumps, a similarity theory should comply with the following
similarity criteria:
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

a)
i „ J.i
40
v*^*-*"
f)
Ahg
30
jo /
(D x J
t /»
12 NPSH, M
H, M b)
= 3f *T
Ahg /
/ / •
A ^5^1
'■%
40
fx. /
i a?
?fl
1 8 NPSH, M
Figure 5.34: Influence of dissolved gas on the suction performance of inducer-

type pump (n = 14500 rpm). (a) for g = 9 • 10~ 2 = ^ 1 " ; (b) for £ =
g.10-2 m^tn
o - deaerated water, x - water with dissolved COi-
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

246 Cavitation
1. Geometric similarity, ie LM/LN = L = const.

Here, L is the simulation scale. Parameters with subscript M and N
indicate the model and actual conditions. If an actual (ie prototype)
pump is tested in a test rig, this condition will be naturally satisfied.
2. Kinematic similarity, ie. QMlnMDM = QN/UND^ — const.

If the actual pump is tested in a test rig, then QMI^M — QN/KN ■
3. Thermodynamic property similarity (liquid), ie KM = KN = const.

Here, K is the criterion number of phase transform.
4. Viscosity similarity, ie. Re M — Re N — const.

Here, Re is the Reynolds number.
5. Surface tension similarity, ie. W6 M — We N = const.

Here, W6 is the Weber number. This similarity can be achieved by
adding some active surface substance to change the surface tension a
(refer to Fig. 5.35).
6. Specific volume similarity (gas soluble in liquid), ie. fgM = fg^ =

const.
This can be achieved by a proper selection of the model liquid, the
soluble gas concentration and the temperature TM-
It is impossible and unnecessary to satisfy all these similarity criteria

simultaneously. For most low-temperature liquids and liquids with raised
temperature, it is often the case to follow the specific volume similarity (gas
soluble in liquid) because the actual gas concentration is not large. But
other similarity criteria, say for the phase transform K, will not be followed.
Usually, clear water is the best choice for the cavitation experiment sim
ulating different liquids. Because it is not dangerous to handle, its thermo-
physical parameters can be changed easily, and the solubility of some gases
in the water is also quite high. The required viscosity of the model operating
medium can be achieved easily by heating up the water in the closed-test
system to a suitable temperature. In order to simulate the thermo-physical
parameters, some soluble gas together with a certain amount of active sur
face substance can also be added to the water. Our many years of experience
using these test rigs have proved the effectiveness and reliability of this ex
perimental approach for pump cavitation simulation.
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

50
t
o 10 a)
(Kg C 2 )
40
1
30
(1
20
vv.
80
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
o io
b)
(Kg C2)
70 /
1
\
50
\
40
V
ti
30
2
20
0.2 0.4 0.8 1.0 V%
Figure 5.35: Surface tension coefficient of water for different ASS concen
trations and temperatures, (a) ASS - Initial alkyl sulphate, T = 293K, (b)
ASS - OA - 10 ( 1 - T = 29ZK, 2- T = 353/0;
V - ASS concentration.

248 Cavitation
5.3.2 Thermodynamic Property Simulation

Following eqn (5.9) and assuming Ahg = 0, the cavitation state in the pump
can be written as
NPSHN = NPSHH2o-^htN^- (5.14)

PN-9
Where NPSH at the state of cavitation when
NPSHH2O-
the actual pump is pumping clear wa
ter (m)
AhtN = A tp-PN-9: Relative correction value of thermody
namics when the actual pump pump
ing a medium other than water.
It should be noted that the actual operating medium and the clear water
are pumped at the same rate, Q, and speed n. From eqn (5.14), the value of
NPSHN can be obtained by summing up NPSHH2O and AhtN. It is not
so difficult to evaluate NPSHH2O either by experiment or by calculation.
Therefore, the kinematic similarity of the actual pump with the model pump
is assured, and the NPSHH2O value for two pumps can be correlated as
NPSHH2O.N _ NPSHH2O.M ,,,,,

„ 2 D2 - 2 D2 ^-1{V
nN.uN nM.uM
The above method for NPSHH2O evaluation based on the experiment
data of a model pump in clear water has been widely used in practice, no
need for further discussion. However, it is much more complex to evaluate
the value of AhtN based on a model experiment.
According to eqn (5.10), K, Re and W£ must remain constant in order
to satisfy the similarity of cavitation phenomena between the actual and
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
model pumps. That is
AhtN &htM ,. 1 C .
1 - AhtN 1 - &htM (5.16)
or
AhtN = &h~tM if &h~tN « 1 and AhtM « 1 (5.17)
To ensure an accurate simulation, K must be kept constant by adjust
ing the model liquid temperature, the constant Reynolds number can be
achieved by changing the model pump speed (TIM = ^ f - ^ 4 1 ) . To satisfy the
Li "
Weber number criterion, an active surface substance can be added to the

model liquid, together with the change of speed njf, to achieve the required
value ofCTM,making He M — W§ # . The required value of OM is
■fiPM " M
OM = ON-L> —.—5-
However, an accurate simulation is not always easy to achieve. In prac

tice, it is often the case that an approximate simulation is employed using a
global similarity criterion, for instance, let
(K05Re 05
He 025)N = (K°-5Re 05
He °- 25 ) M (5.18)
Assuming a proportionality relationship of the inflow velocity W\ with the
shaft speed n, eqns (5.10) and (5.18) can be written as
L \VNJ \PM-<TNJ \KMJ

Thus, in order to maintain the cavitation phenomenon similarity between
the actual and model pumps, assuming the medium of model pump is clear
water, the shaft speed of the model pump should be
„ f t o . ! " ( H f i ) " /SLfSB.)" ( * » _ ) " (5 ,9)

L \ uN J \PH2O-°NJ \KH2OJ
However, if necessary, the shaft speed of the model pump, nô, can
be further adjusted by adding an active surface substance to change water
temperature and by altering the geometric scale L ( = DM/DN) based on
eqn (5.18).
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
There are other approximate simulations. Among them the authors have
proposed the following equality of similarity criterion,
Q
= const; Mi = ^K05 = const; M 2 = Re°'8 H e 0 2 5 - const. (5.20)
nD3 ' pg
The result is shown in Table 5.1. It is obvious from Table 5.1 that the best
medium for model pump is clear water with active surface substance (ASS)
by a comprehensive adjustment of all factors involved. By doing so, for many
actual liquids, accurate or approximate simulations can be achieved at the
temperature of the model liquid.

to
O
9
Actual liquid Model liquid Remarks
a-
nuUu A^M o
Name T,K K Name T, K TY* <7
TH7 *J
Liquidus 20 0.246 Water 600.5 0.675 k1 1 1 Accurate o
hydrogen simulation
479.5 1 1.27 0.612 1 Approximate
simulation
Water+ASS 479.5 0.788 1 1 1 2.63 "
Liquidus 107 0.52 0.75 1.75 "u
nitrogen 102 0.55 0.81 1.5 1
Liquidus 90 0.049 Water 441 1.5 1.38 0.52 "
oxygen 384.3 2.0 1.15 0.845 1 "
Water+ASS 441 1.09 1 1 1 3.6 Accurate
384.3 1.75 1 1 1 1.54 Approximate
Liquidus 74.2 1.15 0.88 1.32 "
nitrogen 75.5 1.1 0.88 1.24 "u
Tetroxide 313 0.059 Water 460.5 1 1.55 0.35
nitrogen Water+ASS 460.5 0.65 1 1 6.9 Accurate
401.5 0.92 1 1 1 4.2 Approximate
Asymetric 313 0.01 Water 367 0.9 1.54 0.375 "
xylic Water+ASS 367 0.59 1 1 6.35 Accurate
hydrazine
Table 5.1: The cavitation simulation programmes for four different prototype
2O+ASS'
liquids. Here, a — <TH°HiO
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

Cavitation Phenomena in Hydraviic Machinery 251
5.4 Cavitation Detection Techniques

E. Egusquiza
5.4.1 Introduction
Detection of cavitation in hydraulic machinery is important to optimise the
operation of the machine and to reduce erosion. During the process of cav
itation, the bubble growth and collapse produce pressure waves, generating
noise (as detailed in §3.9). If the collapse is near the solid boundary the
high-intensity short-duration pressure waves and/or micro-jet impingements
can produce erosion. The noise propagates throughout the fluid and the
structure. Therefore, the methods to detect cavitation in real machines are
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
based on the measurement and analysis of the fluid-borne or structure-borne
noise.
Various types of cavitation are different not only in their shape, which
is related to the cavitation coherent structure but also in their noise and
erosion. An important parameter for erosion is the local pressure gradient
in the zone where the bubble implosion takes place. The proximity of bubble
implosion to solid surface will depend on local flow patterns and turbulence.
The optimum detection system should determine whether cavitation
exists (onset of incipient cavitation), the type/location of cavitation and
whether it produces erosion (surfaces affected). When we detect cavitation
in an actual machine we are faced with several questions: where to measure,
which sensor to install, which signal treatment to use, and finally which
acoustic or vibratory parameter to calculate for quantifying the degree of
erosive cavitation.
Fluid-borne noise is a relatively direct method if the transducer can be
located near the cavitation spot, but the transfer function between the place
where cavitation noise is generated and the sensor changes during the devel
opment of cavitation. Moreover, cavitation noise can not distinguish whether
bubbles are collapsing close to or at a distance from surfaces. That is, there
can be a large amount of fluid-borne noise produced by bubbles collapsing
in the fluid but without erosion. It is also a disadvantage that we have to
drill a hole in the machine although much of the mechanical noise can be
eliminated by doing so. This method has been used successfully in pumps
where it is possible to locate sensors in the inlet near the blades. In hydraulic
turbines, it is more difficult to locate a pressure sensor in the right position.
Structure-borne noise may be more easily corrupted (a noise effect from

252 Cavitation
mechanical and electrical origin) and attenuated, but on the other hand,
it may be more representative of erosive cavitation. Although measuring
using vibration sensors is relatively easy the signal may be corrupted and/or
include components from sources other than cavitation.
5.4.2 G e n e r a t i o n of N o i s e a n d V i b r a t i o n
Bubbles
Imploding bubbles generate high frequency pulses with irregular duration

and amplitude in the form of spikes which are superimposed onto the existing
background noise. They radiate noise like a mono-pole source. The mono-
pole source is a more efficient radiator than a dipole (vibrations in structure)
and a quadrapole (turbulence) for low Mach numbers if the wavelength is
larger than the dimension of the source [5.26]. This is the case for bubbles
of around 1 mm diameter if the propagation speed has not been reduced by
bubbly fluid, so cavitation noise can be distinguished from the background
noise at least during incipient cavitation.
The noise generated depends on the single bubble collapse characteristics,
the number of bubbles collapsing and the radiation efficiencies (which depend
on several parameters, such as the relationship between the characteristic
dimension of the source and the wavelength).
There are two important parameters which are the intensity of the noise
generated and its frequency. For a single bubble, the intensity depends on
the pressure of the implosion which basically depends on the bubble size
and the local pressure gradient. For low void fractions, the implosion in a
cluster is higher than for individual bubbles (higher local pressure gradients);
with large void fractions the compliance of the medium increases, and as a
consequence the implosion pressure decreases. The frequency is a function
of the bubble-collapsing time-scale. The frequency of the greatest energy
density will increase with the flow velocity and the cavitation coefficient. If
bubbles collapse near the surface high frequency vibrations are generated.
Cavities
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
When a fixed cavity develops, there are basically two effects: an increase
in the low frequency noise and shielding of the high frequency noise. Low
frequency noise is produced by turbulence and instability of the interface
that separates the fluid from the vapour-gas mixture of the cavity.

Subject to the hydrodynamic conditions, the cavity can be stable or

unstable. With unstable characteristics, the fluctuations are comparable to
its length. During the fluctuations cavities (U shaped vortices) are shed from
the main cavity (referring to §6.2.1 and Figure 6.7). The amplitude of the
cavity fluctuations increase with the length of the cavity. These U shaped
transient vortices that collapse perpendicular to the structure at the end of
the cavity have an explosive nature and are very erosive [5.4, 5.8 , 5.9, 5.13].
If the frequency of the cavity fluctuation coincides with a natural fre
quency of the structure there is a hydro-elastic coupling. Then some discrete
amplitudes can appear in the broad band noise, which are associated with
the shedding frequency of structures from the cavity (vortical cavities, etc.).
These discrete frequencies can be defined in terms of a Strouhal number
St — f ■ L/U, where / is the frequency; L the cavity length and U the flow
velocity.
During the operation of a hydraulic machine different types of cavitating
structures may appear. In zones of low pressure and low pressure gradients
bubbles develop. This occurs at the blade outlet of Francis turbines (suc
tion side) and at the blade inlet of pumps working around the design point.
This type of cavitation can cause deterioration of performance and it is very
dependent on the cavitation coefficient . In the presence of a discontinuity
or with a sudden pressure gradient the development of an attached cavity
is possible (for instance, a leading edge cavitation in Francis turbines), and
cavitation bubbles and vortices form on the blade surface. During the oper
ation of a turbine, unsteady flow is created in the blades by the interaction
of the guide vanes and the runner blades, which periodically vary the angle
of incidence of the fluid on the blades. This process induces low frequency
cavity fluctuations that modulate the implosions of bubbles.
5.4.3 Propagation
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
As the noise generated by bubble collapses is of high frequency its attenua

tion is a very important fact. As generation and propagation takes place in a
closed space, the acoustic characteristics of the chamber should be taken into
account. Considering that the source dimensions are small compared with
the chamber and random incidence on the wall the classical noise formulas
can be supposed and can be used to calculate the direct and the reverber
ating acoustic field [5.16]. The absorption coefficients can be determined by
measuring the reverberation time of the chamber.
When cavitation develops, the two phase flow changes the local impedance

254 Cavitation
of the water and reduces the pressure-wave propagation-velocity. In criti

cal conditions the pressure waves can induce resonance in gas bubbles. The
sound generated by gas bubbles may be significant in fluids with gas contents
and a high nucleus concentration.
As there is some distance between the cavitation and the sensors a trans
fer function has to be determined to take into account the signal variation
caused by the propagation. The transfer function is very complex and has
to be determined from on-site measurements. The transmissibility of signals
from the blade to the measuring points can be evaluated with an exciter
located in the blade of the turbine where erosion occurs [5.25]. One possi
bility is to use an impact hammer with a hard tip although the maximum
frequency of excitation is around 10 kHz. In Figure 5.36 the frequency re
sponse function between a sensor located in the blade of a Francis turbine
and a sensor located in the turbine guide bearing can be observed.
C)
125m/
Hz RCLD 12.8k
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
400m .
mag
0 H z RCLD iJFik Hz RCLD
Figure 5.36: Transfer function between an accelerometer located in the blade

of a Francis turbine and another located in the turbine guide bearing. Signal
from an impact hammer with a hard tip. a) Input Power b) Output Power
c) Coherence (range 0-12.8 kHz)
5.4.4 Background Noise

In a circuit with a hydraulic machine, there is considerable background noise,
from hydraulic and mechanical origins. Periodic low frequency and broad

band noise are produced by the machine. This low frequency noise is pre
dominant in many cases and has to be filtered out if we want to detect
cavitation noise.
Several methods can be used. The simplest is to use a high pass filter so
that frequencies below a certain value (e.g. 10 kHz) are excluded from the
signal. This is a valid approach for the detection of bubble noise but has the
inconvenience that all the low frequency contents disappear.
Once the signal has been acquired, other methods can be used to exclude
the noise of the machine. One possibility is to use the time-average with an
external trigger synchronised with the machine shaft. The time-averaged
signal is excluded from the raw signal, therefore, only the components not
related to the rotating frequency remain. Another possibility is to calculate
and edit the spectrum to remove the unwanted components.
Another method for extracting unwanted noise is adaptive noise can
celling with two transducers. One is located near where the cavitation takes
place and the other at a some distance where the high frequency cavitation
noise has been damped out.
5.4.5 Cavitation D e t e c t i o n in Frequency D o m a i n
There are several techniques available to detect cavitation using the signals
measured with sensors located near the cavitation place: in the frequency-
domain, in the time-domain and with time-frequency transforms.
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
High Frequency Signals
In the frequency-domain, cavitation shows a broad band pattern especially

in the high frequency range. The high-frequency part of the power spectrum
increases with the appearance of cavitation, as can be seen from Figure 5.37.
The variation of pressure fluctuation measured with a piezoelectric pressure
transducer in a model Francis turbine with different values of cavitation
coefficient has been demonstrated.
Cavitation signals have been measured to include frequencies of more
than 1 MHz. As industrial accelerometers have maximum resonant fre
quencies of around 100 kHz, they are not useful for detecting high frequency
vibration. Acoustic emission (AE) sensors have a broad band resonant re
gion and can measure vibrations of more than 500 kHz. Figure 5.38 shows
the spectra of the signals measured with an acoustic emission sensor located
in the guide vane of a Francis turbine with cavitation problems. There is an

256 Cavitation
100
i i i i : i i :
! ! ! ! ! ! ! !
12.5 • i i i i i i i
i i i i t i i i
i i i i i i i i
/div i i i i i i * i
i i i i i i i i
i i i t i i i i
i i i • i i t i
i i i i i i i i
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
._ r _ _ T _ _ ^ _ T f~~T ~t 1 ~
1 1 1 1 1 1 1 1
< 1 1 1 1 1 1 1
1 1 1 1 1 1 1 1
mag 1 1 1 1 1 1 1 1
_ _ r _ _ T j _ T y , _j. j _ —r~ "
1 1 1 1 1 1 • 1
rms
"TITTM IT —,
Po -4-4-4-4-/ \4-4- -1—

• —
! ! ! Jf <*\l i
i . i i .. .i . .' j i i....
Hz
30k Lin Hz 81.2k
Figure 5.37: Variation of high frequency pressure signals in a model turbine

for different values of the cavitation coefficient. a\< a 2< a 3 (Range 30-81
kHz)

Amplitude Spectrum
20E-3
20000
Amplitude Spectrum
20E-3
20000
Figure 5.38: Comparison between the power spectrum of an AE sensor lo

cated in the guide vane of a Francis turbine at different loads, (a) 80% load
(b) 100% load (range 0-200 kHz)
increase in the acoustic emission when the machine increases the load from
80% to 100%.
Overall rms Levels of a High Frequency Band
This is one of the simplest approaches. The rms of vibration signals af
ter high pass filtering (or for a given frequency band) can be a significant
parameter for the use of cavitation detection (at least 15 averages). Mean
square values of acceleration signals in different frequency bands (5 to 45
kHz, 15-35 kHz, etc) have been used to map cavitation regions [5.8, 5.9].
Some of these bands represent the degree of cavitation and can be correlated
to the cavitation intensity.
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

258 Cavitation
Sound Pressure Levels
Sound pressure levels (Lp) and sound power levels (Lw) have been proposed
by many authors. Normally the Lp value of band filtered fluid-borne noise
is the variable that is measured. Both Lp and Lw in pump inlets can be
related with cavitation intensity (assuming a plane wave propagation in the
inlet duct of pumps). It is useful to know the cavitation-source strength,
which is related to the erosion process [5.22].
For pumps, the relationship of Lp with cavitation coefficient is as fol
lows. With incipient cavitation, the noise level Lp increases, with cavitation
developing it reaches a maximum and finally it decreases again. The de
crease is caused by the shielding of the diffused air bubbles, which reduce
the implosion pressures.
The value of Lp at 40 kHz of 1/3 octave band has been correlated for
pumps [5.21]. A cavitation inception number was defined and correlated
with the tip speed of the pump, as well as a peak noise power ratio to pump
impeller tip speed. Using the cavitation inception number, the presence of
cavitation could be detected and, if so, the peak noise correlation should
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
assist in determining the cavitation degree in the pump.
Low Frequency due t o a Cloud of Bubbles or t o a Cavity
With a cavitation cloud, there is a change in the density of the cavitating

area and the hydraulic circuit becomes a two-part system with two different
frequencies. A low frequency pressure fluctuation is generated in the cavitat
ing area. With a continuous decrease in a the amplitude of the fluctuation
increases to a maximum and then decreases. The maximum occurs when
the natural frequency of the cavity approaches the natural frequency of the
circuit. This is an indirect method of detecting the appearance of cavitation
[5.20].
In Francis turbines with draft tube vortex cavitation at part loads, the
cavities exhibit characteristic low frequency fluctuations. When the cavita
tion core develops, the spiral vortex formed is easily detected because there
it generates a pressure fluctuation that occurs at around 30% of the runner
rotation speed (Figure 5.39).
The amplitude of the pressure pulsations generated by a machine at the
shaft and blade passing frequencies are affected by the degree of cavitation.
These signals also vary, depending on the transducer locations (eg. inlet
pipe, outlet pipe, or volute). These signals can be used diagnostically for

a)
.4 ntMBWa
Figure 5.39: (a) Spectrum of the low frequency pressure pulsation generated
by the draft tube vortex in a pump-turbine rotating at 600 rpm (10 Hz).
Draft tube vortex appears at a frequency of around 3 Hz. Sensor located in
the draft tube; (b) Coherence between this signal an the signal of another
pressure sensor located at 90° in the draft tube.
some types of pump, but the accuracy of this approach is low because the
effect of cavitation can be difficult to distinguish from the possible effect
of other factors which can also produce similar symptoms, such as some
damage (or geometric variation) to the machine rotating parts.
5.4.6 C a v i t a t i o n D e t e c t i o n in T i m e D o m a i n
Analysis in the time domain can also be used to detect high frequency cav
itation noise. The detection can be in the form of pulse counting (acoustic
emission), statistical distributions, etc.
High Frequency Signals

By observing an instantaneous high-pass filtered time-domain pressure sig
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
nal, the pressure bursts produced by cavitation can be detected. At the

onset of cavitation, when there are few implosions, the root mean square
value (rms) (which represents the energy of the signal) will remain virtually
unchanged. However, the peak value will increase considerably, therefore,
the crest factors (peak/rms value) are more significant than the rms values
for incipient cavitation. Statistical moments can also be used to detect the
change of the distribution.

260 Cavitation
The short-impulsive-stress-waves caused by cavitation pulses will propa

gate throughout the structure. Acoustic emission techniques, such as pulse
counting etc, that capture these high frequency transient signals have been
used successfully to detect cavitation. The process consists of counting the
number of occasions that a signal (high amplitude short duration pulses)
exceeds a predetermined level (a threshold to rms level) for a given period
of time. The number of events are represented as a function of its amplitude
[5.6, 5.7].
Synchronised T i m e D o m a i n Signals
Cavitation affects the periodic noise generated by the machine. When cavi
tation starts to develop inside the impeller the circulation around the blades
is modified and so are the pressure pulses generated by the impeller. Av
eraged time-domain signals with the aid of a shaft synchronised trigger are
representative of the pressure pulses produced by each blade. When cavita
tion appears the rms and shape of each pulse change, which can be detected
by some pattern recognition method [5.12] (Figure 5.40). This method is
useful for developed cavitation in pumps.
l<ffl«t« U T K M 7 | MtaCV«4
ttracvi? KT»CVI4
1 * 12S 3 4
s' 3
z (
-.1
-1
* .•
RMS
Figure 5.40: Pressure pulses measured in a centrifugal pump after averaging

with a shaft synchronised trigger. Pump with 6 impeller blades rotating at
3000 rpm. a) No cavitation detected. The 6 pulses produced by the pump
can be easily seen, b) Developed cavitation detected. It can be observed
that the variation in the shape of the pressure pulses when decreasing the
cavitation coefficient. Comparing the signals, the rms of each pressure pulse
and the normalised cross correlation coefficient (NCCC) change considerably.
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

D e m o d u l a t i o n Techniques
As fluctuating cavities seem to generate the most erosive power, it is these

that it is most important to detect. In hydraulic machinery the incidence
angle of the flow entering the machine varies as the runner rotates, because
of the runner-diffuser interaction and/or the pressure differences between the
top and the bottom of the machine in low head machines. This makes the
cavity at the leading edge of the blade change continually, leaving imprints
of cavitation bubbles at the end of it. The greater the unsteadiness of the
flow is, the more likely the erosion is to occur.
Because of this fluctuation, the cavitation signal is modulated by the
blade-passing frequency. Accelerometers mounted on the hydro-turbine hous
ing can detect acoustic emissions. If cavitation exists the demodulated signal
will exhibit discrete frequency tones at the blade and shaft-passing frequen
cies. The wicket-gate link and the turbine guide bearing mountings have
been found to be useful in detecting this type of cavitation [5.1].
Full wave rectification spectral analysis or Hilbert transforms can be used
to demodulate the high pass filtered signals produced by cavitation. Rec
tifying the waveform introduces a new low frequency component. Because
averaging is needed, it is necessary to store a long signal string. The need for
a large storage capacity in a computer is a disadvantage when working with
high frequency signals. Therefore external demodulators are preferred when
monitoring cavitation in machines. One possibility to avoid this is firstly to
average the signal with a shaft trigger and then to demodulate it.
Demodulation seems to be one of the best methods although quantifying
the intensity and erosion rate using this method is not easy. In an actual
machine the demodulation produces different values subject to the sensor
position and the selected frequency band. Figure 5.41(a) shows the demod
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
ulation in a frequency band around 60 kHz of the signal from an AE sensor

located on the guide vane of a Francis turbine. Peaks at the blade passing
frequency fb and harmonics (especially 2 fb) can be detected. The rms
of the demodulated spectrum is a useful parameter to quantify cavitation
(Figure 5.41(b)). At maximum load the rms is high. Reducing the load to
80%, the peaks disappear.
Figure 5.42 shows the result of a demodulation for different frequency
bands. It can be seen that the selection of the appropriate frequency band
for demodulation is important to the success. Therefore, it is necessary to
analyse the variation of power spectrum with loads and on-site measured
transfer functions. Moreover, the demodulation of the same frequency band

262 Cavitation
^f
fj W A/
(0 M 100 120 140 It* IM 200
(a) (b)
Figure 5.41: a) Demodulation of the spectrum of Figure 5.38b (AE sensor

on a Francis turbine guide vane). The machine rotated at 250 rpm, the
runner had 15 blades and there were 25 guide vanes. The blade passing
frequency is fb = 62.5 Hz. Peaks at 62.5, 125 and 187.5 can be identified
in the demodulation spectrum, b) Variation of the rms of the demodulated
spectrum with load.
uaa u 1—£ 1 . a—. . 1 arj ra S I jîr i a !-"* a ,

^
• .«■
nl /aw hi
Lw 1
■J *»/
'r hfc
" |rv
"V 111/ toV^NiM
fllV
•v iiyV "V ^ -\y\ w r*\ lyv\ * \ J fW^
«.• i - s r i7E i—an r ~ '— ' L
¥BT ■•■V »—» T T B ( i—XI r TWI
Figure 5.42: Demodulation of the signal of an accelerometer located on the

turbine guide bearing of a Francis turbine with inlet cavitation. a) Band
5-10 kHz, fb and 2 fb can be observed, b) Band 10-15 kHz. The blade
passing frequency can not be observed.
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

but obtained from accelerometers located in different positions will result in

different spectra. Using frequency bands higher than 50 kHz, the results
are almost independent of the band selection.
Dual Channel Analysis

Dual channel analysis, such as correlation and frequency response, can be
used to improve the identification process of cavitation, by comparing two
signals from different sensors located in the same position or the same type
of sensors located in different positions. For example, by calculating the
coherence of the signals from a pressure transducer and an accelerometer
located in the same position, it is possible to exclude the noise of mechanical
origin.
5.4.7 Cavitation Detection with Time-Frequency Analysis

One difficulty in cavitation detection is the variability of signals. Time-
frequency analysis transforms (Wigner-Ville, wavelets, etc) are useful tools
to get better understandings about what we are measuring and when we
should detect cavitation in an actual machine.
Figure 5.43 shows the signal variation measured by an acoustic emission
sensor for two operating conditions in a hydraulic turbine. At 80% load,
there are only small pulses with low amplitudes (Figure 5.43a). At maximum
load, the high frequency vibrations caused by implosive cavities can be seen
in the spectrogram (Figure 5.43b) [5.31].
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

264 Cavitation
m I IMtorit ^MCttUHM
100 oo-
4.0E+O-
20E*0-
mmmmmm—m
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
O0E*0-^fMMMH
2.0E«0-
-4.0E«O-
0.000 0.010 0.020 0030 0040 aosoittcj
w|tK^HI tp<rtMW|
00.00-
50.00-
,1 »
0O00-R.
5000-
-4.0E«0
O100 0110 0120 0130 0140 ai50(wc)
Figure 5.43: Time-frequency analysis of signals of the machine described

in Figure 5.40. Frequency band 0-200 kHz, AE sensor on a guide vane a)
Machine at 80% load b) Machine at 100% load.

References
5.1 Abbot, P.A. et al. (1986), 'Cavitation Monitoring of Two Axial-Flow
Hydroturbines', International Symposium on Cavitation and Multi
phase Flow Noise, ASME Winter Annual meeting, FED Vol 45, De
cember 7-12, pp 7-15.
5.2 American Hydraulic Institute Standards (1976)
5.3 Anderson, H. (1980), Centrifugal Pumps Trade and Technical Press.
5.4 Avellan F, Dupont Ph (1988), 'Cavitation Erosion of Hydraulic Ma

chines: Generation and Dynamics of Erosive Cavities' Proc. 14th IAHR
Symposium, Trondheim Norway, pp 725-738
5.5 Avellan F, Farhat M. (1988) 'Shock Pressure Generated by Cavitation

Vortex Collapse', Proceedings Inter. Symp. on Cavitation Noise and
Erosion, ASME Winter Annual Meeting , FED 88 pp 119-125.
5.6 Beyer, J.R., Smith, T. G. (1982), 'Acoustic Emission Characteristics of

a Model Francis Turbine under Cavitating and Non-Cavitating Condi
tions', Proceedings International Symposium on Cavitation Noise, The
Winter Annual Meeting, ASME, November 14-19, pp 87-92.
5.7 Beyer, J.R. (1986), 'The Use of Acoustic Emission to Detect Cavita
tion in Model Prototype Applications'. International Symposium on
Cavitation and Multiphase Flow Noise, FED Vol 45„December, pp
7-12.
5.8 Bourdon, P et al. (1990), 'Vibratory Characteristics of Erosive Cavita

tion Vortices Downstream of a Fixed Leading Edge Cavity', Proceed
ings Inter. Symp, IAHR, Belgrade, H-3
5.9 Bourdon P., Simoneau R., Avellan F. (1993), 'Erosion Vibratory Fin
gerprint of Leading Edge Cavitation of a NACA Profile and of a Fran
cis Model and Prototype Hydroturbine', FED-176, Bubble Noise and
Cavitation Erosion in Fluid Systems, ASME.
5.10 Bush, A. R. et al (1976), 'Coping with Pump Pulsations etc', Pump

World (Worthington Pump Inc), 2(1).
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

266 Cavitatton
5.11 Deeprose, W, M et al. (1994), 'Cavitation Noise, Flow Noise and

Erosion', IMechE Conference 'Cavitation', Heriot-Watt, Edinburgh,
September.
5.12 Egusquiza, E. et al. (1987), 'Cavitation Detection in Pumps by Signal

Analysis', Euromech Colloquium 222, Wageningen June 1-3, Paper 5.7.
5.13 Farhat, M. et al (1992), 'Pressions Instationnaires Generees par une

Poche de Cavitation Partielle', Societe Hydrotechnique de France, 2me
Journees Cavitation, March 18, pp (1-20)
5.14 Gongwer, A. C. (1941), 'A Theory of Cavitation Flow in Centrifugal

Pump Impellers', Number 63. Trans ASME.
5.15 Grein, H. (1990), 'Cavitation Pitting and Rain Erosion of Pelton Run
ners', Proc. IAHR Symposium, Paper H6, Belgrade.
5.16 Guelich J. F. (1990), 'Cavitation Noise in Centrifugal Pumps', Vibra

tion and Wear in High Speed Rotating Machinery, Kluwer Academic
Publishers, pp 87-112.
5.17 Ishii, Y. (1962), 'The Two Types of the Runaway State of Propeller
Turbine under Cavitation', Proc. IAHR Symposium, Sendai, pp 247-
268.
5.18 Lakshminarayana, B. (1982), 'Fluid Dynamics of Inducers-a Review',
JNI Fluid Eng, Vol 104, pp 411.
5.19 Lewis, W. P. (1964), 'The Design of Centrifugal Pump Impellers for
Optimum Cavitation Performance', EMG (2), Eng. Trans. Inst. Eng.,
(Aust).
5.20 Li, S. C. (1992), 'Pressure Fluctuations in Cavitating Draft-Tube
Flows', FED 136 ASME , pp 1-6
5.21 McNulty, P. J. (1986), 'Monitoring the Presence of Cavitation in
Pumps by a Single Fluid-Borne Noise Measurement', Proceeding In
tern. Conf. on Condition Monitoring, BHRA, May 21-23, pp 167-182.
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
5.22 McNulty, P. J., Pearsall, I. S. (1982), 'Cavitation Inception in Pumps',

Journal of Fluids Engineering, March, Vol. 104, pp 99-104
5.23 Oschner, K. (1998), 'Stand der Technik be Inducern fur Kreielpumpen

Chemis-Technik', 17:174-80.

5.24 Pearsall, I. S. (1973), 'Design of Pump Impellers for Optimum Cavi

tation Performance', Proc IMechE, 187, pp 667.
5.25 Pereira, F., Farhat, M., Avellan, F. (1994), 'Dynamic Calibration of

Transient Sensors by Spark Generated Cavity', Bubble Dynamics and
Interface Phenomena, Kluwer, pp 227-240.
5.26 Ross, D. (1987), Mechanics of Underwater Noise, Peninsula Publish

ing.
5.27 Tillner, W. (1990/1993), 'Avoidance of Cavitation Damage', Original

published in German by Export Verlag (1990); English translation
edited by Turton, R. K. published by Mechanical Engineering Publi
cations Ltd, London (1993).
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
5.28 Turton, R. K. (1986), 'Studies of Inducers for Centrifugal Pumps',

IAHR Symposium, Montreal.
5.29 Turton, R. K., Furukawa, A. (1986), 'Studies of the Cavitating Be

haviour of Inducer for Conventional Centrifugal Pumps', IAHR Sym
posium, Trondheim.
5.30 Turton, R. K., Tugen, Z. (1984), 'Use of Inducers as a Way of Achieving

Low NPSE Values from a Centrifugal Pump', World Pumps, Part II,
September.
5.31 Vizmanos C , Egusquiza E. (1996), 'Cavitation Detection in a Francis

Turbine', Modelling, Testing & Monitoring for Hydro Powerplants II,
HydroPower & Dams, Lausanne.

--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

Chapter 6
Cavitation Damage to
Hydraulic Machinery
6.1 General Mechanism of Cavitation Damage

Y. Iwai and T. Okada
6.1.1 Introduction
The impact pressure generated by bubble or cavity collapses in fluids has
been extensively studied. Some results from these theoretical and experi
mental investigations are summarised in Table 6.1 [6.29, 6.18, 6.20, 6.36,
6.24, 6.11 and 6.49]. These figures show that the impact pressure can reach
values from several thousands to several ten thousands atmospheric pres
sure. At present, there are basically two different viewpoints towards the
mechanism about the erosive-power generation. One attributes it mainly to
the impact pressure produced by collapsing bubbles; the other attributes it
to the shock wave and the micro-jet produced by collapsing bubbles.
6.1.2 Cavitation Bubble Collapse Pressures and Damage

However, it has been generally agreed by many researchers that cavitation
erosion is a mechanical failure caused by the impulsive pressure generated
by the cavitation bubble collapses. Nevertheless, the detailed information
about the magnitude of pressures and the process of surface fracture caused
by these impulses are still not available. It is still an ongoing research sub
ject. Cavitation in hydraulic machinery often appears in the cloud form
269
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

270 Cavitation
Methods for
Investigator Bubbles calculation or measurement Results Ref.
of collapse pressure
Spherical 1260 atm at the stage

Rayleigh Empty bubble incompressible of 1/20 of initial radius [1]
Hickling Gas bubble

Spherical 4
and Initial gas pressure: <2 X 10 atm
compressible [2]
I Plesset
Ivany
3
Pg=10 " atm
Gas bubble
Initial gas pressure: Spherical 6.77 X 10 4 atm
and 3
Pg=10 " atm compressible [3]
Hammitt Pg-10 "* atm 5.82 X 10 5 atm
Plesset Based on
and Vapour bubble micro jet velocity 2X 1 0 3 atm [4]
Chapman
Jones
Spark-induced Piezoelectric pressure-bar
and 1 0 4 atm [5]
hemisherical bubble guage
I
Edwards
Fujikawa Time duration:
f and
Akamatsu
H2 gas in
a water shock tube

Pressure guage
Holographic interferometry
2-3 u s
1 0 4 -10 5
atm
[6]
i Tomita
and
Shima
Spark-induced
bubble
Pressure transducer
Photoelasticity
Several 10 MPa [7]
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
Table 6.1: Bubble collapse pressures

Cavitation Damage to Hydraulic Machinery 271
consisting of many cavitation bubbles. When such a cloud collapses on or

near the face of a boundary, the surface is exposed to such bubble-collapse
pressures that possess different magnitudes. This is because the bubble size,
shape and the collapsing location etc are all different. Each of these im
pulses acts in a localised tiny area and over a very short duration. The
pressure acting area and the duration are important factors which affect the
mechanism of cavitation damage. Nowadays, some measuring techniques
enable us to estimate these collapse pressures to a certain accuracy. They
are the acoustic-impedance approach by using a hydrophone installed near
the cavitation cloud [6.38 and 6.2], the counting method of the cavitation
pitting rate on a piece of aluminium or mild steel attached on a testing body
[6.26 and 6.41], the technique of measuring the impact pressures by a pres
sure sensitive film [6.25] and the pulse-height-spectrum approach by using a
pressure-bar probe [6.8].
As an example, Figure 6.1 shows the distributions of accumulated counts
of pulse height (impact load) generated by the cavitation bubble collapses
in a venturi facility [6.21]. The pulse height is a maximum value of the
impact load produced in a sampling interval of 110 psec sensed by a pressure
detector (made of piezoelectric ceramic) with a measuring area of 2 mm
diameter. The sample size is 2.69 x 106 pulse heights taken in 5 min. A
general trend can be extracted from these studies. That is, the number of
large impact load always decreases exponentially although their distribution
curves are different subject to their own cavitation conditions.
Figure 6.2 shows the pits on the eroded surface of a mild steel exposed
to cavitation for 1 hour. On a ferrite with body-centred cubic lattice, micro
fractures occur on the surface caused by the very large impact pressures,
resulting in pits. The distributions of pit size on the mild steel, which are
measured in the same area and duration as those impact pressures are mea
sured, are also shown in Fig.6.1. The minimum observable size is about 1 pm.
The distributions of pit sizes resemble the pulse height distributions. The
number of large impact loads which cause pits are remarkably small. The
investigation of pits on a soft aluminium specimen by Knapp [6.26] shows
that only one in 3000 of the travelling cavities causes damage. Robinson
and Hammitt [6.41] also state that only one in 104 of the collapsing bubbles
forms a pit on the surface. Therefore, the severe cavitation damage may not
simply proceed from the accumulation of these impact fractures.
The slip lines and the network of the cracks in the ferrite grains exposed
to cavitation in a vibratory device, which is the most common accelerated
erosion test, is shown by Figures 6.3 and 6.4 [6.34]. The rate of erosion-pit
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

272 Cavitation
Impact load N
40
10,71.1 10 20 30
o (7=0. 68 Pulse distribution
ov=25m/s"]
• 30m/s > Venturi
» 35m/sJ
Pit distribution (Mild steel)
D v=25 m/s "I
■ 30 m/s f Venturi
H 35 m/s J
0 20 40 60 80 100
Pulse height
I I I L.
0 2 4 6 8 10
Pit size ii m
Figure 6.1: Distributions of accumulated counts of pulse height and pit size
on the mild steel surface at various velocities in the venturi facility
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

Figure 6.2: Scanning electron micrograph of pits on the surface of a mild

steel in the venturi facility
Figure 6.3: Slip lines on the ferrite grain of a mild steel (vibratory device;
frequency: 22 kHz; amplitude: 40 pm; 25° C ion-exchanged water; test
duration: 10 min)
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

274 Cavitation
Figure 6.4: Plastic deformation of ferrite grains (vibratory device; frequency:

22 kHz; amplitude: 40 fim; 25° C ion-exchanged water; test duration:
15 min)
production becomes smaller with exposure time owing to the work hardening
on the ferrite grains. On the other hand, many slip lines appear on the grains
along these pits caused by the repeated small impact pressures which are
below the critical value that can impulsively form a pit. Subsequently, the
fatigue cracks initiate along these slip lines and large erosion particles fall off,
resulting in the characteristic spongy erosion-surface due to this fatigue-type
fracture.
For copper and aluminium with a face-centred cubic lattice, pits form
slightly but with a large number of indents because these materials are more
likely to undergo plastic deform. However, in these cases, the repeated small
impact pressures below the value that can impulsively forms a indent also
contribute to the cavitation erosion. Figure 6.5 shows a series of photographs
of the progression from the virgin surface to the severely damaged one [6.34].
On the surface exposed to the cavitation in the vibratory device for only 20
seconds (Fig. 6.5(b)), the metallographic structure has become clear due
to the mutual slide along the grain boundaries. In Fig. 6.5(c), the surface
becomes rough because of the local occurrence of large indents; the slip lines
appear on each grain with different directions; and, the secondary slips form
and the plastic deformations rapidly develop. This process is repeated and
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
Figure 6.5: Cavitation damage of the surface of a copper (vibratory device;

frequency: 22 kHz; amplitude: 40 /xm; 25° C ion-exchanged water)
(a) Virgin surface; (b) Test duration 20 n sec; (c) Test duration 60 LI sec;
(d) Test duration 180 /i sec.

276 Cavitation
the large erosion particles keep falling off due to the fatigue-type fracture.
As to the relationship between the cavitation bubble collapse pressures
and the erosion, De and Hammitt [6.8] showed a linear relationship between
the acoustic power of collapse pressures and the erosion power based on
the calculated weight loss. Recently, a new approach has been reported
[6.22]. That is, the comparison of the distributions of the impact pulse with
the hypothetical S-N curves of fatigue shows that the relationship follows
the Miner's law of fatigue fracture, regardless of cavitation conditions and
materials. Therefore, it may be appropriate to state that the cavitation
erosion at more advanced stage is mainly due to the accumulation of the
micro fracture fatigue caused by the repetition of the small impact pressures.
6.1.3 Correlation between Erosion Resistance and Mechani

cal Property
The correlation between the erosion rate and certain bulk mechanical proper
ties of materials is being extensively investigated. The erosion resistance (for
example, the reciprocal of volume loss rate) does not show a good correlations
with a single mechanical property such as tensile strength or hardness. How
ever, a much better correlation can be obtained based on the proof resilience
(= 1/2 x g S l ), or ultimate resilience (= 1/2 x f ^ £ ) or
work done to cause fracture (= (proof stress + 2/3 x (tensile strength —
proof stress)) x elongation), which means the strain energy to fracture
[6.19 and 6.16]. However, no unified relationship over a wide range of ma
terials has been found yet. This is presumably because the fact that the
bubble collapse acts in a very tiny area, which makes some particular ma
terials deviate from the correlations based on the bulk material properties.
As an example, the materials containing brittle structures (for example, cast
iron) come off the relationships because the remarkable decrease in their mi-
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
croscale strength compared with the bulk strength [6.33]. In addition, the
materials with high ability to work hardening such as stainless steel also de
viate from these correlations because their surface micro-structure changes
owing to the repeated action of the small impact pressures.
The corrosiveness of liquids also affects the micro-scale strength and en
hances the mechanical damage compared with the electrochemical dissolu
tion [6.32]. Therefore, the corrosion sensitivity of the material is an impor
tant factor for the cavitation damage in corrosive liquids.

6.2 Cavitation Damage in Turbines

S.C. Li
6.2.1 Introduction
The material damage of machine components caused by cavitation 1 will
• deteriorate machine performance, such as losses of efficiency and out

put;
• introduce additional hydraulic oscillations and mechanical vibrations;
• cause lengthy downtime and expensive repairs, and incur some indirect
costs such as the loss of capacity availability by unscheduled outages
and the reduction of machine life by repeated extensive repairs.
The accurate prediction of cavitation damage is still impossible. However,

it is known that the damage rate is subject to the following three factors:
• Erosive Power of Cavitation

There are basically two types of erosive cavitation in hydraulic ma
chines, i.e. travelling bubble cavitation and vortex cavitation 2 ,which
cause damage. Their location and severity depend on machine type,
design and manufacturing quality as well as operating conditions, set
ting level and air content in the water etc.
• Material Resistance
The mechanism of cavitation damage depends on the type and crystal
structure of the material. The cavitation resistance is also affected
by the heat treatment and machine finish involved in manufacture
and repair processes and by the status and level of stress within the
material.
• Contents of Liquid Medium

Apart from the air content, solid particles and acid content etc in the
liquid will also affect the process of cavitation erosion, and usually
enhance the cavitation erosion through synergism mechanism.
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

278 Cavitation
X ~ X/Lch
Smooth & transparent _ , . X/Lch
interface Pulsation at closure Wavy & unstable V o r t e X S h e d d i n ' g
interface :—.
3 3 Strong pulsation
at closure
Stable sheet cavity, smaller value of OCi Unstable sheet cavity, larger value of CCi
Sheet cavitations CCi = ^°~ as
0 5 1
Cp Cavitating vortices
-1.0
X/Lch
Travelling bubbles Collapsing area
Travelling cavitation Separation cavitation

0Ci*0O-3° CXiXXs
Figure 6.6: Four types of cavitation (the incidence of the foil, aj, is defined
according to the zero-lift line; aa is the incidence at which the flow separation
takes place; only the pressure distributions on the upper surfaces are shown)
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

The four basic types of cavitations which may develop on a cavitating 2D

hydrofoil are briefly depicted here to explain the correlation of erosive power
with cavitation pattern. The type of cavitation developed depends mainly
on the characteristics of pressure distribution if there is no cavitation, as
shown in Fig. 6.6. When the incidence, a*, is small, for say a< = 0° — 3°,
the pressure distribution on the suction side presents a smooth variation
without negative pressure peaks 3 . The cavitation developed under such flow
condition will be travelling bubbles. The explosive growth of bubbles starts
from the beginning of the low-pressure region and subsequent collapse takes
place in the pressure-recovery region. The incipient cavitation number, a,-n,
is strongly influenced by the air content. The bubbles collapse rather ran
domly in both temporal and spatial domains, producing weak erosive power.
When ai increases to the range of ai = 3° — a s , where as is the value at
which the flow separation takes place, a negative pressure peak appears at
the leading edge. Then, sheet cavitation will develop at the leading edge with
a value ofCTJ„being higher than that for travelling bubble cavitation and not
susceptible to the air content. When the value of aj is relatively small, for
say ai = 3° — 4°, the magnitude of the negative pressure peak is small and
the sheet cavity developed is thin and stable with smooth and transparent
interface, sometimes possessing a small pulsating tail at its closure. This
type of cavitation produces little erosive power except some damage capac
ity with its pulsating closure. When the value of ai is large, the cavity will
develop into a thick and unstable one with a strong pulsation at its closure.
Its unstable and wavy interface is highly turbulent, shedding cavitating vor
tices. This is the most erosive type of cavitation. When a* > as, owing to
flow separation a free shear layer structure results within which a Karman
vortex cavitation can be generated at a much higher value of CTjn owing to
strong vorticity. The cavitating vortices are entrained away from the foil
and, therefore, have little damage potential to the foil itself, but may cause
severe damage to other components if they collapse nearby.
As to the mechanism and erosive power of such unstable and erosive sheet
cavitation, a mechanism, similar to the boundary shear flow, explaining the
' T h i s cavitation damage is also referred as cavitation pitting or cavitation erosion.

2
Sheet cavitation is not included here because if the sheet cavitation is smooth and
transparent it has little erosive power to the material; while it becomes erosive, only its
vortex portions are responsible for the damage. See the following part of this subsection.
3
T h e corresponding pressure distribution pattern for a given value of Oi is subject to
the geometry of the foil. T h e values of a> cited here are only for the purpose of conceptual
explanation.
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

280 Cavitation
instability development and vortex shedding was proposed by Avellan et al,

1991 [6.4], and Dupont and Avellan, 1991 [6.10], which postulates that, ow
ing to the strong shear stress on the cavity surface, the interface of the sheet
cavity plays a similar role as a solid boundary does (referring to Fig. 6.7).
The cavity detachment starts from the spanwise spots forming cone-shape
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
Vorticity lines |iop vl«w|
Figure 6.7: Conceptual sketch of unstable leading-edge cavitation on a 2D

hydrofoil (Avellan, Dupont & Farhat, 1991 [6.4])
joints downstream. With small incidence, say ai < 3°, and small Reynolds
number, the cavity is stable and most of the interface is smooth and transpar
ent. When the values of incidence and Reynolds number increase, the cavity
spots move forward with smaller spacing and the point of transition from
a transparent surface to a wavy/unstable surface, which might be caused
by the transition to turbulence, also moves forward accompanied by violent
pulsations at the cavity closure. Owing to the Kelvin-Helmholtz instability,
the spanwise vorticity lines, which are strongly perturbed by the streamwise
rolling-up of the cavity cones, are transformed into U-shape vortices, which
are lifted off the surface and shed at the Strouhal frequency /
/ = ^ S . (Hz)
'c

where Uoo '■ upstream flow velocity (m/s);

lc : cavity length (m) which can be approximately expressed
38
L = F(ZZc7r')> h e r e ' L 1S t h e foil c h o r d l e n gth, Oin
is the incipient cavitation coefficient, a is the cavitation
coefficient, and aj is the incidence;
St : Strouhal number, valued approximately as 0.3.
The cavity formation in the core of U-shape vortices is due to the vorticity
increase resulting from the vortex stretching and/or the vapour entrained
from the main cavity. The damage area on the foil, which is caused by the
collapse of whirl vortices generated from the main cavity closure, is concen
trated downstream of the closure. The erosive power of such an unstable
cavity is highest, and is at least an order of magnitude greater than that of
its stable counterpart. The focusing collapse of the bubble cloud 4 in the
cores of the vortices under high pressure recovery immediately downstream
of the main cavity is responsible for the production of this very high ero
sive power. According to Avellan & Dupont, 1988 [6.3], the impact pressure
produced by such cavitating vortex collapse could reach 1950 — 2540 MPa
which is capable of damaging any industrial materials, even Stellite. Avellan
et al, 1991 [6.4], estimate the erosive power, P, produced by such a leading
edge cavity to be
P = \pF (Cp,max + a) StUllc (W) (6.1)
where p: liquid density (kg/m3);

F{CPtmax + cr) : a characteristic function of the hydrofoil rep
resenting the influences of cavitation number
a and incidence, here Cp<max is the maximum
pressure coefficient;
St : Strouhal number relating to vortex shedding.
However, when applying the above knowledge based on 2D foil study to

the cavitation erosion problems in turbines and pumps, precautions must be
taken in respect of the following factors.
• The runner cascade is a fully 3D blade system. The uneven spanwise

distribution of incidence along the blade inlet edge makes a significant
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`
4
As to the focusing collapse of the bubble cloud, please refer to the cluster collapse
model of multi-bubbles by Hansson & Moch in §3.7.3 "Bubble-bubble Interaction".

282 Cavitation
contribution to the 3D effect of flow, and particularly accentuates the

streamwise rolling-up, forming strong streamwise vortices.
• The secondary flow formed in the boundary layers around the inter
sections of the blades with the crown and the band (Francis type) or
with the hub (axial-flow type) often results in the formation of stream-
wise vortex flows around the intersections 5 , as shown by Fig. 6.8.
Figure 6.8: Secondary vortices at the intersection of a blade-cascade with

the bottom plate (Kuethe & Chow, 1986 [6.27])
These vortices will cavitate at much higher cavitation numbers, then

collapse somewhere downstream, depending on the machine operating
conditions, and in the collapse process produce highly erosive power.
• The inlet flow towards the runner blades is not uniform, and in the
case of high-specific-speed machines the spanwise variation is very sig
nificant thereby emphasising the 3D nature of the flow between blades.
Furthermore, the wakes of the guide vanes cause temporal variations
of the relative inlet flow to the runner blades, which modulate the
leading-edge cavity oscillations and the vortex shedding at a guide-
vane passing frequency, fgv. Therefore, the Strouhal number, St, in
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`--
5
The mechanism of such secondary flow formation is well documented, e.g. Kuethe &
Chow, 1986 [6.27], pp 433-436 and Hamn-Ching Chen, 1995 [6.6].

Pressure side
Travelling bubble cavitation
X/Lch
0.0
\Leading edge cavitation ' ° (relative chord length)
(inlet) > - _ (outlet)
Figure 6.9: Conceptual sketch of pressure distributions on turbine and pump

blades
eqn.(6.1) should be replaced by a modulated Strouhal number based

on the modulated frequency fgv.
• The mean pressure along the flow passage of a machine runner de
creases in turbines and increases in pumps, referring to Fig. 6.9; while,
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
for the 2D foil flow, it is basically the same at the inlet and the outlet.
Consequently, the erosive power of leading-edge cavitation in pumps
is strengthened owing to the higher pressure gradient experienced by
the collapsing cavitated vortices; while it is weakened in turbines. For
the same reason, travelling-bubble cavitation associated with damage,
which usually appears at the mid-chord or near the trailing edge, is
often observed in turbines but seldom in pumps.
The cavitation-damage intensity in machines is generally proportional

to the 6th power of velocity. A typical specimen of cavitation damage on a
high-head Francis pump-turbine is shown in Fig. 6.10. This deep sponge-like
damage on the pressure side of stay vanes after only 4000 hours of service
from commissioning is caused by the leading-edge cavitation produced by
negative incidence in the pumping mode (Li, 1987 [6.28]). This case demon
strates that the leading-edge cavitation can occur in a very high mean pres
sure area (approximate 400 m pumping head) and produce extremely high
erosive power.

284 Cavitation
Figure 6.10: Leading-edge cavitation-damage (pump mode) on the pressure

side of the stay vane of a Francis pump-turbine (Li, 1987 [6.28])
N = 300 MW, if = 494 - 541 m; Operating duration: 4000 hrs.
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

6.2.2 Propeller and Kaplan Turbines

The cavitation damage zones often observed on the blades are shown in
Fig. 6.11. The leading-edge cavitation damage zone Is on the suction side
Pressure side Suction side
Figure 6.11: Cavitation damage zones on the blade of Propeller and Kaplan
turbines
is usually due to the operating head being higher than the design head
(consequently, resulting in a larger inlet incidence); while its counterpart
zone Ip on the pressure side is due to the operating head being lower than the
design head (resulting in a negative incidence). For Kaplan turbines, these
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
two zones are much less prone to damage because of "on cam" operation.
The damage in zone IIIS along the mid-chord length on the periphery of
suction side, and on zone IV, i.e. the tip end along the mid-chord region,
are caused by tip vortex cavitation. This damage is often severe due to the
strong erosive power of such vortices, particularly, on high-head machines.
Travelling cavitation is often responsible for the damaged zone IIS observed
in the area from mid-chord to the tail, which occurs at high flow rates (i.e.
the cavitation number a of machine is in its high-value region) and/or at high
machine setting level (i.e. the cavitation number ap of plant is low). This
travelling cavitation reduces performance (efficiency) remarkably although
its erosive power is usually low. The hub cavitation developed in Kaplan
turbines may cause damage on both the spherical part of hub and the suction
side of the blades near the blade axis. Such damage is usually not severe
but it is typical in Kaplan turbines.
Damage zones on non-rotating parts are shown in Fig. 6.12. Tip-vortex

286 Cavitation
vvn/
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
Figure 6.12: Cavitation damage zones on the non-rotating parts of Propeller
and Kaplan turbines
cavitation also damages the discharge ring, producing an erosion band (zone
V). The discontinuity between the discharge ring and the draft-tube liner
often results in a cavitation erosion band (zone VI). The cavitated secondary
vortices shed from the bottom of guide vanes may also cause damage on the
bottom and/or discharge rings (zone VII) as well as on the runner blades.
The damage in zone VII often appears immediately downstream of the wake
of the guide vanes intermittently in peripheric direction 6 . Damage caused
by guide-vane-leakage cavitation (in closed position) often appears on the
vertical end of the guide vanes (its counterpart appears on the top plate of
the bottom ring) indicated as zone VIII, and its horizontal seal areas (zone
IX). High-head machines are more susceptible to this cavitation damage.
6.2.3 Francis T u r b i n e
The damage to a Francis runner is shown in Fig. 6.13. In a similar fashion as

before, the leading-edge cavitation damage zone Is on the suction side is due
t o t h e o p e r a t i n g head being higher t h a n the designed value. For Francis tur-
See the Example 2 in §6.6.4 "Examples" where the damage area on the bottom ring
in the wake of guide vanes of a propeller turbine is such a case

Suction side of blade
Pressure side of blade,
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
Figure 6.13: Cavitation damage zones on the runner of a Francis turbine
bines (particularly the high-specific-speed ones), the meridian velocity field

is strongly non-uniform with the highest meridian velocity around the corner
of the band which thus accentuates the negative pressure on the blade suc
tion side behind the leading edge. Therefore, a secondary streamwise vortex
flow structure forms there in which the vortex cores are very susceptible to
cavitation. The collapse of these cavitating vortices produces very aggres
sive erosion in zone Is on both the blades and the band, which is a typical
damage pattern for Francis turbines even operating under design conditions.
When operating at a head lower than the design head, the leading-edge cav
itation damage occurs on the pressure side as shown by zone Ip. These two
damage zones are related to the inlet incidence in respect of its deviation
from the design value rather than to the value of the plant cavitation num
ber Op. However, low-specific-speed Francis runners show little damage in
these two leading-edge damage zones because they are not susceptible to
the variation of operating head. The cavitation damage zones IIC caused
by low-load operation are complex owing to the uncertainty of the strong
secondary vortex developed at the inlet of the blades, which is mainly caused
by the incidence variation from the hub to the band. This cavitating vortex
may appear in various forms and positions somewhere in the space between
two adjacent blades; sometimes, two parallel vortices may appear depend-

288 Cavitation
ing on the design and operation conditions. Hence, the observed damage
patterns vary in location. If the vortex attaches to the intersection of the
blade-inlet edge with the crown and extends down through the space be
tween the two adjacent blades into the draft tube, the attached area at the
intersection may be damaged by the collapse of the vortex. A typical sketch
of such secondary vortex flow simulated with a potential flow model by Gode
& Cu^nod, 1988 [6.12], is shown in Fig. 6.14. The second position that
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
Figure 6.14: Part-load secondary-vortex at the inlet between the blades with
the upper end attached to the intersection of the blade leading-edge with the
crown (Gode & Cuenod, 1989 [6.12])
the vortex may attach to is the midway of the crown between the blades.
The attachment may also move further downstream towards the outlet of
the blades. Thus, the damage zones on the crown could appear at various
places. If the vortex moves further downstream to the centre of the run-

n
Part-load damage Low-head damage
zone He zpnelp_
Full/over-load damage
zone IIs
11
Figure 6.15: Conceptual sketch of the operation-range influence on cavitation

damage.
ner then a cavitating vortex-core flow forms in the draft tube and produces
huge pressure pulsations, but with insignificant damage to the draft-tube
liner. The trailing-edge damage zone IIS on the suction side near the band
is often observed; it appears from mid-chord towards the trailing edge and
is caused by travelling-bubble cavitation. This damage is often caused by
operation at full flow rate and/or at a lower value of the plant cavitation
number ap. Besides, the leakage through the seals on the band may cause
leakage cavitation (vortex type) damage on the band, and on the discharge
ring and the draft-tube liner. Figure 6.15 shows the influence of operation
range on cavitation damage (where unit speed: n n = ^ £ ; unit discharge:
Qii DJVH ) ■
Cavitation damage on non-rotating components is often observed on the

guide vanes at corresponding locations to those on propeller and Kaplan
turbines (referring to Fig. 6.12), which is caused by leakage when the guide
vanes are closed. High-head machines are more susceptible to leakage cavi
tation damage.
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

290 Cavitation
6.2.4 Pelton Turbine

Two cavitation coefficients need to be identified for Pelton turbines. One is
related to the lowest pressure developed on the surface of the needle, which
can be evaluated either by calculation or by experiment; the other is related
to the lowest pressure developed on the inner surface of the buckets, which is
difficult to evaluate. The understanding of cavitation and associated damage
on this type of machine, particularly for the buckets, is limited and mainly
obtained from observations on prototypes.
High head
P (kg/cm 2 )
Low head
Figure 6.16: Pressure distribution and cavitation damage
At present, the IEC code stipulates no damage for Pelton turbines. Nev
ertheless, damage occurs at various locations, particularly for high-head and
high-specific-speed (with a high number of jets) machines. The often ob
served cavitation damage on the surface of the needle is shown in Fig. 6.16.
The pressure in the low-pressure zone may reduce to the vapour pressure
under conditions of partial load and high head. Proper profile design can
improve cavitation performance, reducing the risk of this damage. The dam
age on buckets (Grein, 1990 [6.13]) is shown in Fig. 6.17. These damage
zones are caused by cavitation erosion and/or rain erosion 7 . Damage zone
Rain erosion is caused by repeated impacts of travelling liquid droplets which may come
--``,```,,``````,,,,``,`,,

Figure 6.17: Damage on a Pelton bucket (Grein, 1990 [6.13])

(1) Back of splitter tip (2) Back of cut-out (3) Front side of splitter tip (4)
Upper zone of splitter (5) Bottom of bucket (6) Upper zone of inside bucket
wall (7) Upper zone of outside bucket wall
/ at the back of the splitter is caused by the repeated collapse of a small

cavity developed at the back every time the tip cuts into the jet. The rain
erosion resulting from the uneven distribution of energy within the jet might
be also responsible for this damage. Zone II at the back of the cut-out is
often observed for high-specific-speed and multi-jet machines, where the out-
layer of jet is often a layer of droplets resulting from the disturbance caused
by the out flow ejected from previous jet and produces this rain erosion at
zone II. Zone V at the bottom of the buckets is often detected. This is
probably because the jet is often disturbed by the back of the next bucket
entering the jet, forming droplets and/or cavitated high-speed vortex flow
in the disturbed jet and causing subsequent rain erosion and/or cavitation
erosion. Zone VI is observed on the bucket with a large angle of outflow,
which leads to lower pressure and higher cavitation-damage risk. The re
duction of this angle can reduce the risk of such cavitation damage and also
improve machine efficiency, but it may cause the tangent impact of droplets
to zone VII of the subsequently entering bucket, resulting in rain erosion
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
damage. At zone III, i.e. the front edge of the splitter tip, and zone IV, i.e.
from jet interaction and splashing. The rain-eroded surface has very similar appearance
to t h a t of cavitation-damaged surface.

292 Cavitation
the upper surface of splitter, cavitation damage may be observed for buckets
with a poor profile design and surface finish.
Cavitation damage as a secondary (subsequent) effect of silt erosion of
ten occurs in Pelton turbines. The Pelton turbine is highly susceptible to
silt erosion especially with high-head applications. The silt-eroded surfaces
on the needle, nozzle and buckets will worsen flow conditions and promote
secondary cavitation damage.
6.2.5 Cavitation Guarantee

The cavitation guarantee is a contractual agreement between the machine
manufacturer and the owner, which provides a permissible loss of metal
from cavitation damage for a selected duration of operating hours within a
guaranteed number of years. The amount of cavitation damage depends on
the following factors 8 :
• The type and design of the machine;
• The material and surface condition of the parts subject to cavitation;
• The setting of the machine in the plant, i.e. the ap value of the plant;
• The duration of operation and operating conditions.
The guarantee is usually expressed in terms of allowable metal-loss rate

W
Where W: allowable metal loss rate (Ib/hr or kg/hr);

K\: cavitation intensity coefficient (lb/ft2 or kg/m2);
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
D: runner discharge diameter (ft or m);

Ho'- cavitation guarantee duration of operation (hrs).
Several cavitation guarantee standards, provided by the International

Electrotechnical Commission (IEC), the U. S. Corps of Engineers (COE),
the U. S. Tennessee Valley Authority (TVA) and the U. S. Bureau of Recla
mation (USBR) etc., are often referenced. The IEC code gives the value of
# 1 = 0 . 0 9 7 - 0 . 3 9 lb/ft2 (0.48 - 1 . 9 kg/m2) for the runner only. The machine
8
IEC Publication 609: Cavitation Pitting Evaluation in Hydraulic turbines, Storage
Pumps, and Pump-Turbines.

Allowable metal loss on runner

per 8000 hours (lb)
Period installed : ■ /
• 1950-1959 : /
0 1960-1969 • ! •!* /
+ 1970-1984 : t O0 /
:0» 7 :
i• i /•
1 _,_ • ..
1EAJ- upper limit • •
( 1978) KJj=0.39
J rt ' ^ / • +': j
:•
:O
:+• ?V; •£-; /
:0 / _m /
i ! $<><*% \ l l . . .
■• \ • / + i /:
:ziz*:*2JfTtiz
\+
•• •
I I...Z i
•:
ifjfKYI 1 IEC lower limit

(1978) K 1=0.097
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
o
10 15 20 30 40 50
Runner discharge diameter (ft)
Figure 6.18: IEC cavitation damage guarantee compared with 729 units
installed in the USA since 1950 (Sinclair & Rodrigue, 1986 [6.47])

294 Cavitation
units of total 729 installed in USA since 1950 were examined against the
IEC code by Sinclair & Rodrigue, 1986 [6.47], as shown in Fig. 6.18. They
concluded that the use of the lower limit of IEC code is permissible. For
large runners (runner diameter above 7.3 m) and for generating units where
repair outages will result in generation and capacity value losses, a reduc
tion of the K\ value to 75% of the IEC lower limit is recommended. For
propeller and Kaplan turbines, the damage on the non-rotating parts should
not exceed 50% of runner's limit value; while for Francis turbine and pump-
turbine, it should not exceed 33%. The guarantee duration of operation,
Ho, is 8000 hrs and for peak operating units and pump-turbines reduces to
3000 hrs. The recommended guarantee period is two years.
For Pelton turbines, a guarantee code is not yet available although an
IEC draft is in preparation.
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

6.3 Cavitation Damage in Pumps

S.C. Li
6.3.1 Introduction
Cavitation damage in pumps occurs mostly at the inlet area of the blades,
and is more severe than in turbines because the following two factors. Firstly,
for pumps (particular at off-design conditions and with small commercial
pumps), the flow in the inlet area is a highly unstable vortex flow, which
is very susceptible to cavitation 9 and provides most erosive power 10 . Sec
ondly, as mentioned before, for pumps the increasing mean pressure along
the flow passage accentuates the pressure gradient of the pressure recovery,
which provides a much stronger driving force for the collapse of cavitation
bubbles/clusters. The resultant damage rate is usually several times higher
than on turbines. In particular, the damage on the pressure faces of blades
caused by such cavitation is in some cases 50 times higher than that to the
suction faces, because of the even higher driving force on the pressure side
(Giilich fc R6sch,1989 [6.15]).
6.3.2 Axial Flow Pumps

The causes and locations of cavitation damage in axial flow pumps are more
or less similar to those in axial flow (propeller and Kaplan) turbines, as
shown in Fig. 6.19. Damage zone J3 at the closure of the leading-edge
cavity on the suction side is most often observed, particularly for fixed-blade
pumps, when the flow rate and NPSH value are low. Damage zone IIS on
the suction side as well as damage zone III on the tip end of the blade and
its opposite surface on the discharge ring are caused by tip-vortex cavitation.
For adjustable-blade units, there might be hub cavitation damage due to the
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
gap leakage between the blade and the hub. Damage zone Ip at the closure
9
I t is known from §3.7.4 "Bubble-Flow Field Interaction" (Single/Individual Bubble
Study) t h a t the ratio, fi, of the pressure drop caused by the vortex (Rankine model) to
the ambient pressure is il = -p£—(^rfs-)2, ^ dramatic pressure drop at the core centre
can be expected as ac is at its smallest value. Furthermore, owing to the capture effect,
the number density of nuclei in the core is much higher and the nuclei can stay for a much
longer time period in the core centre to cavitate, thus promoting the cavitation occurrence
in the core even under fairly high ambient pressure.
'"Unstable vortex cavitation provides more severe damage, at least an order of magni
tude higher than its stable counterpart; this is often the case for pumps operating at low
flow rate and/or low NPSH value.

296 Cavitation
Pressure side Suction side
Figure 6.19: Cavitation damage zones on axial flow pumps
of the leading-edge cavity on the pressure side may also occur, particularly
for fixed-blade units, when operating at full flow rate or above.
6.3.3 Centrifugal P u m p s
The damage locations in centrifugal pumps are even more uncertain than in
Francis turbines, since the pattern of vortex flows and the location where
the cavitated vortices implode are not easily predictable. They vary dra-
Cavitation inception
NPSH
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
Permissible cavitation damage
-•-+-..: (e.g. 5 years)
3% flow rate drop

(developed cavitation)
Q/Qo
0.5 1.0 1.5
Figure 6.20: Variations of cavitation inception, permissible damage and 3%

(flow rate drop) NPSH line with the relative flow rate
matically with operating conditions as well as with the impeller design.

(a)
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
Figure 6.21: Cavitation damage on the suction side of an impeller

(a) Photograph of the damage ( Palgrave & Cooper, 1986 [6.35]) (b) Damage
zones

298 Cavitation
Cavitation damage can be basically categorised as follows. At, or around,

the optimum flow rate, -g- « 1 (referring to Fig. 6.20), only thin and smooth
sheet cavity might present on the pressure- and/or suction faces of blade
leading edges, producing little damage potential. With the deviation of flow
rate from $- = 1, damage occurs at the closure of the leading-edge cavity
(on the suction side for $- < 1 and on the pressure side for $- > 1). Figure
6.21 shows such damage on the suction side on a centrifugal pump. Zone Is
is located at the closure of the leading-edge cavity. Zones II, and IIIS are
damaged by the cavitated secondary vortices formed at the fillets of blade-
hub and blade-band. With a further reduction of flow rate, flow recirculation
may develop at the pump inlet as well as at the impeller discharge, generat
ing vortices whose pattern and location are not easily predictable 11 . These
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
vortices may attach to the pressure face of the blade, producing extremely
damaging power there. Besides, the shed vortices from the leading-edge cav
ity on the suction side can also provide erosive power on the pressure face of
the adjacent blade if these shed vortices collapse near/on it. The unstable
vortices resulting from such recirculations are persistently present when the
flow rate is particularly low, i.e. in the region from B to A (referring to
Fig. 6.20) where the NPSH line descends with the reduction of flow rate,
resulting in a huge flow-rate surge and damage if the NSPH level is low
enough to cavitate vortex cores. For details, see, e.g. Palgrave & Cooper,
1986 [6.35].
6.3.4 Pump-turbines
Cavitation damage to pump-turbines is similar to their counterpart turbines
when operating in turbine mode; and, similar to their counterpart pumps
when operating in pump mode.
However, the setting level of pump-turbines is determined by the pump
ing mode which gives a larger value of plant crp than that required by the
generating mode. Therefore, cavitation damage of pump-turbines operating
in turbine mode, particularly over the trailing part of the blade suction side
(travelling-cavitation damage), is much less compared with the counterpart
n
F o r high-energy pumps, with large inlet-eye diameter (i.e. high inlet peripheric ve
locity) and small blade-blade overlap, inlet recirculation is more likely to occur, which is
characterised by the highly unstable shear vortices generated between the rejected back-
flow at the outer eye diameter and the incoming flow near the hub. Severe damage will
be expected if these vortices attach to the surface of the flow passage. Similar damage at
the impeller discharge and/or in the diffuser can be caused by discharge recirculation.

turbines.
For both operating modes, severe leading-edge cavitation damage may
be observed when the entering flow deviates. Damage on the suction side of
the blade caused by travelling cavitation may also be detected. Tip-vortex
cavitation damage occurs for propeller and Kaplan pump-turbines and hub
cavitation damage occurs in Kaplan pump-turbines.
When operating in pump mode, the flow recirculations at the runner
entrance and at runner discharge can cause severe damage. In this mode,
leading-edge cavitation damage may also appear on the guide vane, even
on the stay vane as shown by Fig. 6.10, either due to the improper de
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
sign of the vane entrance or due to the periodical variation (at the blade
passing frequency) of the incidence relative to the guide vane caused by the
peripherally nonuniform discharge flow of the runner. This damage is often
severe owing to the unstable nature of the leading-edge cavity and the high
pressure/pressure gradient there.
6.3.5 Cavitation Guarantee

For storage pumps and pump-turbines, the IEC code, publication 609 as
depicted in §6.2.5 "Cavitation Guarantee", is applicable.
For other pumps, the cavitation guarantee is usually given in terms of op
erating life during which the pump can provide proper service. One criterion
assumes that 100,000 hours or 10 years (80,000 hours) of operating life can
be expected if the threshold value of cavitation damage 12 is not exceeded.
The other criterion often quoted is 40,000 hours or 5 years of operating life,
which allows limited cavitation damage as shown in Fig. 6.20. The NPSH
line for permissible cavitation damage (5 years) is somewhere between the
inception line and the 3% line.
For particular applications, the operating life guaranteed may vary from
case to case. For example, the expected life of a high-pressure oxidiser
turbopump for a space-shuttle main engine is only 7.5 hours, allowing ap
preciable cavitation in such adverse operating conditions (i.e. the extremely
high suction specific speed, head rise and flow rate).
Based on accumulated data, the operating life of pumps, with known
geometry/material/operating conditions, can be estimated by various ap-
12
The threshold value is determined by the cavitation intensity and the material resis
tance, i.e. the threshold value is considered to be reached or exceeded if the implosion
pressure produced by cavitation is equal to or greater than the fatigue strength of the
materials.

300 Cavitation
proaches. For example, by relating the leading-edge cavity or its noise level
to the erosion rate, Gulich & R6sch,1989 [6.15], proposed an approach for
radial and semi-axial centrifugal pumps to calculate the probability, W, at
which the pump can reach the required operating life, lL,Req- The following
equation is used to correlate the properties of the cavity with the erosion
rate ER at the location possessing the maximum material loss,
12
P2)3Fa
ER = CL
J
cav,R
(P0 -
^Fr mat (?n=)(*r<-*> (6.2)
where a constant, being 2.3 x 10~ 12 for the suction side and
1.1 x 1 0 - 1 0 for the pressure side (m3/Ns);
x2: being 2.83 for the suction side and 2.6 for the pressure
side;
Lcav,R reference cavity length, being 10 (mm);
J-'cav measured cavity length (mm);
PQ static pressure at impeller inlet (N/m2);
Pv vapour pressure (N/m2);
Fcor factor accounting for corrosion effect;
Rm tensile strength (N/m2);
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
Fmat factor accounting for material properties;

aR reference gas content, being 24 (ppm);
a gas content (ppm);
a« reference sound speed in water, 1490 (m/s);
a sound speed in the water used (m/s);
P'k reference saturated vapour density of water, being
0.0173 (kg/m3);
saturated vapour density of the water used at the given
temperature (kg/m3);
Erosion rate ER for each operating condition is evaluated and multiplied

by relative duration r . The calculated operating life is
0.75EB
h,Cal = (*)
ZTER
Here, EB is the blade thickness (m). Thus, the probability, W, that the
pump can last for the required operating life, It,Req, can be obtained from

Factor RL
2.0
1.5
1.0
0.5
0
20 40 60 80 100
Probability W (%)
Figure 6.22: Probability functions of the operating life (Gulich & Rosch,
1989 [6.15])
Curve 1: cavity property method; Curve 2: cavitation noise level method
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
the curve 1 of Fig. 6.22 by evaluating the life factor RL,
h,Cal
Similarly, the probability of reaching the required operating life can also
be estimated by evaluating the cavitation noise level (CNL) (for details see
Gulich & Rosch, 1989 [6.15]). By evaluating the factors influencing the dam
age, the operating life can also be estimated. For example, Doolin, 1986
[6.9], proposed an approach to estimate the operating life by the production
of seven factors. They are: (1) Ratio of available NPSHa to rated (3% head
drop) NPSH; (2) Thermodynamic properties of liquid; (3) Corrosive effect
of liquid; (4) Cavitation resistance of material; (5) Speed of pump; (6) Off-
peak operation; (7) Suction specific speed. Based on the life factor FL, i.e.
the product of these seven factors, the operating life, II, can be estimated
as
IL = 10FL {yrs)
The evaluation of these seven factors are given in [6.9].

302 Cavitation
6.4 Silt-Laden Water Effect

S.C. Li
6.4.1 Introduction
The abrasive wear of hydraulic machinery caused by silt-laden water has
become a world-wide problem. It is a multidisciplinary subject which is as
yet not well understood, particularly when it occurs together with cavitation
erosion. The mutual influence mechanism between silt erosion and cavitation
erosion, i.e. the synergistic action which is far more devastating to machine
parts, needs to be fully understood so that comprehensive guidelines can
be provided to cover scheme planning, machine selection and layout, the
hydraulic design of the machine, material selection/coating application, and
machine operation/maintenance/repair etc. The current trend to the use
of higher specific-speed machines makes machine parts more susceptible to
such synergistic damage 13 . Therefore, it is vital for machines used in the
environment of high silt concentration, that the penalties arising from sand
erosion and cavitation damage are understood and evaluated.
6.4.2 Silt Erosion

The morphology of silt-eroded surfaces is characterised by micro-cutting
grooves in the flow direction. Figures 6.23 and 6.24 show the eroded guide
vane (Francis turbine) and nozzle needle (Pelton turbine) respectively, on
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
which the typical flow-direction grooves are clearly seen.

Up-to-date knowledge shows that the erosion process is governed by the
following three factors.
Silt Properties
Large hard particles with sharp edges provide the most erosive poten
tial. The content of entrained particles usually comprises the very hard and
sharp quartz particles, siltstone (rock) particles and very fine mud parti
cles. Among them, are quartz particles ranging from a few /j,m to hundreds
of fxm which arrive seasonally in rivers fed by melting snows in mountains
13
The rate of silt erosion alone increases with 3rd power of relative flow velocity and the
rate of cavitation erosion alone increases with 6th power of relative flow velocity; whereas,
the rate of synergistic erosion increases at a much higher rate than the summation of the
above two, owing to the synergistic action.

Figure 6.23: Eroded guide vane (Grein & Schachenmann, 1992 [6.14])
Francis turbine: N = 4.4 MW, H — 154 m; Material: 14/1 stainless steel;
Operating hours: 41,000 hrs; Silt concentration: up to 1.5 kg/m3
Figure 6.24: Eroded needle (Grein & Schachenmann, 1992 [6.14])

Pelton turbine: 2 nozzles, N = 8.4 MW, H = 540 m
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

304 Cavitation
Screening (%) Not retained (%)
100 ' ' 1 ^ ' ——^ ' 1 1 0

5 10 20 50 100 200 500
Grain size ( |i m)
Figure 6.25: Screening characteristics of solid particles in the intake water

of some power stations (Grein & Schachenmann, 1992 [6.14])
1: Mean characteristics at Grand Dixence pumped storage station (Switzer
land) 2: Stefel (Switzerland) 3: Nathpa Jhakri (India) 4: Kabul river (Pak
istan) 5: Min Jiang river (China) 6: Shuikou power station (Min Jiang river,
China) 7: Zermatt (Switzerland)
such as the Himalayas and the Alps. These particles are mostly erosive
although their concentration is usually not high (only a few kg/mz). The
remaining particles are mostly siltstone, which are usually milled down by
each other to small/fine sizes (a few /*m up to 100 /j.m) during the trans
portation process. However, in seasonal floods their concentration may rise
to some hundreds of kg/m3. These particles are also erosive and their ero
sion rate is reported to be approximately proportional to their concentration
(Grein & Schachenmann, 1992 [6.14]). However, their erosive power may be
retarded by accompanying extra fine mud particles in which they are en
trained during flood seasons. Therefore, the composition of the entrained
particles, their screening characteristics, and their shapes and hardnesses are
necessary information for the planning/designing of anti-erosion measures.
Typical screening size characteristics of entrained particles for some rivers
are summarised as shown in Fig. 6.25 (Grein & Schachenmann, 1992 [6.14]),
which shows the Himalayan rivers (i.e. the curve No.5 for Min Jiang river,
China) having coarser particles than the others.
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

Material Properties
In general, for metals, material hardness is a key parameter in evaluating

the erosion resistance, i.e. resistance increases with hardness. If the hard
ness exceeds that of erosive particles, little erosion is expected. However,
some soft but elastic materials such as rubber etc. do show certain erosion
resistance, presumably, owing to their ability to absorb impacts from par
ticle impingement, where material toughness plays a dominant role rather
than its hardness.
Driving Forces
The forces driving the particle impinging and cutting the material surface
are complex.
Firstly, in many cases the major driving force results from the relative
velocity of the flow to the particle. It exerts on the particle in the relative
flow direction. The erosive power of the particle driven by such a force
increases with the flow velocity. Hence, for locations where the flow velocity
is high, severe erosion can be expected. Furthermore in turbulent flows, the
particle does not only obtain increased momentum but it can also regain
this momentum repeatedly after making microcutting (i.e. losing its cutting
potential) and being carried into the outer layer of the boundary flow where
it is re-energised.
Secondly, the inertia forces play a dominant role in some flows. One
such case arises when the centrifugal force perpendicular to a concave sur
face presses the particle against the surface while it is being dragged over
and in so doing cuts the surface. Therefore, for applications where large par
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
ticles and high flow velocities are encountered, precautions must be taken to
avoid sudden changes in flow direction, and/or corresponding anti-erosion
measures should be employed. The particles entrained in large eddies or
vortex cores with their axes parallel to the surface is another such a case
where the centrifugal force perpendicular to the surface causes the particle
to impinge on and cut the surface. The herringbone-like scouring on the
bottom ring surface in front of the guide vane (Francis turbine) shown in
Fig. 6.26 is a typical example of such microcutting by the particles in the
secondary vortex formed there. See Figs. 6.8 and 6.27 comparing with Fig
ure 6.26 for the evidence of the agreement between the vortex flow and the
resulting scour pattern.
Thirdly, gravity and buoyancy should be taken into account in some cases
although they play less significant roles.

306 Cavitation
Figure 6.26: Herringbone-like scouring on the guard-ring surface in the front

of the guide vane (Grein & Schachemann, 1992 [6.14])
Francis turbine: N = 43 MW, H = 420 m; Operating hours:12,000 hrs
Various silt-erosion models are available but none of them can simulate
all the above factors. The model by Sato et al, 1990 [6.42], is introduced here,
which estimates the particle relative penetration depth Dp(= h/a), referring
to Figure 6.28, based on the assumption that the silt particle will penetrate
into such a maximum depth, Dp<Tnax, that its kinetic energy reduces to zero,
DP,max = O.64p*(sin0)lV*H~*
where p: particle density, (kg/m3);

9: incident angle, (degree);
V: impact velocity of particle, (m/s);
H: material hardness, valued as Vickers hardness x 10 7 ,
(kg — m/s2 /m).
Sato et al, 1992 [6.43], further classify silt erosion into three modes,
namely ploughing, wedging and cutting, subject to the maximum pene
tration DptTnax and the particle-surface shear strength / as shown by the
abrasion mode diagram, Fig. 6.29. For hydraulic machinery, ploughing and
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

Figure 6.27: Secondary vortex-core formed at the intersection between col

umn and bottom plate (Grein & Schachemann, 1992 [6.14])
wedging modes are commonly observed.
6.4.3 Synergism of Silt and Cavitation Erosions

In silt-laden water, the silt abrasion and cavitation erosion will enhance
each other through synergistic action. The enhancement takes place in the
following ways.
• Promotion of Cavitation
The particle entrainment will increase the nuclei concentration of water
and, consequently, promote the cavitation occurrence. A 10 — 15%
increase of the incipient cavitation number, CTJ„, in silt-laden water
compared with tap water was reported by Toshiba et al, 1991 [6.50].
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`--
• Strengthening of Impingement

308 Cavitation
FsinG
Figure 6.28: Silt penetration model (Sato et al, 1992 [6.43])
The particles will gain much higher velocities when coupled with cav
itation. In particular, the microjets from the collapsing bubbles give
particles a very high velocity towards the surface. Therefore, particle
impingement with incidence 6 ss 90° and high velocity is much stronger
than that in non-cavitation conditions.
As an example, the study of damage to stainless steel 18Cr8Ni on
a hydrofoil by Jin et al, 1986 [6.23], is presented. The silt erosion
is dominated by the ploughing mode which produces smooth eroded
grooves with a maximum width of 1 mm and a maximum depth of
0.15 mm; whereas, under cavitating conditions, the grooves become
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
deeper and longer with a maximum length of 7 mm and a maximum

depth of 5 mm. Owing to the deeper penetration, numerous particles
are embedded in the bottom surface of the pits damaged by cavitation
microjets, as shown in Fig. 6.30(a). Figure 6.30(b) shows the cutting
on the bottom surface of cavitation pits. The wedging mode is also
observed as shown in Fig. 6.30(c).
Alteration of Flow Pattern

The occurrence of cavitation is a remarkable alteration to the original
non-cavitating flow pattern, often resulting in an altered flow pattern
in which more violent flow structures are involved such as the unstable

a
>
o
2
u
T3
O
/\/N
^
^^^^H^"'' if
T
S
S3
Ht||j|^
<L>
Ploughing Wedge Cutting

Degree of penetration Do
0.6
—
0.5 — I \k Dp=h/a
0.4 —
\affi SL
0.3 —
Cutting
0.2 —
— ^^^^^^""^-^
Wedge
Submerged jet
\ (SUS 304)
0.1 -
Ploughing Water channel
(Aluminium)
,
1 1 \
0.2 0.4 0.6 0.8 1.0
Shearing strength at the contact interface f
Figure 6.29: Abrasion mode diagram (Sato et al, 1992 [6.43])
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

310 Cavitation
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
Figure 6.30: Damage by synergism of silt- and cavitation erosions (Jin et al

1986 [6.23])
(a) Embedded particles on the surface of pit-bottom (Magnification 500 x)
(b) Cutting scars on the surface of pit-bottom (Magnification 750x) (c) A
wedged particle on the flank surface of pit (Magnification 1750x)

leading-edge cavity with strong vortices at the closure and coupled

with vortex shedding. The particles entrained in these flow structures
possess strong damage potential. The study on the silt erosion with
and without cavitation in a venturi flow device by Sato et al, 1992
[6.43], demonstrates such an effect of flow alterations involved in a
2D leading-edge cavity as shown in Fig. 6.31. Without cavitation, the
portion A of the bottom specimen is severely eroded, owing to the high
flow velocity and the strong shear vortices there, whereas the portion
C is less eroded. When cavitation occurs, the violent closure of the
cavity and the collapse of the shedded vortices accelerates both the
silt- and cavitation erosions on portion C, whereas erosion on portion
A is reduced since it is covered by the relatively less violent cavity
sheet which retards the silt impinging, referring to Fig. 6.31(a). The
erosion (where, say, assessed by the total weight loss) on the bottom
specimen, if evaluated as a whole, seems not to be affected. However,
the overall erosion on the side specimen is accentuated by synergistic
action. This study evidently shows the close relation of synergistic
action with the flow alterations. It also explains why some studies
performed on different devices where detailed observations on the flow
pattern alterations are ignored lead to controversial conclusions about
the synergism.
On the other hand, the alterations of pressure distribution caused by
the presence of solid particles will enlarge the extent of the cavitation,
thus increasing the cavitation damage power.
• Subsequent Damage
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
The damage on the surface caused by either silt erosion or cavitation

erosion, or by their synergistic action, will worsen the profiles of ma
chine parts, which may induce violent changes to the flow pattern such
as cavitating vortices in the immediate downstream, resulting in more
damage. Such subsequent damage is often observed in leading-edge
erosion where the worsening of the leading-edge profile will further de
teriorate the flow pattern and cause yet more damage, resulting in an
accelerated erosion process.
The mechanism of material fracture by synergistic action varies from

material to material. Based on observations of the microstructure, Zhao
et al, 1993 [6.51], postulate a model for the metals often used for hydraulic
machinery, namely, hard alloys, stainless steels and carbon steels. The weight

312 Cavitation
Cavitation cloud Flow

Bottom-specimen
(abrasion)
Erosion depth
( um/h) Test material :Aluminium
Velocity: 40 m/s
Silt concentration: 30 g/1
(abrasion+cavitation) Silt granule size: 39 u m
(a)
Erosion rate (cm 3/h) (X10"3 ) Erosion rate (cm 3/h) (X 10"3 )
Bottom specimen Side specimen
' wc
=1.0
CTWC=1.5 80
(abrasion) Abrasion
/ .A Cavitation
6 —
//
:
4 — il .=1.5
/A.
// awc=1.0
A
2 — / Abrasion
(,-, •+ •
Cavitation
i
|
1 1 1
10 20 30 40 50 10 20 30 40 50
Silt granule size (lim) Silt granule size (\i m)
(b) (c)
Figure 6.31: Influence of flow-pattern variation on erosion rate. Test mate

rial: aluminium; Velocity: 40 m/s; Silt concentration: 30 g/l. (Sato et al,
1992 [6.43])
(a) Depth of erosion on the bottom-specimen (b) Erosion rates on the
bottom-specimen (c) Erosion rates on the side-specimen
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

loss rate, Wc+S, of synergistic erosion is the weighted sum of the individual
rates of silt erosion, Ws, and cavitation erosion, Wc, which relates to the
mechanical properties as follows,
Wc+S = AWC + BWS = « 7 ^ i + Pjj
where A, B: weighting coefficients for cavitation erosion and silt

erosion respectively;
a, (3: constants relating to weighting coefficients A and B
respectively;
tf. material true fracture strain in static tensile test,
used to indicate its fatigue-ductility coefficient e',
since e'f = (0.35 ~ 1.0)e/;
H: material hardness (HRC).
This model assumes that the process of cavitation erosion is a low-cycle

strain fatigue whose erosion rate is inversely proportional to the fatigue life,
i.e. (e/H)2 according to Manson-Coffin theory of strain fatigue; and the
process of silt erosion is a microcutting whose rate is inversely proportional
to the hardness H. The weighting coefficients A and B, or constants a and
/?, are subject to the following factors.
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
• The silt characteristics such as silt concentration and composition, par
ticle hardness/shape/size.
• The flow characteristics such as flow and cavitation patterns, flow ve
locity, turbulence level and cavitation severity.
• The type of material.

314 Cavitation
6.5 High Resistance Materials

S.C. Li
6.5.1 Introduction
Apart from cavitation damage, corrosion and silt erosion are also responsible
for the destruction of construction materials encountered in hydraulic ma
chinery. Depending on the environment, these three causes may be present
singly or in combinations, making the material selection more stringent. Fur
thermore, the resistances of material to these damage processes are not the
only criteria considered in the process of material selection. Other factors
also affect the selection, such as material strength, machinability, weldability,
formability and costs 14 .
In hydraulic machines subject to cavitation damage, it is often the case
that carbon steels are used for the body fabrication or casting while zones
prone to cavitation damage are protected by stainless steels. Simoneau,1984
[6.46], has shown that such machines are exposed to two different types of
cavitation. One is the high-intensity cavitation, which is often encountered
in high-head/high-flow-velocity machines. It damages small localised areas,
such as blade leading edges, at a high erosion rate (e.g. 0.2 — 3.0 mm/year
on austenitic stainless steels), and examination shows that mechanical de
struction plays a dominant role. The other is the low-intensity cavitation
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
combined with galvanic corrosion, which occurs on the carbon steel at the
interface between the carbon steel and the stainless steel. This damage af
fects much larger areas with erosion rates varying from 0.5 to 2 mm/year.
As explained, corrosion plays a significant role 15 in this type of damage (not
only in sea water even in river water) since the attack from cavitation alone
is weak.
6.5.2 Fused Materials

Owing to good formability, cast irons are extensively used for making parts
of small/ordinary pumps, such as their impellers, diffusers and casings, since
14
These costs include the cost for the raw material itself and the additional cost incurred
in casting, machining and heat treatment owing to the particular difficulties encountered
for the selected material.
15
This corrosion role is often underestimated, although it is the worst part of the whole
cavitation-damage problem in some particular applications where the water possesses a
low value of pH or contains chloride, see [6.46].

welding is not necessary on these parts either during the manufacturing or

in future repairs 16 .
Low-manganese steels are widely used for turbines and large pumps (be
cause they are stronger and weldable although they are still vulnerable as
cast irons to cavitation attack), with cavitation-prone areas protected by
austenitic stainless steels which are either clad or overlayed. The low-Mn
steel often used contains 0.2% carbon and 1% manganese.
Present-day practice is to use the materials with better cavitation resis
tance such as the martensitic stainless steel (containing 13% Cr and 4% Ni
or 17% Cr and 7% Ni) to replace the low-Mn steel for fabricating machines,
particularly for high-head/high-flow-velocity machines 17 . These machines
show less cavitation damage with repair intervals varying from 2 to 4 years.
The damaged areas are then repaired with a properly selected stainless steel
either by cladding or overlaying. Having an expected service life of 60 years
for turbines, the electric utilities often wish to operate the machines with
out major maintenance for at least 30 years. The experience (Simoneau,
1984 [6.46]) from some existing excellent machines do show the possibility of
meeting such a requirement if the runners are fully of stainless steel and the
high-intensity cavitation zones are protected by cobalt-based alloys such as
Stellite 21 overlay after the first field repair.
Many stainless steels have been developed for fabrication and repair pur
poses. The systematic study on this subject done by Rao & Kung, 1987
[6.39], covers a wide range of commonly used metallic overlay materials,
which are presented in Table 6.2. In contrast to the well-known correlation
of cavitation resistance with the material's mechanical properties, such as ul
timate resilience UR and hardness BHN (e.g. Hammitt, 1983 [6.17])18, the
cavitation resistance of stainless steels, nickel-equivalent (e.g. cobalt-based)
alloys can be better correlated with their chromium and nickel-equivalent
parameters, Creq and Nieq. The values of Creq and Nieq can be estimated
by Delong's formulae:
16
For small pumps, in their service life (5 — 10 years) no repair welding is expected on
these cast-iron parts. In contrast, turbines and storage pumps are designed to last a much
longer life, for say 60 years, and repair welding on the runner etc. is often expected.
17
Austenitic stainless steels have excellent formability, castability and weldability and
machinability. They are also less prone to weld cracking and have better cavitation re
sistance than martensitic stainless steels. Therefore, some practices (Medley, 1992 [6.30])
suggested the use of low-nickel austenitic stainless steels for the construction of turbine
runners, although their strength, abrasion and galling resistances are lower.
18
For most materials, the cavitation resistance might be further treated to be roughly
proportional to BH£ , here E is the elastic modulus.
--``,```,,``````,,,,``,`,,`,`,``-`

316 Cavitation
Composition (weight %)
Alloy Common
Name C N Mn Ni Co Cu Si Cr Mo Nb V Ti W P S Fe
A TSN .04 - 0.4 7.0 . 1.5 0.7 15.0 - 0.7 . 7 ? Bal

B 17-4 .08 - 0.3 5.0 - 3.2 0.4 16.4 0.2 .3 .05 - - .027 .015 Bal
C 301 .07 - 0.9 8.0 - 0.2 0.5 17.3 0.3 - .05 - - .022 .007 Bal
D 303 .03 - 1.9 9.0 - 0.5 0.6 17.8 0.5 - .05 - - .035 .190 Bal
E 347-L .06 - 0.8 9.7 - - 0.8 18.2 0.2 0.3 .06 - - .019 .014 Bal
F 308-L .03 - 0.9 9.9 - - 0.8 19.4 0.3 - .05 - - .017 .014 Bal
G 308 .08 - 0.7 9.6 - - 0.8 19.1 0.2 - .07 - - .025 .012 Bal
H 316-L .05 - 0.9 11.2 - - 0.7 16.2 2.5 - .05 - - .027 .012 Bal
I 316 .10 - 1.0 11.0 - 0.2 0.7 17.0 1.8 - .08 - - .021 .015 Bal
J 309 .08 - 1.1 12.4 - - 0.6 22.7 0.1 - .07 - - .021 .020 Bal
K XRD .04 - 0.8 16.7 - - 0.7 16.3 5.5 - . .032 .024 Bal
L P5 .04 - 1.1 13.2 - - 0.9 22.0 2.8 - .06- - .032 .026 Bal
M P4 .10 - 2.0 12.9 - - 0.2 22.0 2.8 - . . . .026 .017 Bal
N 310 .11 - 1.1 19.0 - - 0.6 26.9 0.1 - .08 - - .020 .016 Bal
O 254-SLX .07 - 1.2 22.7 - 1.6 1.0 20.8 4.2 - . . . .027 .023 Bal
P 320 .04 - 0.7 31.4 - 0.9 18.9 2.0 - . . . 7 ? Bal
Q 3A-395 .11 .20 1.6 25.0 - 0.7 15.2 6.2 0.7 . .040 .020 Bal
R 330 .22 - 1.8 32.7 - - 0.5 15.3 0.6 - . . 7 ? Bal
s FAIVPW .03 .13 0.7 9.4 - 0.1 0.8 21.6 3.5 - .06 - - .030 .020 Bal
T F-255 .05 .18 1.0 6.7 - 1.7 0.4 26.1 3.8 - .06 .03 - .006 .009 Bal
U 312 .09 - 0.6 8.1 - - 0.8 30.8 0.1 - .06- - .024 .009 Bal
V 446 .15 .15 1.4 0.5 - - 0.2 27.2 - .016 .016 Bal
w CA6NM .04 - 0.6 4.0 - - 0.5 12.9 0.7 - . . . .012 .018 Bal
7
X 13/2 .06 - 0.5 2.5 - 1.1 0.1 13.3 - - ? Bal
a 1-182 .04 - 7.1 63.5 - 0.1 1.0 12.9 0.05 1.7* - .3 - .011 .004 13.2
b 068-HH .08 - 5.1 64.9 - 1.0 18.1 0.7 2.1* . .009 .005 8.0
c 1-112 .17 - 0.4 55.1 - 0.3 3.5 19.9 6.8 3.6* . . .010 .006 5.9
d H-716 1.10 - 1.0 23.0 11.5 1.5 26.0 3.0 - - - 3.5 7 ? 29.0
e H-25 .08 - 1.1 9.4 43.8 - 0.3 19.1 - - - - 8.3 7 ? 17.1
f E-25 .14 - 1.1 10.9 50.0 7 20.5 - - - - 14.2 .015 .019 2.3
g C-227 .38 - 1.0 3.5 60.5 0.1 0.1 27.1 4.4 - . . . ? .012 2.7
h S-6B 1.10 - - - 66.9 - 28.0 - - - - 4.0 7 7
j S-21 .22 - 0.5 2 3 59.0 - 0.1 29.3 4.9 - - ? .013 3.4
* Ta + Nb
Table 6.2: Stainless, nickel and cobalt alloys (Rao & Kung, 1987 [6.39])
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

Cr,„
eq Md30 8 Fcrrite Erosion rate mg/h Erosion ratio
Alloy Common Cr Ni
cq
eq
Name C
'eq+Nieq °C Vol % Low High High/Low
Intensity Intensity
A TSN 16.4 9.1 0.64 +123 4.0 11.0 2.8

B 17^1 17.4 9.0 0.66 +125 2.9 7.0 2.4
C 301 18.3 10.5 0.63 +29 4.4 12.0 2.7
D 303 19.3 10.8 0.64 +29 2.8 - -
E 347-L 19.8 11.9 0.61 +9 4 4.2 14.1 3.2
F 308-L 20.8 11.3 0.64 -2 2 4.2 12.8 3.0
G 308 20.5 12.4 0.62 -11 7 4.0 12.1 3.0
H 316-L 19.7 13.1 0.60 -31 3 4.5 13.3 3.0
I 316 19.9 14.5 0.58 -51 1 4.9 13.8 2.8
J 309 23.7 15.3 0.60 -112 5 4.7 18.6 3.9
K XRD 22.9 18.3 0.56 -193 19.8
L P5 26.1 15.0 0.63 -156 7 5.6 13.4 2.4
M P4 25.1 16.9 0.60 -145 9 6.0 12.0 2.0
N 310 27.8 22.9 0.55 -349 1 7.6 25.0 3.3
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
O 254-SLX 26.5 26.1 0.50 -294 6.6 22.3 3.4

P 320 22.3 33.0 0.4 -462 9.4 28.1 3.0
Q 3A-395 22.8 35.1 0.39 -498 10.4 32.7 3.1
R 330 16.6 39.5 0.30 ^188 7.2 24.5 3.4
S FAL/PW 26.3 14.4 0.65 -164 15» 5.1 6.8 1.3
T F-255 30.6 14.8 0.67 -178 35« 2.1 5.9 2.8
U 312 32.1 11.0 0.74 -202 70* 2.8 8.6 3.1
V 446 27.6 10.2 0.73 -38 85* 8.1 26.6 3.3
W CA6NM 14.5 5.4 0.73 +200 6.7 15.0 2.2
X 13/2 13.5 5.0 0.73 +232 8.9 20.0 2.2
a 1-182 15.2 68.3 0.18 8.3 24.6 3.0
b 068-HH 21.4 69.9 0.23 9.2 29.9 3.2
c 1-112 33.8 60.4 0.36 4.4 12.9 2.9
d H-716 33.8 61.2 0.36 6.8
e H-25 25.5 30.4 0.46 2.6 10.0 3.8
f E-25 30.8 36.1 0.46 6.1
g C-227 31.7 40.2 0.44 0.4 1.2 3.2
h S-6B 30.9 60.4 0.34 1.3
j S-21 34.4 33.4 0.51 0.3 1.3 4.3
* Metallographically estimated figures
Table 6.3: Chromium & nickel equivalents, 6 ferrite content and erosion rate
(Rao & Kung, 1987 [6.39])

318 Cavitation
Nickel Equivalent
•
R
• Q
•
p tff>
y
Austeniie
0» ^/ / ^y
K
•
^ \ A+M \ \ i • J,
Martenite
t+F>
\ F+M
Ferrite
M+F ^--^
1
10 20 30 40
Chromium Equivalent
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
Figure 6.32: Schaeffler diagram (Rao & Kung, 1987 [6.39])

Erosion rate (mg/h)
40
Distilled water
Q«
32
i■b f\
\ p \
\ \ N : • v;
24 a jj.Y.f.R
■ \ \.v;
:
16
w;
c ■
4
!.U.J
d >D \ ^
\
\
0.2 0.4 0.6 0.8
C'eqêq+Nieq)
Figure 6.33: Erosion rate correlation with chromium and nickel equivalents,
Creq and Nieq, for stainless steels, nickel based alloys and nickel equivalent
(e.g. cobalt) based alloys (Akhtar, 1986 [6.1])
Creq = Cr + Mo + 1.5St + 0.5JV6,

Nieq = Ni + 0.5Mn + 30{C + N);
here, each element is calculated based on its weight percentage. For some al
loys containing copper and cobalt, the contributions from them are weighted
as 0.44Cu and OAlCo to be added to the nickel equivalent Nieq. The calcu
lated values of Creq and Nieq for these materials are listed in Table 6.3 and
also shown on the Schaeffler diagram, Fig. 6.32. The cavitation resistances
of stainless steels, nickel-based alloys and nickel-equivalent (cobalt-based) al-
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

320 Cavitation
loys listed in Table 6.2 are well predicted by this correlation (see Fig. 6.33).
For stainless steels (the right curve), there is a minimum of erosion rate
corresponding to
Cr
^ 0.7
Crea + Nieq
where the stainless steel T (i.e. F255), which has a duplex ferrite-austenite
microstructure (35% ferrite), locates, while stainless steels U, V, W and X
(i.e. 312, 446, CA6NMand 13/2), which are mainly of ferritic and marten-
sitic structure, locate on the ascending portion of curve. The fully austenitic
weld metals O, P, Q and R (i.e. 254SLX, 320, 3A395 and 330) with low val-
ues
°^ Cr +Ni P o s s e s s l ° w resistances. Stainless steel R (i.e. 330) behaves
as nickel-based alloy since it locates on the nickel-based curve (the left curve)
rather than on the stainless-steel curve. This is owing to its much higher
nickel-equivalent element (0.22% carbon) which gives a total value of Nieq
up to 40%. Owing to a similar reason, alloys d (H716) can be considered as
a nickel-based alloy as well.
The curve for nickel or nickel-equivalent-based alloys shows that the min
imum erosion rate occurs when the value of Cr ^.'jfo is around 0.42. The
cobalt alloys g, j and h (C227, S6B and S21) possessing strongest cavitation
resistances locate just in this region. It should be noted that although cobalt
alloys g (C227) and j (S21) all contain about 60% cobalt as alloy h {S6B)
does, their carbon levels are much lower (0.38% and 0.22% respectively)
compared with 1.10% carbon for alloy h (S6B). The lower level of carbon
can ease the recontouring of the deposited overlay of these two cobalt alloys.
A subsequently developed high-resistance alloy TS1 (1.0% C, 10% M, 12%
Co, 5% Si, 30% Cr and 42% Fe) with a value of Cr^Nieq « 0.45 is well
predicted by the curve as shown in Fig. 6.33.
Rao & Kung also discuss the metallographical factors governing the ero
sion of materials. Firstly, the stress-induced phase transformation from
austensite to martensite is a measure of cavitation resistance for fully austenitic
and austenitic-ferritic duplex stainless steels 19 . The cavitation resistance is
proportional to the transformation level20 as shown in Fig. 6.34. Whereas,
the thermal martensite is unfavourable to the cavitation resistance, presum-
19
This measure is valid only for the duplex stainless steels containing a small amount of
ferrite. For high-intensity cavitation (±25/im, vibratory test), the limit for the amount of
ferrite is up to 5%; for low-intensity cavitation (±12ftm, vibratory test), it is up to 15%.
20
The transformation level is measured by the temperature Mdm at which 50% marten
site is produced under the action of a true strain of 0.30.
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

Erosion rue (mg/h)
...
\ ^
Q*
V
0
•^^
* ^ _ _ „ _
~^~^K
~^^ E
L 0 \ , w ■»•
OM I • •V! ^-^
c
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
UO s
Amplitude * 25 \l m
^~^«^.^ 7
B ^
T 0°
(dm/dt) , =13-0.035', M
25 d30
200
1
(a) d30 < ° O
Erosion rate (mg/h)
10 »Q \
p ^ \
V
8
\ « N
^« O ^ O * * ^
w»
5
L *^^
Ô « ^ *C
A
4 - U M► E Vj
Distilled water U< >

! ■ ^ ^ ^ ^
•D^\J
Amplitude * 1! |l ■
al
^
(drtfdt) =3.9-0.0118 M „ „
12 d30
n
200
1
d30 < ° c>
(W
Figure 6.34: Erosion rate vs. Md30 value (Rao & Kung, 1987 [6.39])
Filled circle: austenitic alloys (<5 ferrite < 5%); Open circle: duplex
austenitic-ferritic alloys (6 ferrite > 5%); Triangle: triplex austenitic-ferritic-
martensitic alloys.
(a) High-intensity cavitation (b) Low-intensity cavitation

322 Cavitation
ably, because the thermal martensite has a tetragonal structure compared

with the body-centred cubic structure possessed by the stress-induced a'
martensite. This explains why the alloys containing thermal martensite, i.e.
the duplex alloys in the austenite-martensite region such as A (TSN), B
(17/4), W (CA6NM) and X (13/2), have higher erosion rates than their
counterparts (in terms of possessing same Mefao values) fully austenitic al
loys. Secondly, the 8 ferrite plays a favourable role in reducing erosion rate
for the duplex austenitic-ferritic alloys containing 5 — 80% (for high-intensity
cavitation) or 10 — 80% (for low-intensity cavitation) of 8 ferrite, such as L
(PS), M (P4), S (FAL/PW), T (F255) and U (312) which contain 8 ferrite
of 7%, 9%, 15%, 35% and 70% respectively, referring to Figs. 6.34, 6.35
and Table 6.3. The dependence, of the 8 ferrite influence, on the intensity
Change in erosion rate (mg/h)

16
12 Distilled water
Amplitude
8 • ± 25 u m
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
f 12 um
4
LM -
0
-4
-8
-12
0 20 40 60 80 100
% 8 Ferrite
Figure 6.35: The erosion rate difference between duplex austenitic-ferritic

alloys and their fully austenitic counterparts (Rao & Kung, 1987 [6.39])
of cavitation attack is presumably due to the range-compatibility between

the stress induced by cavitation impacts and the dispersed 8 ferrite phase
in the material where the 8 ferrite phase plays an intrinsically different role
in resisting erosion from the austenite phase 21 . Thirdly, for cobalt-based
alloys, their excellent cavitation resistance is due to the fine deformation
21
The 5 ferrite phase is superior to the austenite phase, although it can not sustain
indefinite implosion impacts due to its high strain rate sensitivity, because the value of

twinning accompanying the strain-induced martensitic transformation un

der a low stacking fault energy, SFE, close to the optimum value (approx.
20 mJ/m2), which can be achieved by the judicious selection of alloying
elements, because some elements such as iron and nickel increase the SFE
value, and some, such as chromium, molybdenum and tungsten, decrease
the value. The maximum value of Ni + Fe permissible to obtain an optimum
cavitation resistance is about 11%. The threshold value for C to increase
resistance is 0.2% and further increase of C will not improve erosion resis
tance, which indicates that the outstanding resistance of cobalt-based alloys
is mainly from the matrix phase not from carbide volume fraction.
C Mn Si Cr Ni N
Nitronic 60 0.07 8.0 4.0 16.5 8.5 0.14
304 0.06 1.5 0.5 19.0 9.0 0.03
Table 6.4: Chemical composition of Nitronic 60 (Schumacher, 1989 [6.45])
There are other materials often used. One is Nitronic 60 introduced by

Armoc Inc in 1975. It is a nitrogen-strengthened austenitic stainless steel
with high levels of manganese and silicon as compared with 304, referring to
Table 6.4. Although its cavitation resistance is not as high as that of cobalt
alloys, it provides much lower material and fabricating costs while offers the
best combination performance of corrosion, wear, galling and cavitation re
sistances among all stainless steels tested by Schumacher, 1989 [6.30]. It
is a preferable material for the applications where the cavitation erosion,
wear and galling are equally important, e.g. Nitronic 60 wear rings used
to minimise galling of wicket gates in the sealing plate area, Nitronic 60
weld wire deposited directly to the CA6NM guide vanes and weld patches
of Nitronic 60 on turbine runner blades etc. The relative cavitation resis
tance of Nitronic 60 is listed in Table 6.5 22 . The newly developed Ireca and
Tribonic 20 are another two materials which possess superior comprehen
sive characteristics (Medley, 1992 [6.30]). Ireca is an iron-based alloy with
low cobalt content developed by Hydro-Quebec, Canada, referring to Table
^—^rfn— f ° r ^ f e r r ite phase is more close to the optimum value, 0.7. Take alloy T (F255)
as an example, the value is found to be 0.71 for S ferrite phase and 0.65 for austenite
phase.
22
These data are derived from different sources. Therefore, they are not quantitatively
correct. Table 6.5 should be used as a guide only.
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

324 Cavitation
(A) Data from Rao & Kung, 1987 [6.39]

Material Relative Resistance
Cast iron 0-1 and 0-4
Manganese bronze 0.5
Cast carbon steel 1.0
Carbon steel plate 1.0
Aluminium bronze 3.6
309 stainless steel overlay 5.0
Cast stainless steel {CA6NM) 5.5
Cast stainless steel (CF8) 5.5
Stainless steel plate (304) 5.5
308 stainless steel overlay 7.1
Haynes 25 9.2
Nitronic 60 13.2
Ferrallium 255 14.5
Stellite 21 49.0
Note: The relative resistance is normalised in respect to carbon steel.
(B) Data from Schumacher, 1989 [6.45]
TEST 1 (by Corps of Engineers, USA)
Material Relative Erosion Rate
Nitronic 60 1.00
308L 1.89
304 3.67
CA6NM 6.80
Note: Vibratory test ASTM G32
TEST 2 Weldments (by TVA.USA)
Material Relative Erosion Rate
Cobalt alloy 6 0.76
Nitronic 60 1.00
308L 3.38
316 4.62
Note: High-pressure jet impingement test
Table 6.5: Relative cavitation resistance
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

6.6. Hydroloy HQ 913 is the name for Ireca used for weld overlay purposes.
Ireca possesses high cavitation resistance comparable with Stellite 6 and
Stellite 21 (see Table 6.7), owing to its strain-induced fine twinning asso
ciated with low stacking fault energy, along with its better weldability and
surface grinding characteristics similar to 308 stainless steel. Tribonic 20
c Mn Si Cr Ni Co N
0.20 10 3 17 Nil 8 0.2
Table 6.6: /reca-Chemical composition (weight %) (Medley, 1992 [6.30])
CA6NM CA6NM-L 308 301 S21 S6 Hydroloy Ireca (cast)

15 27 15 6 1.4 0.7 1.2 1.0
Table 6.7: /reco-Erosion rate comparison (Medley, 1992 [6.30])
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
c Mn Si Cr Ni
0.03 14 5 5 5
Table 6.8: Chemical composition of Tribonic 20 (Medley, 1992 [6.30])
is a fully austenitic iron-based alloy, which should not be categorised as a

stainless steel owing to its low content of chromium, referring to Table 6.8.
Its cavitation resistance is four times of 304 and 316 stainless steels along
with a corrosion resistance equal to stress-relieved JÎO and an outstanding
galling resistance, at least equal to and in some cases superior to Nitronic 60.
Tribonic 20 is suitable for the applications such as turbine runners, seal rings
and shaft sleeves etc. More recently, a heat-fusible metallic powder named
SPHG\ and its compound cladding plates named lCrl8Ni9Ti+SPHGi,
developed by the Chinese Research Institute of Hydro Power, have been
successfully employed to protect the Kaplan turbines ZZ360 and ZZ362 at
the Sanmenxia Hydropower Station (China) against the synergistic damage

326 Cavitation
of cavitation and silt erosion 23 , showing at least 10 times higher resistance

than A132 stainless steel overlay and also superior performance than the
previously used epoxy-ceramic compound coating.
Other factors such as stress status, heat treatment and corrosion also
affect the cavitation resistance of the material.
As to the status and level of stress, the vibratory test conducted by Rao
& Kung, 1987 [6.39], shows: (1) For the carbon steels tested, the erosion
rate of specimens containing biaxial compressive stress becomes higher than
at stress-free status, while it decreases from the value for stress-free status
if biaxial tensile stress is applied; (2) For stainless steels, the applied stress
lowers the cavitation resistance of martensitic stainless steel {CA6NM) and
austenitic stainless steel (Rao et al, 1986 [6.40]), and improves the cavi
tation resistance of the duplex austenitic-ferritic stainless steel (F255); (3)
The residual stress caused by weld has a complex influence on cavitation
resistance because of the involvement of three different materials, i.e. the
weldment, base material and their interface, whose chemical compositions
are further changed dramatically due to the dilution process occurring in the
welding. Hence, no general rules can be followed to predict the influence of
weld stress on cavitation resistance. Each case must be assessed individually.
As to the effect of corrosion, there are basically six types of corrosion
encountered in hydraulic machinery, namely general corrosion, corrosion-
erosion, crevice corrosion, stress corrosion cracking, galvanic corrosion and
intergranular corrosion (Miller, 1992 [6.31]). Generally speaking, the corro
sive medium, e.g. sea water, will reduce the low-intensity cavitation resis
tance of the weld deposits of metals by 1.2 — 2.5 times, by the synergistic
effect of corrosion and cavitation erosion. Rao & Kung, 1987 [6.39], pointed
out that the alloying element Mo has a detrimental effect on the cavita
tion resistance of fully austenitic stainless steel weld metals in sea water 24
while a favourite effect for duplex austenitic-ferritic stainless steels, nickel
base and cobalt base alloys by improving their crevice corrosion resistance.
The galvanic corrosion by bimetallic coupling will worsen the cavitation re-
23
Sanmenxia P.S. locates on the Yellow River, suffering severe cavitation and silt erosion
problems. The annually average silt concentration passing machines is 15 kg/m3.
24
The cavitation resistance of stainless steel welds in sea water depends mainly on the
pitting corrosion resistance of the austenitic phase whose resistance will be weakened by
the presence of molybdenum in this phase. For fully austenitic stainless steels this is just
the case, while for duplex austenitic-ferritic stainless steels this is not the case because the
molybdenum being a ferrite stabiliser tends to dissolve in the ferritic phase instead of in
the austenitic phase.
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

Metal and condition Mass loss Galvanic current (ma)

A36 cavitating alone 167.6
308L(F) cavitating alone 45.0
F255(T) cavitating alone 21.3 -
S210) cavitating alone 0.1
A36 stationary coupled 3.7 0.31

308L(F) stationary 0.2

F255(T) stationary 0.087

S21(j) stationary 0.046
A36 cavitating coupled 166.3

308L(F) stationary 0.8 0.10

F255(T) stationary 0.4 0.13

S21(j) stationary 0.3 0.15
A36 stationary coupled 24.5

308L(F) cavitating 21.5 2.61

F255(T) cavitating 20.6 1.97

cavitating 0.8 0.23
S21(j)
A36 cavitating 203.7

308L(F) cavitating coupled 35.9 1.53

F255(T) cavitating coupled 9.7 1.21

cavitating coupled 0.3 0.7
S21(j)
Table 6.9: Erosion rates of weldments under synergism of cavitation erosion

and galvanic corrosion (Rao & Kung, 1987 [6.39])
Base metal: Carbon steel {ASTM A36)
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

328 Cavitation
Metal and condition Mass loss Galvanic current (ma)

13/4 cavitating alone 61.3
308L(F) cavitating alone 45.0 -
F255(T) cavitating alone 21.3
S21(j) cavitating alone 0.1
13/4 stationary coupled 0.043 0.001



13/4 cavitating 42.7

--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
coupled 0.037
13/4 cavitating coupled 44.5 0.024

13/4 cavitating coupled 44.9 0.004

13/4 stationary coupled 0.2 -0.11

308L(F) cavitating 39.9

F255(T) cavitating 27.2

S21(j) cavitating 0.1
13/4 cavitating 51.1 0.06

308L(F) cavitating coupled 36.9

F255(T) cavitating coupled 13.7

cavitating coupled 0.2
S21(j)
Table 6.10: Erosion rates of weldments under synergism of cavitation erosion

and galvanic corrosion (Rao & Kung, 1987 [6.39])
Base metal: Martensitic stainless steel (13/4)

Erosion rate (mg/h)

Heat treatment Alloy Micro-structure Low intensity High intensity
change
As Heat As Heat
deposited treated deposited treated
E 4.2 3.0 14.1 7.8

630°C-24 hours G Sensitization 4.0 3.4 12.0 13.6
Cooledatl2°C/h I 4.9 5.1 13.8 13.5
593°C-4 hours A 4.0 5.1 11.0 14.4

Air cooled B Stress relief
2.9 3.7 7.0 10.9
870°C-24 hours Sigma phase

Furnace cooled T 2.1 1.4 5.9 3.2
(a)
Alloy Condition Hardness (Re)
As deposited 28-33
A
Heat treated 23-28
B As deposited 36-39
Heat treated 34-37
As deposited 23-25
T
Heat treated 35-37
(b)
Table 6.11: Heat effect of post-weld heat treatment (Rao & Kung, 1987
[6.39])
(a) Erosion rate (b) Hardness
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

330 Cavitation
sistance of the coupled material with anodic polarity. The worst galvanic
effect occurs for the weld overlay of duplex austenitic-ferritic stainless steel
308L applied either on carbon steel A36 or on martensitic stainless steel
13/4 base metals, referring to Tables 6.9 and 6.10. For these two base
metals, i.e. A36 and 13/4, the lowest galvanic effect can be achieved by
using cobalt-based alloy S21 as weld overlay material. The sensitisation of
austenitic stainless steels by the welding heat is the main cause responsible
for the intergranular-corrosion affected resistance deterioration 25 . To avoid
the sensitisation effect, one method is to use low-carbon stainless-steels and
low-interpass temperature.
As to the heat effect, apart from the sensitisation caused by welding heat,
the post-weld heat treatment is the most common heat effect encountered in
practice. The triplex austenitic-ferritic-martensitic alloys A (TSN) and B
(17/4) erode more rapidly after heat treatment (stress relief), presumably
due to the reduction of their hardness, referring to Table 6.11. While the
alloys E (347L), G (308) and I (316), which contain small amounts of 8 fer-
rite, exhibit a variation in erosion-rate change as shown in Table 6.11. The
alloying elements Mo and Nb have a beneficial effect on the post-weld heat
treatment as demonstrated by alloy E (347L). The duplex austenitic-ferritic
stainless steel T (F25S) after the heat treatment shows significant improve
ment on cavitation resistance (reducing erosion rate by a factor of two) due
to the formation of a phase which strengthens 6 ferrite and delays its disinte
gration. Therefore, Rao & Kung conclude that the duplex austenitic-ferritic
stainless steel T (F25S) has best overall performance, although its cavita
tion resistance is lower than the cobalt-based alloy j (S21). The excellent
weldability and machinability of T (F25S) make this material the optimum
metallic overlay material for cavitation repair.
6.5.3 Non-fused Materials

--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
Most of non-fused materials are being used on a trial experimental basis.

The data from various sources are often not comparable owing to the lack
of test identity. The material properties and applying procedures provided
25
Sensitisation is such a phenomenon: when an austenitic stainless steel is exposed
to a temperature range between 1150° to 1250° F, the carbon in the material will diffuse
towards the grain boundaries where the high concentration of carbon ties up the chromium
by forming chromium carbides at grain boundaries, which leaves a zone of chromium
depletion in the immediate area around the grain boundary making this area less corrosion
resistant than the bulk material and resulting in intergTanular corrosion (Miller, 1992
[6.31]).

by manufacturers may vary dramatically. According to the survey (Sinclair

& Rodrigue, 1986 [6.47]), about one third of repairs on turbine and pump
designs in the USA have been made with non-fused materials. However,
only one third of these repairs was successful or partially successful. The
information obtained from the pilot test on abrasion and cavitation resis
tant materials 26 organised by the Chinese National Experiment Centre for
Abrasion and Cavitation in Hydraulic Machinery, 1987 [6.48], may be used
as guidelines. The five types of non-fused materials often used are epox-
ies, polyurethane-based plastics, neoprenes, ceramic compounds and plasma
sprayings.
The tough and elastic materials such as various epoxies/epoxy com
pounds and polyurethane-based plastics are suitable to be used as protecting
coating materials in the low-intensity cavitation and abrasion areas, partic
ularly in the areas prone to the synergism of galvanic corrosion and low
cavitation erosion (such as the interface between the stainless steel and the
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
carbon steel base). The spiral casings, nozzle pipes, draft tubes and the low-
intensity cavitation area on the blades are such areas suitable to be protected
by the coatings made of these two types of materials.
Various ceramic compounds have been developed since ceramic itself
alone offers little resistance to cavitation attack although its excellent hard
ness is good for abrasion resistance. Some of these compounds possess good
resistance to cavitation owing to their improved toughness. Some of them
even can be used as weld materials such as the metal-ceramic compound
developed by the Chinese Academy of Steel and Iron [6.48]. Ceramic com
pounds are reported to be successful in protecting the areas suffering severe
cavitation and/or abrasion attack such as the areas on blades and guide
vanes under the attack of leading-edge cavitation.
The often-used plasma sprayings are ceramic, polymeric and metallic
coatings. The performance of these spraying coatings in cavitation/abrasion
resistance is mainly determined by the spraying material itself. That is,
ceramic and hard material coatings offer excellent abrasion resistance but
poor cavitation resistance; polymeric coatings reveal better unified resistance
to the synergistic attack of cavitation/abrasion/corrosion, subject to their
particular properties.
Little information is available on the use of neoprenes (only three cases in
26
Total 27 metal and non-metal materials provided by 18 research academics, design in
stitutions, manufacturers and power stations, were tested on a prototype propeller turbine
ZD661-LH-120 under the same synergistic attack of cavitation and silt erosion.

332 Cavitation
the USA), whose resistance, in general, is not as good as polymeric materials.

The advantages of using non-fused materials are:
• The coatings of non-metal materials provide a good protection against

the synergistic action of cavitation erosion and galvanic corrosion;
• Some of these materials, e.g. ceramic compounds, offer excellent cavi

tation and abrasion resistance suitable for the repair of high-intensity
cavitation-damaged areas.
• The applying of non-fused materials will not introduce heat effect and
is less time consuming compared with the welding repair.
However, the main difficulties with the use of these materials are:
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
• The bonding technique is vital to the success of applying these mate

rials. Particularly, achieving a good bonding strength in wet service
condition under the dynamic loads, resulted either from machine work
ing load or cavitation impacts, is a difficult task.
• The welding work in the areas adjacent to these materials will cause
the failure of bonding and may produce toxic fumes.
• The non-conductivity makes the use of arc gouging impossible when

carrying out a future repair work in the areas protected by these non-
fused materials.
Information about some non-fused material products and their appli

cations is briefly quoted here. The trade names of commonly used epox-
ies/epoxy compounds are Devcon27, Nordbak, Resweld, Epoxylite, Melalox
and Velodur. Fibre reinforced epoxy (Con — Chem) and Carbomastic 15
(coal tar epoxy) are also reported to be successful in cavitation repairing
(Sinclair & Rodrigue, 1986 [6.47]). The ceramic epoxy compounds devel
oped jointly by the Research Institute of Yellow River Committee and the
Tianjin Design Institute of Hydro Power (China) offer good resistance to the
synergism of cavitation and abrasion [6.48]. A similar ceramic epoxy product
with the trade name of Dura Tough developed by Palmer International Inc.
(USA) is reported to be 4 times more resistance to cavitation erosion than
27
Available in three forms: plastic steel putty, stainless steel putty and wear resistance
putty.

CA6NM (13/4) stainless steel and twice as resistant as austenitic stainless

steel overlay (308); and it can be applied in situ for low-intensity cavita
tion and abrasion areas. Duratough DP is applied as a paste to fill areas
severely pitted by cavitation, while Duratough DL is a viscous liquid top
coat. Another similar ceramic compound product is Belzona: Belzona R
used as fillings and Belzona S as top coats. The experience of cavitation
repair on the blades of propeller and Kaplan turbines with Belzona in the
USA shows that this material mitigates cavitation erosion, however it does
pit and requires replacement about every two years. The castable polyamide,
Cast polyamid 12 (Lauyamid) developed by Fluorocarbon Ltd and tested by
Schroder, 1987 [6.44], is claimed to offer excellent cavitation, abrasion and
corrosion resistances superior to the high-quality pump-alloy material, G-X
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
3CrNiMoCuN 24-6 (Noridur). It is particularly recommended for making
the propellers of slurry pumps. As to plasma sprayings, according the exper
iments conducted in China [6.48], the nylon compound spraying employed
in the pilot test at the Suide Power Station can provide fairly good coatings
against cavitation and abrasion.

334 Cavitation
6.6 Repair of Machine

S.C. Li
6.6.1 Introduction
The principles of cavitation-damage repairs for turbines and for large pumps,
storage pumps and pump-turbines are basically the same. For small cast-iron
pumps, welding repairs are not practical or economical. The damaged parts
are simply replaced with new parts. Therefore, only the repair philosophy
for turbines and pump-turbines is briefly depicted and demonstrated by two
examples in this section. As to the details of the procedures and standards,
they are very utility-dependent although there are several sources available
for reference, such as Palaev & Agelly, 1974 [6.37]; Sinclair & Rodrigue,
1986 [6.47] and the USBR Power O & M Bulletin No.24 "Welding Repair
to Hydraulic Equipment", etc. The input information from manufacturers
is often important and should be always sought.
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
6.6.2 C a u s e of D a m a g e
Inspection
For analysing the causes of damage, it is vital to obtain a complete set of

data about both the damage itself and the operation history of the machine
which includes operating hours/modes/ranges and liquid qualities such as
PH value, entrained solid particles and oxygen concentration etc. According
to the experience in the USA (Sinclair & Rodrigue, 1986 [6.47]), for accu
mulating such data, the inspection of turbine runners and other components
is done on a frequency of one to four years. For machines with minimum
cavitation damage, a four-year interval is recommended for turbines and
a two-year interval is recommended for pump-turbines. During inspections,
the damage location, extent, volume or weight loss and its appearance should
be recorded by appropriate methods such as:
• Record forms;
• Photographs;
• Quantities of man-hours used and weld material consumed for repair,

which is an approximate estimation of damage quantity;

• Replication technique which provides accurate records of the individual

volume loss and the damage features such as the surface roughness
(even the size and depth of small discrete cavitation pits) by using vinyl
polysilixane polymer (a material used for making dental impression
named Reprosil and Permagum) to make replicas of damage surfaces.
When conducting inspections, the cavitation damage should be distin

guished from other damage such as silt erosion and various corrosions. How
ever, it is often the case that the damage is caused by synergistic actions:
cavitation-silt erosion or cavitation-galvanic corrosion etc.
Analysis
Based on the mechanism of various types of cavitation damage occurring

in the machines as depicted before, the cause of a particular damage detected
on the machine can be identified by means of the in situ examination, model
study and numerical simulation.
In many cases, the causes can be identified just by in situ examina
tions by investigating the damage extent, location and surface features with
reference to the machine operating history (mode/head/output/hours etc).
There are some basic guidelines for the analysis:
• If the damage is very localised and appears immediately downstream

of some surface discontinuities (caused by either poor manufacture or
design mistakes in profiling or previously improper repairs), then the
surface discontinuity should be treated first.
• If the damage varies from blade to blade (in which some blades are
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
damage free), the cause of damage is obviously either non-identical

profiling or incorrect positioning of the blades. The least damaged
blades (or the damage-free ones) should be recognised and used to
provide template for re-profiling or re-positioning the damaged blades.
• If the damage appears on every blade identically and/or on some parts

(e.g. bottom rings) axisymmetrically, the cause may be due to in
correct machine designs (parametric selection, fluid passage profiling
and blade contours etc) or inadequate plant cavitation number av or
improper operating range (head and output).
• If the damage is persistent or even reveals an aggressive form after pre

vious repairs, then the previous repair method should be questioned.

336 Cavitation
• If at the next inspection the damage reduces after previous repairs,

then one can conclude that the analysis of the cause and the repair
approach used were basically correct, so future repairs, if requested,
should be carried out to the same procedures.
A list of some damage causes and remedial measures can be found in [6.47].
However, some cases of damage do necessitate further model and numeri
cal studies since in situ examination does not always provide satisfactory
answers. Model study can provide the information about the cavitation de
velopment, its pattern and damage 28 . Computer simulation of flows helps
identify the critical flow structures which could potentially induce cavitation
(damage), such as flow separation, minimum-pressure spots and vortices. It
is often the case that these three measures are performed integrally to pin
point the damage causes and to find the satisfactory solutions.
6.6.3 Main Concerns in Repair

Several factors should be considered when making repair programmes.
Repair Method
The repair methods often used are:
• Overlay and cladding

The overlay is the most widely used method. The damaged surface is
removed by arc gouging etc then filled by weld materials with equal
or superior cavitation resistance. Cladding is also a common method.
The damaged area is removed then filled by welding cladding plates
instead of weld overlays. The stainless-steel plates are welded around
the edges and in the inner holes. The method of cladding may reduce
repair time and heat input. However, it may fail under dynamic load,
and it is also prone to galvanic corrosion developed underneath the
plate.
• Non-fused material filling and coating

Non-fused materials are often used for the repair of damaged areas
suffering from low-intensity cavitation because of their low cavitation
resistance. They can also be used for temporary repairs owing to ease
28
Various techniques such as paint removal, pressure-sensitive film and oil-film trace can
be used for this purpose
--``,```,,``````,,,,``,`,,`

of application. Apart from filling repairs, non-fused materials such as

epoxies are also used as coatings to protect the interface of stainless
steel overlay and carbon steel base from the galvanic corrosion 29 .
Adding Anti-cavitation Devices
Various anti-cavitation devices can be chosen to minimise cavitation and

its damage. For example, the aeration of cavitation zones is an effective
method. Small amounts of air can be injected into cavitation areas on blades,
to provide an air cushion, to reduce the blade cavitation damage and noise.
The volume of air injected is about 0.1 — 0.2% of the machine flow rate. For
Francis turbines, air is often introduced into the cavitation zone in the draft
tube to suppress cavitation and associated pressure fluctuations. However,
over-injection of air will reduce machine efficiency. Usually, 1.5 — 2.0% of
the machine flow rate is employed. Other devices, e.g. the fins installed on
the periphery of the blade suction side for propeller and Kaplan machines,
have been adopted for reducing tip-cavitation damage.
Repair Frequency
According to the American experience (Sinclair & Rodrigue, 1986 [6.47]),

the frequency of turbine cavitation repairs varies from plant to plant, depend
ing on the rate of metal removal and the plant owner's philosophy:
• Make all repairs during each inspection period;
• Repair only areas where cavitation damage is 1/8 in (3 mm) or deeper;
• Repair areas on stainless-steel overlay where pitting is 1/8 in (3 mm)

or deeper. On carbon steel, repair areas even with light damage using
stainless-steel weld materials;
• Allow cavitation damage to progress to the maximum depth, 3/8 in

(10 mm), which can be repaired with two weld passes;
• Make repair only when damage threatens to impair the strength of the
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
turbine.
29
In this application, the stainless-steel overlay should be extended to such a large area
that the interface locates in the low-intensity cavitation zone so that the epoxy coating
can stay there without the risk of being removed by cavitation attack.

338 Cavitation
In Russia (Pylaev & Agelly, 1974 [6.37]), the criteria determining the
repair frequency are:
• The damage should not exceed such a severity that the machine can
not still be repaired in situ. The allowable repair depth on stainless
steel without preheat varies depending on the material properties and
the weld procedure. The value of this depth is 5 — 8 mm for 20Crl3Ni;
and 1 0 - 1 2 mm for 0Crl2NiCu;
• For runners made of carbon steel, the criterion is that the deterioration
of machine efficiency should not exceed a financially reasonable level.
Experience shows that the repair work is justified following a 0.2%
efficiency drop.
The optimum repair interval, T (hrs), could be determined in such a

way that the sum of the repair cost per hour, r ^°"", and the hourly cost
incurred due to the efficiency loss, —3^, reaches minimum as shown in Fig.
6.36. The repair cost, C r e p a , r , includes the direct repair cost and the indirect
C repair + C eff
Cost
—»" T
Topt
(Repair interval, hours)
Figure 6.36: Determination of optimum repair interval
cost which consists of both the unit downtime cost (i.e. generating loss and
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

capacity value loss) incurred due to cavitation repair and the cost of machine
service-life loss owing to extensive damage repairs.
Selection of Weld Materials
According to the American experience [6.47], the following is recom

mended.
• For turbine components with carbon-steel base material

Areas where the depth of weld is greater than 3/8 in (10 mm) should
be built up to 3/8 in (10 mm) depth with E7018 mild-steel weld ma
terial (low-moisture type). The remainder of the area, together with
the other areas with depth of 3/8 in (10 mm) or less, should be filled
with 300 series austenitic stainless steel material. The first pass of
material should be 309 stainless steel 30 , the remainder 308 stainless
steel. Stainless-steel material should not be used in areas of deep dam
age to avoid blade distortion resulted from the difference of thermal
expansions between carbon and stainless steels.
• For martensitic stainless steel (CA6NM)

In areas of low stress level and in areas of high stress where depth of
damage does not exceed 10% of the material thickness, repairs should
be made with 308 austenitic stainless-steel weld material. In areas of
high stress, where the depth of damage exceeds 10% of the material
thickness, the turbine manufacturer should be consulted. For this,
a 410Ni-Mo weld material may be appropriate because chemical and
physical properties are similar to those of the base material. This
weld repair material requires high preheat to prevent hydrogen-induced
cracking and post-weld heat treatment at 1000° F (540° C) to temper
the weld metal. Repair with 15Cr25Ni austenitic weld material and a
overlay of a 50% cobalt-based alloy will result in a repair with good
physical and cavitation-resistance properties without post-weld heat
treatment.
• For austenitic stainless steel materials (including overlay)

--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
The damaged areas should be repaired with 308 austenitic weld mate
rial.
30
The high level of ferrite in 309 stainless steel reduces the risk of hot-short cracking
(micro Assuring).

340 Cavitation
In the areas suffering severe cavitation damage, the following weld ma
terials may be chosen for repairs, referring to §6.6.2 "Fused Material".
• The duplex austenitic-ferritic stainless steel containing 35% 5 ferrite

such as F255 (Ferralium 255) possesses excellent weldability and the
weld metal can be easily ground to the required profile. Its cost is
comparable to other stainless steels.
• The nitrogen-strengthened austenitic stainless steel Nitronic 60 offers

the best comprehensive resistances with reasonable cost (much lower
than cobalt alloys).
• The fully austenitic iron-based alloy Tribonic 20 offers even better

galling resistance than Nitronic 60.
• The cobalt (59%) based alloy, S21 (Stellite 21), has far superior cav
itation resistance. However, it can not be used in highly restrained
areas such as deep weldments and cracks since it is vulnerable to hot
cracking (micro Assuring). Therefore, it is only used as a final weld
material, such as the final three passes over the weld metal F255. This
material is expensive and its machinability (grinding) is poor.
• The cobalt (43.8%) alloy H25 (Haynes 25) has better cavitation resis
tance than F255.
• The iron-based alloy Ireca (weld metal named as Hydrology HQ 913),

which has fairly low cobalt (8%), offers the cavitation resistance equiv
alent to 60% of S21 while possessing much better weldability and grind
ing characteristics, similar to 308 stainless steel.
Preheating
Before welding, proper preheating can avoid the hydrogen-induced crack

ing and the excessive ductility loss. The amount of preheat depends on the
chemical content of the base metal 31 , the thickness of the base metal, the
restraint and rigidity of the area being repaired, the actual heat input dur
ing the welding process, and the type of weld material (Sinclair & Rodrigue,
1986 [6.47]):
Particularly, the carbon equivalent, Ccq, of the base metal which is evaluated as
Ceq = C + ±Mn + \{Cr + Mo + V) + ±{Ni + Cu).
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

• Carbon steel weld material and carbon steel base metal

The required preheat may be determined from the carbon equivalent
and the thickness of base material from Table 6.12. The minimum
preheat for typical carbon materials used in turbines is given in Table
6.13.
Base Material Carbon Equivalent

Thickness (in) 0.35 0.40 0.45 0.50 0.55
0.5 60 60 60 125 200
0.75 60 60 125 250 300
1.0 60 60 200 300 350
1.5 60 150 250 350 400
>2 60 200 300 400 400
Recommended temperatures are based on:

-Use of low hydrogen electrodes; and
-Minimum arc energy of 0.30 kj/in.
Table 6.12: Recommended minimum preheat temperature (in degrees Faren-

heit) (Sinclair & Rodrigue, 1986 [6.47])
• Austenitic stainless steel weld material on carbon-steel base

--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
Due to the fact that the austenitic stainless steel can absorb hydrogen
to avoid underbead cracking, the preheat may be 150° F (83° C) less
than those given for carbon-steel weld materials, but should not be less
than 60° F (16° C). However, austenitic stainless-steel weld metal is
susceptible to intergranular corrosion caused by the sensitisation (see
§6.5.2 'Fused Materials'). The low-carbon stainless-steel weld metal,
such as AISI-316L stainless steel (only 0.03% C), is less susceptible
to sensitisation. To avoid sensitisation, a low interpass temperature is
also recommended.
• Austenitic stainless-steel weld material on austenitic stainless steel

A nominal 60° F (16° C) preheat is necessary.
• Austenitic stainless-steel weld material on martensitic stainless steel

(CA6NM)

342 Cavitation
Material Thickness Carbon Minimum

(in) equivalent preheat
(ASTM) (%) (OF)
A27Gr 60-30 0.40 60

1-1/2 0.40 150
>2 0.40 200
A27 Gr 65-35 0.42 120

1-1/2 0.42 190
>2 0.42 240
A27 Gr 70-40 0.45 200

1-1/2 0.45 250
>2 0.45 300
A516Gr55 0.41 120

1-1/2 0.41 190
>2 0.43 260
A516Gr60 0.44 180

1-1/2 0.44 230
>2 0.46 320
A516Gr70 0.49 280

1-1/2 0.49 330
>2 0.51 400
A285 Gr B 0.37 60
1-1/2 0.37 60
2 0.37 120
The above values for carbon equivalent are based on the maximum
allowable carbon and maganese content allowed by the respective
ASTM standard. The actual carbon equivalent may be less, allowing
lower preheat.
Table 6.13: Preheat for typical carbon-steel turbine-materials (Sinclair &

Rodrigue, 1986 [6.47])
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

If the material thickness is less than 4 in (102 mm), a minimum preheat

should be 150° F (66° C); otherwise, should be 250° F (121° C). In
restricted areas, preheat should be increased by 100° F (56° C).
• Stellite 21 weld material
A nominal 60° F (16° C) preheat is necessary.
Surface Finish
Surface roughness is one of the key factors involved in the cavitation

development. In the USA, the following upper limit for surface finishes, Rs,
is recommended [6.47]. For H < 300 ft (91 m), Rs < 250 pin (6.3 (im); for
H = 300-1000 ft (91-305 m), Rs < 125 pin (3.2 fim); and for H > 1000 / *
(305 m), Rs < 90 fiin (2.3 ^m).
There are other precautions that should be taken in repairs. If the re
pair work involves re-contouring or re-profiling, all these changes should be
made in small steps to ensure an optimum shape being reached after several
successive repairs. Besides, the dimension monitoring and the distortion pre
vention (by using braces and strong backs) are also essential to the success
of repairs.
6.6.4 Examples
The successful repairs, using the in situ examination, model investigation

and numerical analysis, are demonstrated by following two examples.
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
Example 1:
This is a case of leading-edge cavitation damage on Francis turbines,

reported by Coulson et al, 1984 [6.7].

344 Cavitation
LG-2 Turbine characteristics

Prototype machine:
Nominal output (MW) 338.7
Rated net head (m) 137.2
Range of net head (m) 127.4-143
Synchronous speed (rev/min) 133.3
Outside diameter of runner (mm) 5940
Runner mass (t) 90
Nominal plant sigma 0.102
Model:
Test head (m) 15
Scale factor 13.38
Table 6.14: LG-2 turbine specifications (Coulson et al, 1984 [6.7])
The Francis turbines at La Grande 2 Powerhouse (Canada) experienced

severe leading-edge cavitation damage which was initially repaired with the
blade inlet modification in situ, resulting in a partial improvement. A fur
ther model study helped to reach a simple and satisfactory cavitation-free
solution.
Description of Damage
The cavitation guarantee clause for the turbines (whose specifications are
listed in Table 6.14) limits the metal removal to 45.5 kg (100 lb) during a
continuous operating period of 10,000 hrs, referring to Fig. 6.37. However,
before the beginning of the guarantee period, during the 3000 hr initial
inspection of the first commissioned machine, severe cavitation damage was
detected which was very localised and repetitive from blade to blade; the
depth of the eroded zones varied from 1 — 15 mm, even exceeding the depth
of the stainless-steel overlay. The damaged areas were centred on the back
of the blades approximately 60 cm downstream from the inlet edge (about
20% of the blade length) and about 30 cm from the blade-band junction as
shown by Fig. 6.39(a).
Initial Modifications
The initial modifications were made at site by making the inlet curvature
changes (on the lower quarter of the blade height) in two ways to reduce the
incidence (Fig. 6.38). One was made on No.2 unit by extending slightly the
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

Weight loss (kg)

50
' 8 jarai itee

/ limi t /
40
B :fore nodi icatk n
/
30 \
/
20 /
/
fterr lodifi :atiof
10 / \,
/
\
\
5000 10000
Operating hours
Figure 6.37: Quantities of metal removal by cavitation damage (Coulson et

al, 1984 [6.7])
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
(b) Unit 3
Figure 6.38: Inflow-edge modifications

346 Cavitation
Figure 6.39: Leading-edge cavitation damage and final modification (Coul-

son et al, 1984 [6.7])
(a) Leading-edge damage on the back of blade (b) Final modification
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

inflow edge and by building up the face side to minimise metal removal on the
back; the other was made on No.3 unit by increasing the curvature of inflow
edge but without extending the blades. The inspections after 3500 hrs (unit
2) and 1300 hrs (unit 3) subsequent to the modifications showed a significant
damage reduction on unit 2 and a reduction by a factor of three on unit 3.
However, these improvements were still not satisfactory.
Model Investigation and Final Solution
A further model study (scale factor: 13.38) was carried out to pin point
the damage cause. A fine and long endoscope was inserted through the spiral
case and distributor to a point near the entrance to the runner in order to
observe the critical cavitation zone with minimum flow disturbance. It was
observed that the cavitation bubbles appeared at about 1 cm downstream of
the inlet edge at the junction between the blade and the band. The change
of inlet curvature on the model showed little effect on the cavitation devel
opment; while modifications to the fillet seemed to have a more significant
influence. Thus, after an intensive modification test on the model, a final
simple and effective solution was reached which was a fillet having a pro
gressively increasing radius, combined with a build-up on the back of the
blade adjacent to the fillet and a thickening of the inlet edge near the band,
as shown in Fig. 6.39(b). The inspection after 2350 hr service since this
final modification showed that no significant metal removal had taken place
on such modified blades (approximately 98% damage reduction as shown in
Fig. 6.37) and a 0.5% efficiency gain resulted.
This example shows that the change of incidence of inlet edge is not al
ways the answer to the leading-edge cavitation damage; the radius of fillet,
which influences the formation of secondary vortices there, is also respon
sible for the leading-edge cavitation damage adjacent to the band; and the
in situ examination/analysis may not be adequate in seeking the final solu
tion, particularly for complicated cases.
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
Example 2:
This is an unusual and complex case involving tip vortex cavitation com
bined with leading-edge cavitation damages on the pressure side of blades
on a retrofitted propeller turbine, reported by Baetz et al, 1989 [6.5].
The retrofitted propeller turbine at the Kembs Power Plant (France) was
found to be damaged on the pressure side of blades after 500 hour operation

348 Cavitation
Blade inlet
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
(b) 2 (m)
Figure 6.40: Flow comparison between the old and retrofit designs (Baetz
et al 1989 [6.5])
(a) Meridian flow (b) Evolution of meridian velocity and boundary layer
along the bottom and discharge rings

Inlet edge
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
Cavitation damage
Discharge ring
Section B-B Section A-A
(a)
Wicket gate trailing edge
Bottom ring (ordinary steel)

Discharge ring (stainless steel)
(b)
Figure 6.41: Damage areas on the retrofit design (Baetz et al 1989 [6.5])
(a) Damage on the blades (b) Damage on the bottom ring

350 Cavitation
and on the bottom ring after 3000 hours. A short-term solution based on
the damage-pattern analysis, which involved the local cutting of blade and
the installation of anti-cavitation fins, was not successful. Then a solution
obtained through flow analysis and confirmed by model test, which modified
the blade inlet edge and rounded up the sharp tip-edge of blades, finally
eliminated the damage problem.
Description of Damage
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
The demand of increasing unit flow rate made the retrofitted turbine pos
sess a larger diameter, a smaller number of blades (from six reduced to five)
and wider opening of guide vanes while the following remained unchanged:
(1) The distributor position and the profile of guide vanes; (2) The draft
tube elbow and diffuser. Thus, the retrofit design of the flow passage had
a very sharp turning at the bottom ring as shown in Fig. 6.40. The conse
quent cavitation damage was found on two areas. One is the very localised
and aggressive damage on the inlet area of blade pressure side, adjacent to
the blade periphery towards the discharge ring as shown in Fig. 6.41(a).
The blades were made of stainless steel (Cr 17%, Ni 4%). The depth of
damage exceeded 5 mm. The other is the low-intensity cavitation damage
on the bottom ring, which was made of ordinary carbon steel, occurring at
a periodic azimuth as shown in Fig. 6.41(b).
Short-term Solution
Owing to the aggressive nature of damage on the blades, a short-term

approach based on the in situ analysis of damage pattern was proposed
before the final solution was reached. The analysis assumed that the damage
was due to negative incidence over 20 mm from the blade tip. Therefore,
the inlet edge towards the tip was cut back forming a inlet corner radius
of R = 50 mm to eliminate the negative incidence there. Furthermore,
based on the nature of tip periphery damage, the anti-cavitation fins were
installed on the pressure side. These initial modifications did not solve the
problem effectively. However, the fact that the damage area was narrowed
down from 110 mm (without fin) to 60 mm (with fin) gave a clue for use in
further modifications that the damage was related to tip-vortex cavitation
rather than to leading-edge cavitation.
The damage on the bottom ring was successfully repaired with a stainless-
steel overlay.

Flow Analysis
The tip cavitation damage on the pressure side near the inlet edge was
unusual and the mechanism of generating such a pressure difference at this
location, i.e. the underpressure on the pressure side and the overpressure on
the suction side, was further studied.
i = -30°
Cylindrical cross section
Figure 6.42: Influences of boundary-layer separation and entrainment on the

incidences of the retrofit and final modified blades (Baetz et al, 1989 [6.5])
Firstly, the meridian flow calculation (potential flow model) for the retrofit
design showed an increased velocity peak at the bottom ring corner, which
generated a vacuum pressure of —4 mWC responsible for the cavitation dam
age on the bottom ring, followed by a steeper deceleration, which gave rise to
the thickening and separation of the boundary layer there, referring to Fig.
6.40. The calculation of the boundary layer along the walls of bottom-ring
and discharge-ring indicated that the retrofit design increased the boundary-
layer thickness up to 40 mm at the location corresponding to the blade inlet,
where a possible separation was very likely to occur if the form factor, L,
was equal to or less than —0.178, making the leading portion of the blade
periphery well within the boundary layer, i.e. the velocity-deficient zone, or
even in the separation zone. Thus, if the separation did occur at the location
corresponding to the blade inlet, the blade cross-section even at a distance
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

352 Cavitation
(for say 20 mm) to the wall of discharge ring would have a negative incidence
of about —30° owing to the angle of relative inflow, (3, being virtually zero,
as shown by Fig. 6.42; or a negative incidence of —5° with no separation
occurring. This alteration of boundary-layer characteristics introduced by
the retrofit design was the real cause which gave rise to the inverse pressure
difference along the leading 10% chord-length of the tip periphery where the
tip-vortex cavitation occurred.
Model Study
Through the model investigation, the boundary-layer separation and the

validity of the proposed final modification to the retrofit design were con
firmed. The modification, referring to Fig. 6.42, was: (1) To cut back such
a large portion of the blade corner as shown in Fig. 6.43 that the draw back
of the leading edge from that boundary-layer separation point would leave
enough space for the re-entrainment effect, which accelerated the boundary
layer, to take place and also favourably reduce the curvature of blade itself
by about 3°, resulting in an approximate zero incidence; (2) To reduce the
strength of tip vortices by rounding up the blade tip.
Figure 6.43: Comparison of damage areas before and after blade modification
(Baetz et al, 1989 [6.5])
Prototype Result
The inspection after completing 18,000 hour service since the final mod
ification showed a significant improvement (only a little cavitation damage
on the periphery), referring to Fig. 6.43.
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

References
6.1 Akhtar, A. (1986), 'Materials Technology for Turbine Performance',
Water Power and Dam Construction, August 1986, ppl3-19.
6.2 Arai, C. (1984), 'An Acoustic Detection Method of Cloud Cavitation',

Trans. ASME, J. Fluids Eng., Vol.106, p.466.
6.3 Avellan, F. and Dupont, Ph. (1988), 'Cavitation Erosion of Hydraulic

Machines: Generation and Dynamics of Erosive Cavities', 14th IAHR
symposium on Progress within Large and High Specific Energy Units,
Trondheim, Norway.
6.4 Avellan, F., Dupont, Ph. and Farhat, M. (1991),'Cavitation Erosion

Power', 1st ASME-JSME Fluids Engineering Conference, Portland,
USA.
6.5 Baetz, P. et al, (1989), 'Increased Unit Output of Kembs Propellers

- Control of Cavitation Damage', Journal of Hydraulic Research,
Vol.27, No.6.
6.6 Chen, H. C. (1995), 'Assessment of a Reynolds Stress Closure Model

for Appendage-Hull Junction Flows', Journal of Fluids Engineering,
V.117, 557-563.
6.7 Coulson, D. M. et al, (1984), 'Solving a Severe Cavitation Problem at

La Grande 2', Water Power and Dam Construction, January, 1984,
pp20-24.
6.8 De, M. K. and Hammitt, F. G. (1982), 'New Method for Monitor

--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
ing and Correlating Cavitation Noise to Erosion Capability', Trans.

ASME, J. Fluids Eng., Vol.104, p.481.
6.9 Doolin, J. H. (1986), 'Judge Relative Cavitation Peril with Aid of

These Eight Factors', Power, Oct. 1986.
6.10 Dupont, Ph. and Avellan, F. (1991), 'Numerical Computation of a

Leading Edge Cavity', 1st ASME-JSME Fluids Engineering Confer
ence, Portland, USA.
6.11 Fujikawa, S. and Akamatsu, T. (1978), 'Experimental Investigations

of Cavitation Bubble Collapse by a Water Shock Tube', Bull. JSME,
Vol.21, No. 152 , p.223.

354 Cavitation
6.12 Gode, E. and Cu6nod, R. (1989), 'Numerical Flow Simulations in

Francis Turbines', Water Power & Dam Construction, May 1989,
ppl7-21.
6.13 Grein, H. (1990), 'Cavitation Pitting and Rain Erosion on Pelton
Runners', IAHR Symposium on Modern Technology in Hydraulic En
ergy Production, Belgrade.
6.14 Grein, H. and Schachenmann, A. (1992), 'Abrasion in Hydroelectric
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
Machinery', Sulzer Technical Review, 1/1992.
6.15 Giilich, J. F. and Rosch, A. (1989), 'Cavitation Erosion in Centrifugal
Pumps', Chemical Engineering Progress, Nov. 1989, pp68-73.
6.16 Hammitt, F. G. (1980), Cavitation and Multiphase Flow Phenomena,
McGraw-Hill, New York, 1980.
6.17 Hammitt, F. G. (1983), Cavitation and Multiphase Flow Phe
nomenon, Lecture Notes (in Chinese) presented at the China Ship
Research Centre, Wuxi, China.
6.18 Hickling, R. and Plesset, M. S. (1964), 'Collapse and Rebound of a
Spherical Bubble in Water', Phys. Fluids, Vol.7, No.l, p.7.
6.19 Hobbs, J. M., 'Experience with a 20-kc Cavitation Erosion Test',
ASTM STP, Vol.408 , p.159.
6.20 Ivany, R. D. and Hammitt, F. G. (1965), 'Cavitation Bubble Col
lapse in Viscous Compressible Liquids - Numerical Analysis', Trans.
ASME, Ser.D, Vol.87, p.977.
6.21 Iwai, Y., Okada, T. and Mori, H. (1990), 'An Experimental Inves
tigation of Cavitation Bubble Collapse Pressure and Erosion Pits in
Flowing System', Cavitation and Multiphase Flow Forum, ASME
FED Vol.98, p.127
6.22 Iwai, Y., Tanaka, S. and Okada, T. (1989), 'A Study of Cavitation
Bubble Collapse Pressure and Erosion, Part 2: Estimation of Erosion
from the Distribution of Bubble Collapse Pressure', Wear, Vol.133,
p.233.
6.23 Jin, H., Zhang, F., Li, S. and Hang, C. (1986), 'The role of Sand Par
ticles on the Rapid Destruction of the Cavitation Zone of Hydraulic
Turbines', Wear 112, ppl99-205.

6.24 Jones, I. R. and Edwards, D. H. (1960), 'An Experimental Study of

the Forces Generated by the Collapse of Transient Cavities in Water',
J. Fluid Mech., Vol.7, p.596.
6.25 Kato, H., Maeda, M. and Nakashima, Y. (1981), 'A Comparison

and Evaluation of Various Cavitation Erosion Test Methods', Proc.
ASME Symp. on Cavitation Erosion Fluid Systems, 1981-6, p.83.
6.26 Knapp, R. T. (1955), 'Recent Investigations of the Mechanics of Cav

itation and Cavitation Damage', Trans. ASME, October, p. 1045
6.27 Kuethe, A. M. and Chow, C. Y. (1986), Foundations of Aerodynam

ics, 4th Edition, John Wiley & Sons, New York.
6.28 Li, S. C. (1987), 'Cavitation Damage on Pump-turbine', Consultant

File.
6.29 Lord Rayleigh (1914), 'On the Pressure Developed in a Liquid during
the Collapse of a Spherical Cavity', Philos. Mag., Vol.34, P-94.
6.30 Medley, D. F. (1992), 'Selecting Alloys for Optimum Turbine Life

and Efficiency', ASME FED 136.
6.31 Miller, R. S. (1992), 'Corrosion in Pumps', Proceedings of 9th Int.

Pump Users Symposium, Houston, USA.
6.32 Okada, T. (1981), 'Corrosive Liquid Effects on Cavitation Erosion',

J. Ship Research, Vol.25, N0.4, p.271.
6.33 Okada, T., Iwai, Y. and Yamamoto, A. (1983), 'A Study of Cavitation
Erosion of Cast Iron', Wear, Vol.84, p.297.
6.34 Okada, T., Iwamoto, J. and Sano, K. (1977), 'Fundamental Studies

on Cavitation Erosion (Observation of Eroded Surface by Scanning
Electron Microscope)', Bull. JSME, Vol.20, No.147, p.1067.
6.35 Palgrave, R. and Cooper, P. (1986), 'Visual Studies of Cavitation in

Pumping Machinery', 3rd International Pump Symposium, Houston,
USA.
6.36 Plesset, M. S. and Chapman, R. B. (1971), 'Collapse of an Initially

Spherical Vapour Cavity in the Neighbourhood of a Solid Boundary',
J. Fluid Mech., Vol.47, No.2, p.283.
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

356 Cavitation
6.37 Pylaev, N. E. and Agelly, U. W. (1974), Cavitation in Hydraulic

Turbines (in Russian), Mashinostroenie, Moscow.
6.38 Ramamurthy, A. S. and Bhaskaran, P. (1979), 'Velocity Exponent for

Erosion and Noise due to Cavitation', Trans. ASME, J. Fluids Eng.,
Vol.101, p.69.
6.39 Rao, A. S. and Kung, D. (1987), Metallic Overlay Materials for the
Optimum Cavitation Performance of Hydraulic Turbines, Report for
the Canadian Electric Association, CEA No.135 G273.
6.40 Rao, A. S., Kung, D. and Akhtar, A. (1986), 'Effects of Stress on

Cavitation Erosion', IAHR Symposium, September 1-6, Montreal,
Canada.
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
6.41 Robinson, M. J. and Hammitt, F. G. (1967), 'Detailed Damage
Characteristics in Cavitation Venturi', Trans. ASME, J. Basic Eng.,
Vol.89, No.l, p. 161.
6.42 Sato, J. et al, (1990), 'Comparison of Silt Erosion Characteristics of

Materials through Different Types of Test Methods', 3rd China-Japan
Joint Conferences of Fluid Machinery.
6.43 Sato, J., Usami, K., Okamura, T. and Tanaba, S. (1992), 'Basic Stud
ies of Coupled Damage Caused by Silt Abrasion and Cavitation Ero
sion', ASME FED 136.
6.44 Schroder, V. (1987), 'Cavitation Erosion Studies with Construction

Material Samples Made of Cast Chromium-Nickel Steel and Cast
Polymide', Pipes & Pipelines International May-June, 1987.
6.45 Schumacher, W. (1989), 'Reduced 0 & M Costs through Better Met

allurgy', Water Power '89.
6.46 Simoneau, R. (1984), 'The Optimum Protection of Hydraulic Tur

bines against Cavitation Erosion', 12th IAHR Symposium, Sterling,
UK.
6.47 Sinclair, J. P. and Rodrigue, P. R. (1986), Cavitation Pitting Miti

gation in Hydraulic Turbines, EPRI AP-4719, Palo Alto, California,
USA.

6.48 Sui-de Hydropower Station and National Experiment Centre for

Abrasion and Cavitation Control in Hydraulic Machinery, (1987), '
Summary of Pilot Test for Cavitation and Abrasion Resistance of Hy
draulic Turbine in Suede Hydropower Station', Journal of Abrasion
and Cavitation in Hydraulic Machinery (in Chinese), Vol.1 No.5.
6.49 Tomita, Y. and Shima, A. (1986), 'Mechanisms of Impulsive Pressure

Mech., Vol.169, p.535.
6.50 Toshima, M., Okamura, T., Satoh, J., Usami, K. and Tanabe, S.
(1991), 'Basic Study of Coupled Damage Caused by Silt Abrasion
and Cavitation Erosion', J. JSME (B) V.57 No.539, pp20-25.
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
6.51 Zhao, K., Gu, C , Shen, F. and Lou, B. (1993), 'Study on Mechanism
of Combined Action of Abrasion and Cavitation Erosion on Some
Engineering Steels', Wear No.162-164, pp811-819.

--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

Chapter 7
Cavitation Caused
Vibrations
7.1 Cavitation Pressure Pulsations in Turbines

7.1.1 Blade Cavity Pulsations
J. Sato
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
In modern hydraulic turbines, cavitation is far more common in rotat

ing cascades, such as runner blades, in the lower pressure region than in
stationary cascades, such as guide vanes, in the high pressure region. The
cavitation developed from the runner blades will not reach any other rotat
ing or stationary cascades even if it breaks away from these blades. How
ever, the cavitation will cause pressure pulsations in the water and produce
acoustic noise. These pulsations may propagate to all components of the tur
bines and further to the concrete structure, even causing a vibration of the
power-house. Nowadays, there is an increasing demand for noise reduction
in hydroelectric power plants.
The radial fluctuation force caused by cavitation has been measured on
five different models of Francis pump-turbines with specific speed ranging
from 28 to 60 {rpm, m, m?/s) [7.53]. In pumping mode (Figure 7.1 (a)), as
the cavitation number, cr, decreases, the fluctuating amplitude remains con
stant until a certain cavitation number, ac, is reached, and then increases.
After it reaches a local maximum, it begins to decrease rapidly as the cavita
tion becomes more severe. In turbine mode (Figure 7.1 (b)), the amplitude
remains virtually constant for any cavitation numbers, which indicates a
359
360 Cavitation
Hioote25.3m
4
/GVO=5'>M ?0.]
q=0.33
*q=0.56
7:
■«c -GVO=lO*,q%0.24
-GVO=60%,q^0.24
0.1 0.2 0.3 0.4 0.5

cavitation number a
(a)
—i r
GVO=30%.QI=0J0U
O : n 1 =223 tpm/v r m
O : n 1 »205 rpm/v m
Unit of Qi ! n r / s / v m
,GVO=60%,Qi=0.024
iVO=30%,Qi=0.014
GVO=l 10%, Ql *0.042

GVO=60%,Ql%0.027
0.2 0.3 0.4 0.5

cavitation number a
(b)
Figure 7.1: Amplitude of radial fluctuation force versus cavitation number
(medium specific speed) (a) Pumping mode; (b) Generating mode.
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

Cavitation Caused Vibrations 361
weaker influence of the cavitation-number variation.

As described in Chapters 5 and 6, various types of cavitation arise in
turbine runners. The most pulsating ones appear at the off-design points,
where the incidence angle at the leading edge of the runner is extremely
positive or negative. A positive incidence angle usually occurs at the high
head operating points, resulting in violent vibrations [7.42, 7.51], while a
negative incidence angle usually occurs at the low head operating points. At
partial load or overload, similar extreme deviations of incidence angle occur,
arising from complicated flow patterns between the guide vanes and runner
vanes. The followings are some typical cavity pulsations.
• A typical example of such severe case which occurred at off-design

points for a medium-specific-speed Francis runner is reported [7.57].
At very low head operating points, the flow separation on the pressure
side of the leading edge (see Figure 7.2) produces vortices there, form
ing a vortex line starting from the crown and extending downwards
(i.e. towards the runner outlet). The core of this vortex is usually
Rotation
Pressure side
Suction side
Vortices by \
separation ^
- Shroud Band
High cavitation
intensity
Vortex line
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
Figure 7.2: Vortex line at low head.

362 Cavitation
Figure 7.3: Inter-blade vortices at high head.
cavitating. If the vortex touches the suction side of the adjacent blade,
erosion may occur there. The pulsation caused by this kind of cavita
tion is very strong. In some circumstances, the cavitating vortex may
extend onto the suction side down to two thirds of the blade length,
causing the most violent vibration [7.42, 7.51], and subsequently severe
erosion at the blade outlet.
• Another example is the cavitation on the suction side of Kaplan run

ners. At very high head operating points, the cavitation develops from
the leading edge. This kind of cavitation is very unstable and causes
strong fluctuations, producing an intensive erosion power immediately
down stream.
• At very low partial load or high head operations, inter-blade vortices

will appear between two adjacent blades, close to the suction side.
Because the machine cavitation number is usually relatively low at the
high head operations, this vortex is always cavitating. These interblade
vortices may not cause damage to the blades if they do not touch the
blades, but they always cause certain pulsations in the machine (Figure
7.3).
--``,```,,``````,,,,``

m=170.5ipm (high head)
ni=l73.5ipm (normal head)
m-179.2rpm (low head)
cavitation inception
(visual observation)
■ti O C O
20 40 60 80 100
Guide vane opening GVO (%)
(a)
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
(b)
Figure 7.4: Inlet cavitation of pump turbine (model, generating mode), (a)
acoustic emission; (b) visual observation.

364 Cavitation
However, such cavitation caused vibrations can be used to advantage as

a cavitation level index. It is particularly useful for monitoring prototype
turbines in situ. The presence of cavitation can be successfully detected by
measuring the acoustic emissions from the cavitation (see Figure 7.4), by
using accelerometers attached on the outer wall of the turbine.
7.1.2 Blade Wake Cavitation Pulsations

J. Sato
The trailing edge of the runner is usually shaped thinly and asymmetri
cally, in order to avoid the Karman-vortex-shedding, which might otherwise
cause severe pulsation and singing noise. This asymmetrical shape, however,
is often the cause of cavitation, which also induces pulsations. In some cases,
it may also damage the trailing edge.
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
7.1.3 Draft Tube Vortex Core Cavitation Pulsations

P. Henry
For Francis turbine runners, which have the non-adjustable blades, the
Figure 7.5: Vortex cavity at the runner outlet, helical at part load (left)
and axially centred at full load (right), (Courtesy of Mr. Tanaka, Toshiba,
Japan).

10
as
IWEf
0.8
0.7
06
M
»/Topt
lausf
10
v = 0.220
09
m m ■
:■ 2Z5mCE
0.074
08 //A fit Opt

<X2S a50 Q75 100 125 150 ITS
Figure 7.6: Draft tube cavitation patterns versus operating conditions of a

Francis turbine.
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,

366 Cavitation
•-/Emplacement du
/ capteor
<P/<PA
AJj *&.
2 f/h
Figure 7.7: Draft tube pressure oscillations and cavitation patterns versus
flow coefficient.
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

Forf chary Chorp* por»«ll«
1 v-0.640
LO i II a -Q310
i H «10mCE
1
»^ i—^
AH/H-fcl
v-0.220
a* ff »ao7*
:Li
H «22.SmCE
OJ
Figure 7.8: Part load pressure oscillations and vortex cavity at part load
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
and full load for high and low specific speed turbines.
residual swirl will cause a vortex in the draft tube when operating at off-
design flow-rates. The vortex rotates in the same direction as the runner at
part load and in the opposite direction at overload [7.8, 7.9]. Its structure
depends on both the draft tube design and the runner outlet velocity distri
bution. For very small flows, this vortex is quite disorganised. From 50 %
flow-rate up to nearly the best efficiency flow, a precession motion at 0.2 to
0.3 times the runner rotational frequency generates a spiral vortex pattern
in the whole draft tube cone height (Figure 7.5, left). The spiral vortex
may extend all the way to the draft tube bend. Beyond the best efficiency
flow, the vortex is axially centred in the draft tube cone, with a very small
spiral termination (Figure 7.5, right). When local pressures are low enough,
the vortex core cavitates and the vortex pattern may be visually observed.
The observation of draft tube cavitation yields interesting information on
the unsteady flow fields (Figure 7.6).
Draft tube cavitation will also create a compressible volume within the
flow. This is strongly dependent on the Thoma number a. This cavitation
compliance, together with the water plug inertia in the draft tube, determine
the frequency of free oscillations of the draft tube flow [7.22]. It is normally
less than twice the frequency of the runner rotation.
Pressure oscillations [7.25] naturally go with the unsteady flow fields
(Figure 7.7). Below 50 % of the best efficiency flow, only a broad band of
nearly random noise with occasional shocks is usually measured. Between
50 % and 80 %, a rotating sine pressure field associated with the precession
in the draft tube dominates the pressure signals (Figure 7.8) [7.2, 7.36]. As

368 Cavitation
the precession motion is more or less off-centre, the cross-sectional area-

averaged pressure is not constant and some excitations are radiated through
the runner and transmitted into the penstock. Strong surging may occur if
the precession frequency is close to one of the free oscillation frequencies of
the draft tube or the penstock. At around 80 % of the highest efficiency
flow, unstable flow patterns generate sharp bursts of pressure pulses above
the rotational frequency [7.23]. At full load, the draft tube free oscillations
are excited by a low-power, broad band noise (figure 7.8). This is normally
quite gentle if there is no dynamic amplification from the penstock, but
self-amplified surges may develop with some turbine designs.
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

7.2 Cavitation Induced Pulsations in Pumps

7.2.1 Introduction
R. K. Turton
In recent years, the effect of re-circulation in the suction zone of a pump
has been studied extensively. The choices open to a pump designer faced with
low NPSHA are to increase the suction size, or fit an inducer. If the suction
diameter is increased the flow rate at which NPSHR rises approaches, and
can exceed NPSHA, gets closer to the design flow rate, as already illustrated
in §5.2 (referring to Figure 5.25).
Recirculation thus has an effect on hydraulic performance and NPSHR,
and associated effects are hysteresis in the head/flow curve as sketched in
Figure 7.9, and an increase in NPSHR as flow reduces which gives a peak
in the curve, as illustrated in Figure 7.10, with NPSHR reducing as flow
rate reduces.
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
1
Rovratt a
(a)
Howrrtt. fl
00
Figure 7.9: Hysteresis in a pump characteristic due to cavitation at part

flow, (a) dip in a Q - H curve; (b) hysteresis in a Q - H curve.

370 Cavitation
I
I
in
a.
Flovrate Q
Figure 7.10: The "surge" in NPSHR at part flow discussed by Yedidiah

[7.56].
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
CAVITY
Figure 7.11: A stable central cavity caused by recirculation described by

Yedidiah [7.56].

--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
Figure 7.12: A comparison of NPSHA curves for a high and a low friction
system with the 1% decay NPSHR curve and associated surge (after Massey
[7.33]).
Yedidiah [7.56] comments that as NPSHR reduces at low flow rates low
frequency oscillations (in the region of 1 to 6 cycles per second) become
very intense, but that as NPSHR reduces further the noise also reduces.
He observed the flow in a transparent 5 in suction line using stroboscopic
illumination. The pump had a specific speed of 2000 (rpm, US gall/min, ft)
and was fitted with a constant pitch inducer. When tested with reasonable
NPSHR the machine ran quietly right down to zero flow, even though very
intense prerotation could be seen. At part flow it was observed that as
NPSHR reduced, a large single cavity appeared periodically as sketched in
Figure 7.11. This cavity collapsed vigorously, giving rise to the vibration
and noise already mentioned. The approximate dimensions of the cavity
are noted in the figure. As NPSHR reduced further, prerotation ceased,
and vapour pockets formed in the inducer. Yedidiah suggests that these
acted as cushions, reduced noise, and were also responsible for reduced head.

372 Cavitation
Experience gained by other engineers supports this view. Massey [7.33]

describes experiences with two machines with specific speeds of 26.1 and
147.5 (rpm, m3/hour, m) running at 50 Hz frequency. He also observed a
frequency range of 1-6 Hz, with variations in the net generated head of up
to 20%.
He commented that the surges observed had important effects on the
suction and discharge systems, with the instability affecting valve controls,
automatic leak-off systems and automatic level controls.
Figure 7.12 based on his analysis suggests the relationship between the
system NPSH and the NPSHR based on the 1% head drop criterion, and
allows a judgement of the risk of surge with a low friction system.
7.2.2 Characteristics of Cavitation Induced Pulsations

Y. Tsujimoto
The cavitation induced pulsations plotted on suction performance curves
of an inducer [7.26] and a centrifugal pump [7.54] are shown by Figure 7.13
and 7.14 respectively. These pulsations usually occur
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
9 Routing cavitation
0Low cycle oscillation
o.io
0-0.113
*—0.123
0.05' #-0.131
T3
t . i
0.05 0.10 0.15

Cavitation number, a
Figure 7.13: Cavitation induced pulsations in an inducer (Kamijo et al, 1977

[7.26]).

0.5 —I -J—.g»Q028,
t % • »"0 O ■ « » > « OOQ
®=0.056
/ i») in mm w=j
04 '/ 1 '. C J.'"..»='
2 03
• Heovy OSciHolO"
02 9 W « * osdllation
(>=1.5m I
I* = 206m
Mi »3000rpm
01
002 005 01 02 0.5 1.0
Cavitation coefficient K*
Figure 7.14: Cavitation induced pulsations in a centrifugal pump (Ya-

mamoto, 1980 [7.50]).
1. even near the design flow coefficients where the head decreases with
the increase of flow coefficients; and,
2. at the cavitation coefficient which is just above the head breakdown

without significant head reduction.
The above feature of item 1 is quite different from the conventional surge
and rotating stall observed in fans and compressors. This indicates that the
cause of cavitation induced pulsations is by nature different from that of the
surge and rotating stall, which is due to the positive slope of the head-flow
coefficient characteristics. This makes the problem more difficult to tackle
because we can not escape from it simply by restricting the operating points
as is often done to avoid the surge and rotating stall. The second feature
suggests that the pulsations are not caused by the head decrease due to
cavitation. Rather, the pulsations cease once head breakdown occurs, as
shown in Figure 7.13 and 7.14.
Two types of pulsations are shown in the figures. As the cavitation
number reduces, firstly there is a rotating cavitation, which is shown as
"weak oscillation" in Figure 7.14. Figure 7.15 shows the variation with time
--``,```,,``````,,,,``,`,,`,`,``-`-`,,

374 Cavitation
of cavity length on three blades of the inducer. It shows that the propagation
of the cavity is in the direction of blade number 3 -» 2 —► 1, i.e., in the
direction of impeller rotation. The propagating velocity ratio (— rotational
velocity of cavitated region / rotational velocity of impeller) was 1.16 for
Figures 7.13 and 7.15, and 1.25 for Figures 7.14 and 7.16, both rotating
faster than the impellers. This feature is quite different from that of the
rotating stall in which the separated region rotates slower than the impeller
does.
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
Time t (sec) 0.008»«c
Figure 7.15: Cavity length oscillations on three blades of the inducer (Kamijo
et al, 1980 [7.26]).
At smaller cavitation numbers, there is "low cycle oscillation" or "heavy

oscillations". Figure 7.16 shows the effect of suction pipe length on the
frequencies of the pulsations shown in Figure 7.14. It shows that the rotat
ing cavitation shown as "weak oscillation" is not affected by the suction pipe
length (l\), while the frequency of "heavy oscillation" depends on the length.
Although not shown, significant inlet flow rate fluctuation is observed for the
"heavy oscillations", but it is much smaller for the rotating cavitation. For
the cases of "low cycle oscillations" or "heavy oscillations", the cavity volume
changes in the same phase for all of the blades. From these observations,
it can be concluded that the rotating cavitation or "weak oscillations" is a
local flow oscillation just like the rotating stall, while the "heavy" or "low

:-a*;-c8'
« « 0.056
Nm ■ 3O00rpm
t t - 5.0m
■ D (i»16.6m
♦ O d »t3.6m .
M X ti »10.6m
A A It * 7.6m .
Red marks ! Heavy Osdltation
0.1 0.2 Q3 Q4
Covitotion coefficient X*
Figure 7.16: Effects of suction pipe length on the frequencies of the pulsations
in the centrifugal pump (Yamamoto, 1980 [7.54]).
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
A— 0.1 SECOND
Figure 7.17: Typical wave form of inlet pressure oscillations under heavy
cavitation surge (Sack & Nottage, 1965 [7.45]).

376 Cavitation
cycle" oscillations is the system instability similar to "surge" in the com

pressor system. Here, we term the cavitation induced system instability as
"cavitation surge". Figure 7.17 shows a typical wave form [7.45] of inlet
pressure oscillations under heavy cavitation surge. It is U shaped with its
flat bottoms corresponding to the vapour pressure.
Since the pioneering work of Acosta [7.1], efforts have been made to
identify whether or not the cavitation has a propagating nature and to clarify
its relationship with the system instability. It should be mentioned that the
examples shown here, in which the rotating cavitation was clearly identified,
are very rare cases. The propagating velocity ratio approaches 1 as we reduce
the cavitation number [7.26, 7.44]. This situation with the ratio 1 is more
often observed and causes a severe synchronous shaft vibration. For an even
number of blades, it is called the alternate blade cavitation [7.1].
7.2.3 Mechanisms of Cavitation Induced Pulsations

Y. Tsujimoto
Firstly, we consider a pumping system shown in Figure 7.18. We assume
a cavity with volume Vc + Svc at the pump inlet. The cavity volume is a
function of inlet pressure and inlet axial velocity, and thus we can express
the disturbance as follows:
JL
Vc + OVc U* + 5ua
Ui + 5ui Pi + 5p>
P.
Const
Suction pipe Cavity Pump Throttle
^R>>0
Impeller performance Throttle characteristics
Figure 7.18: Cavitation surge model.
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

Svc = -K*6Pi - M*5ux (7.1)
where K* = —dvc/dpi and is called (dimensional) "cavitation compliance"

and M* — —dvc/dv.i is "mass flow gain factor". For each component we
have the following relations,
Continuity Su2 -Sui = -Jii^c) (7.2)

rformance 5p0 - 8pi = -Rp8u2 (7.3)
Inletline Ph dt - SJH (7.4)

Outletline Sp0 = RT8V,2 (7.5)
Now, there are five relations for five unknowns. From these equations we
obtain, for example,
phK + { M } + 6ui = (7 6)
~W R7TRT~ - ^T ° -
This equation shows that if
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
M-
> WTTr <"'
the system has a negative damping and pulsations will occur. Eqn (7.7)
indicates that
• the positive mass flow gain factor is the cause of cavitation surge; and,
• the likelihood of cavitation surge increases if the length of the inlet line
reduces and the outlet resistance {Rp + RT) increases.
It is intuitively understood that a positive mass flow gain factor causes

cavity-shrinking due to an increment of inlet flow, thus promoting a further
increase of inlet flow to fill up the reduced volume. The second point above
suggests the different nature of these pulsations from the conventional surge.
A negative pump resistance (Rp < 0), which would be the cause for a
conventional surge, here will have an effect of damping cavitation surge.
Although the outlet resistance (RT > 0) promotes cavitation surge, the
inlet resistance has a damping effect. A shorter inlet line together with a
larger outlet resistance would lead to higher amplitudes of inlet disturbances
compared with the outlet ones. On the other hand, the angular frequency,

378 Cavitation
o>, at the neutral stability (i.e. M* = pl\/(Rp + RT)) can be determined

from eqn (7.6) as follows,
w = (phK*)-''2 (7.8)
which is independent of the downstream line parameters and is proportional

tol^' . Since K* correlates with U^2 as shown in eqn. (7.11), the frequency
u> of cavitation surge is proportional to the impeller speed UT, in contrast
to the constant frequency of the conventional surge.
Thus, the cavitation surge is basically a phenomenon at the inlet line.
The head decrease due to cavitation has the effect of damping cavitation
surge. Note that the cavity causes the instability through its volume change,
not through the deterioration of performance. For more details, refer to [7.45,
7.58 and 7.17].
For rotating cavitation, a two-dimensional model has been proposed
[7.47]. Using relations similar to eqns (7.1-7.5), a third order characteris
tic equation corresponding to eqn (7.6) has been obtained.
(A*-A:J)(fc*-&5)(&*-A:5)=0 (7.9)
where k* = kR + jk] is a complex reduced frequency, and k*R and k) are
the propagating velocity ratio and the damping rate of the disturbance,
respectively. It was found that k* and k% with k\R > 1 and A;^ < 0 represent
rotating cavitations propagating faster than or in the opposite direction to
the impeller rotation, respectively, k* and fcj are mainly dependent on mass
flow gain factor M and cavitation compliance K and are nearly independent
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
of the pressure performance of the inducer. The approximate criteria for

rotating cavitation is
M>2K{l + a)cot0i (7.10)
where a is the cavitation number and J5\ is the angle of relative flow mea
sured from the axial direction. Hence, rotating cavitation is also caused by
the positive mass flow gain factor. It was also shown that rotating cavita
tion is a phenomenon at the inlet. Although k* simulates experimentally
observed rotating cavitation fairly well, the existence of a backward mode
has not been found in experiments. k% with k%R < 1 represents rotating stall
which is mainly dependent on the pressure performance and relatively in
sensitive to the cavitation characteristics. Note that the rotating cavitation
and the rotating stall are represented by different characteristic solutions.
This suggests that they are completely different phenomena., i.e., "rotating
cavitation" is not a cavitation in rotating stall which is represented by k%.

OB
0.7 »v —5
— —tr
—Q/Q.
0.6
05 0.6 0-7
(a) (c)
0.8
*>
06 ~~**
>
h'^**" ^"1 ■ W - . --« 7%* 0.7
1-
05
t
-»-• —
:
i 7
r*-~-i. "-•».. *
— o/o, 0.6
04 /
0,2 0.3 0.4 0.5 0.6 0.7 5 — Q/Q,
0.5
05 0.6 0.7
(b) (d)
Figure 7.19: Cavitation in rotating stall (Murai & Takeuchi, 1969 [7.38]), (a)
sketch of various cavitations; (b) propagation velocity ratio for C\ cavitation;
(c) Ci cavitation; (d) C3 cavitation.
Cavitation in rotating stall has been observed experimentally. Figure

7.19 illustrates the types of cavitations observed in an axial flow pump at a
low flow rate below the peak of head characteristics, under light cavitating
conditions without head decrease. C\ ~ C3 cavitations possess a propagating
nature and the propagation velocity ratios are shown in the figure. They are
significantly different from each other subject to the cavitation type, suction
head Hs, and flow ratio Q/QN- However, all of them rotate more slowly
than the impeller, which is a characteristic of rotating stall.
The relationship between flow instabilities in turbomachinery systems is

summarised in Figure 7.20.
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

380 Cavitation
Cause and onset flow range Local Instability | System instability
Rotating
Surge
Flow. O stall
Positive stooe of
pressure performance.
3P/3Q)0
onset
>J3 Rotating Cavitation

to o Design
Ftow.O cavitation surge
Positive mess flow
gah factor.
M—3V/aO>0
Figure 7.20: Relation between flow instabilities in turbomachinery (Tsuji-

moto et al, 1993 [7.47]).
7.2.4 Cavitation Characteristics - Mass Flow Gain Factor

and Cavitation Compliance
Y. Tsujimoto
In this section, we discuss how the massflowgain factor can become pos
itive. Non-dimensional mass flow gain factor M and cavitation compliance
K are defined as follows.
UT dVc K = pU$ dVc
M = - (7.11)
AiHdC^ 2AiH dpi
where UT is the tip velocity at inlet, Ai is the suction area, H is the blade
spacing, Vc is the total volume of cavity, Cz\, is the axial velocity at inlet
and pi is the inlet pressure.
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
Firstly, we consider blade cavities. It is generally accepted that K > 0

since the cavity is smaller with higher inlet pressure. The velocity triangle
at the impeller inlet shows that the incidence angle is smaller for a larger
axial velocity. Therefore, the cavity volume is smaller for a larger axial

cr
o
2
<
o
u
u.
1
u
z
o
u
a
1
a.
§
K
3
%JDI 0.02 0.03 0.04 0.05 0.06 0.07
TIP CAVITATION NUMBER, <rT
Figure 7.21: Comparisons of mass flow gain factor MB and cavitation com
pliance KB calculated from quasi-steady blade surface cavitation analysis
with experimental values (Brennen & Acosta, 1976 [7.5]).
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`-

382 Cavitation
velocity, and thus the blade cavity has a positive M. This explanation is
also applicable to the tip cavities. Figure 7.21 shows a comparison of the
results of a quasi-steady blade cavity calculation with experimental values
[7.5]. Although general tendencies are in good agreement, the experimental
values are larger. This might be attributed to tip vortex, backfiow and
bubble cavities in experiments. It is noticeable that K increases faster than
M when the cavitation number reduces.
The most complete experimental data on unsteady cavitation character
istics of inducers are given in [7.7]. They were simulated by a bubbly flow
model [7.6], which assumed that the void fraction of the produced bubbles
was proportional to the incidence angle.
For centrifugal pumps, cavitation surge at part load is of special impor
tance due to its severity. In such cases, a significant reverse flow is often
observed as described in §5.2 (Figure 5.24). Due to the blockage effect of
the reverse flow near the pipe wall, the axial velocity at the centre might
increase when the total flow rate decreases. From Bernoulli's equation, the
pressure there will decrease and the inlet cavity volume will increase; and,
hence the mass flow gain factor, M, will become positive. However, under
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
oscillating conditions, the development of reverse flow occurs, with a certain

time delay, after the reduction of flow rate. Therefore, to allow the above
effect to take place, i.e. for the mass flow gain factor to become positive, this
time delay should be small compared with the period of the possible system
oscillation which depends on the inlet pipe length and the cavitation com
pliance. Therefore, the mass flow gain factor M is a function of frequency;
and, a linear treatment assuming M to be a constant is inappropriate in this
case. This mechanism is proposed and examined in detail in [7.55].

7.3 Influence of Operating Conditions

P. Henry
7.3.1 Preliminary Discussion
The operating conditions of a Francis turbine can be defined by the flow and
energy coefficients, <p and ip, and the Thoma number a.
Q , 2E NPSE
= rp = a=
* ^jy Rip'' ~E—
This section will discuss the influences of these parameters on various
dynamic phenomena associated with turbine operation [7.25].
7.3.2 Part Load Precession

The rotating flow in the draft tube of a Francis or propeller turbine in off-
design operation may organise itself into a precession motion [7.13]. Exper
imental investigations with vortex-generating cascades [7.39, 7.40] showed
that this instability indeed occurred beyond a critical value of the ratio of
moment of momentum to through-flow momentum (that is, below a critical
value of <f>). The search for similitude rules, based on Strouhal and Euler
criteria, led to the non-dimensional frequency and amplitude parameters, / *
and Ap*. Translating to model test operation parameters, these lead to:
fD3 / 1 8 ITR3U> f 4 , !
/• =
Q n 2TT i
Ap* = D 4 Ap Ap E De Ap w2 2E 64 TT2/16U;2
2 2
pQ ~ pE D Q2~ pE 8 R2w2 2
7T W 2
Q2
PE 7T2 V V
This means that if runner flows are in similitude, the precession frequency
is proportional to the runner rotational frequency, and the associated pres
sure oscillation amplitude is proportional to the test head. The resulting
non-dimensional values are appropriate for the description of all geometri
cally similar turbines operating in analogous conditions.
The draft tube cone angle and the distance between runner and bend are
statistically dependent of the specific speed v. The precession movement,
strongly influenced by these two parameters, must then also be influenced
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

384 Cavitation
by v. Investigating the oscillations occurring at 60 % of the best efficiency

flow on a number of turbine models ranging from v — 0.216 to v = 0.640
does not lead to clear conclusions about the frequency. It varies between
0.22 and 0.31 times the runner rotational frequency, with a mean value close
to the f/n = 0.28 suggested by Rheingans. Note that some authors report
on higher precession frequencies for low specific speed machines [7.12].
Figure 7.22: Influence of specific speed v on the frequency and amplitude of

the part load pressure oscillation, without resonance (statistics from IMHEF
tests).
The oscillation amplitude, off-resonance, grows linearly with v (Figure

7.22): 2 to 3 % RMS at the downstream side of the draft tube cone seem
quite normal for a high specific speed turbine, but 1 % already looks too
high for a low specific speed one. Close to resonance, amplitudes may be
expected to reach twice the normal level, still at the downstream side of the
draft tube cone. This proportion is of course purely approximative. The
precession motion is of course most influenced by the runner outlet velocity
distribution. This accounts for the large scatter in tendencies versus v.
The relative precession frequency f/n varies slightly with <j>. The vortex
machines theory gives as a first guess a frequency going down from f/n — 1
at zero flow to f/n = 0 at the zero exit swirl flow [7.43]. But the tests
with a turbine runner show nothing about this. The precession oscillation
is organised only between 50% and 90% of the best efficiency flow. In this
range, f/n stays fairly constant. Slight ups and downs appear subject to
the runner design. The energy coefficient rp and the Thoma number a influ-
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

AH/H C4n» awn tup.

m (v.. KMSJ f/n 0.»
as
/ 0.4
I AP ^ > « —
020 5M SO 5SO
a2
TRP«
*=r
♦/*.
4H/H
( * IMS]
v = 0.640
Oi
..*-*
02. , . 1—Z^L_^=*ei— *
T " . . *"«*
O A D pressurefluctuationsassociated with precession
• ▲ ■ pressurefluctuationsassociated with the 2nd harminic of precession
jy A ff pressure fluctuations associated with free oscillations
O relative frequencies
A relative amplitudes
□ phase shift (referring to cone downstream)
Figure 7.23: Influences of </>, ip and a on part load oscillations.

--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

386 Cavitation
ence the cavity volume, and so do the local velocity fields and the relative
precession frequency.
Figure 7.23 gives an example of the representation of an analysis of pres
sure fluctuations associated with the operation of a Francis turbine. Circles
stand for frequencies, triangles for amplitudes and squares for phase shifts.
Precession phenomena are plotted as solid or (in the upper left diagram for
instance) dashed lines. Free oscillations of the water column in the draft
tube are represented by dotted lines or by frequencies without indication of
amplitude. The upper left diagram shows for instance that the precession
oscillation is active from 0.5 to 0.9 <f>opt around 0.3 n with a maximum am
plitude near 0.7 <j>opt- I n this range of flows (where the vortex cavity volume
is greatest), the precession frequency is higher than elsewhere. The free
oscillations frequency goes up until <popti t n e n gets lost and appears again
beyond 1.2 âpt a s t w o distinct bands. The lower left diagram shows that the
relative precession and its second harmonic frequency (black circles) don't
change much within the explored range of ip. The free oscillations frequency
seems to be going up beyond tpopt, but this is not quite clear. The precession
amplitude goes down as ip increases. The right hand diagram gives an exam
ple of water column resonance in the draft tube, excited by the precession:
at o = 0.20, the precession frequency (white circles and solid line) is equal
to the free oscillations frequency (dotted line). The amplification of torque
oscillations is remarkable.
The unsteady pressure field resulting from the precession motion in the
draft tube cone is not purely rotating. Due to non-symmetric features of
the channel (mainly the bend), the rotation is off-centred. This means that
the surface integral of pressure forces acting on a draft tube cross-section
is not zero. Globally speaking, we can split the observed oscillation into a
rotating pressure field and a pulsating (or synchronous) one. This separation
may be performed in the time or in the frequency domain. It qualitatively
clarifies the influences of the bend (variations of cross-sectional area, depth
and piers) on the distortion of the precession motion.
In the frequency domain (Figure 7.24), pressure oscillation vectors from
various locations around a draft tube cone section are assembled [7.36]. Ide
ally, each vector should point to the position of the corresponding sensor.
The circle radius gives the rotating pressure field amplitude and the distance
from the origin to the centre of the circle gives the pulsating pressure. Only
three pressure signals are theoretically required to split the two components
of the part load oscillation. Actually, four signals give a better definition.
In the time domain (Figure 7.25), examining signals delivered by only
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

rut • l.lllll «■/> • 1%

Mir - •.»!•
t.«IS-»l Ilf • 1.114
Ilf - a.1(1 «1« • ».»11 llf - 0.311
Figure 7.24: Vectorial assembly of pressure oscillations and splitting of the

part load fluctuation into rotating and pulsating fields.
"to
•20- '«
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
o
-20-
WY wfwfw
w w -V
- 0 -
-20-
AAA
0 1 2 i" 3 0 1 2 t/i 3 B 1 2 t(t 3 <D 1 2 »T 3
P<0
1-20-1 -20-
0
.'TV/fa-rA - 0 -
il A' \\l
W il
>
11/ i
-20 -20-
Figure 7.25: Time domain assembly of pressure signals and splitting of the
part load fluctuation into rotating and synchronous fields: time-averaged
pressure oscillations on four sensors 90° apart on the draft tube cone (Pi),
rotating (Pro) and pulsating {P3y) components extracted from signal pairs
(Nishi, 1992 [7.41]).

388 Cavitation
two sensors 90° apart in a draft tube cone section allows a separation of
rotating and pulsating pressure fields [7.41]. The quad phase shift helps in
eliminating the rotating field. The use of this method is however restricted,
as it does not allow the analysis of signals with a beat. Actually, even strong
harmonies may corrupt the results.
7.3.3 80% Load Oscillations
The 80% load oscillation [7.23] is plotted in bold lines on Figure 7.26. It
Figure 7.26: 80 % load oscillation at the draft tube wall.
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

consists of a group of peaks above the runner rotational frequency. Higher

amplitudes are met when the helical vortex cavity extends down to the lowest
in the draft tube. This phenomenon may be connected with the precession:
signals from the downstream and upstream sides of cone are in phase for the
first peak, in opposite phase for the second and more or less in phase for the
third. The greatest amplitude goes with the first or second peak.
Experience showed that these oscillation frequencies increase strongly
with <j> and tp. They are not harmonics of the precession. They also strongly
increase with a, thereby showing the hydraulic nature of the phenomenon.
Their global evolution is similar to that of the draft tube free oscillations
between 75% and 85% of the best efficiency flow. Relative frequencies are
little influenced by large variations of the test head. Highest amplitudes are
observed around 0.8tpopt- Air admission completely eliminates these oscilla
tions.
7.3.4 Draft Tube Free Oscillations

Resonance of the Part Load Fluctuation
Experiments with vortex machines showed the influence of cavitation param

eter K on the precession motion frequency and on the associated pressure
oscillation amplitude [7.39, 7.40]. For a given moment of momentum number
m, the frequency and amplitude are not affected by the variation of K as
long as cavitation is non-existent or not very strong. At the critical value of
K, the frequency drops a little bit and the amplitude increases. If K is fur
ther reduced, the amplitude goes down again. The frequency decreases then
ceases to be perceptible. If the rotating pressure field largely dominates the
high K fluctuations, the oscillations at critical K are nearly in phase within
a draft tube cone cross-section. This organisation of the unsteady pressures
(surge) is visualised by the volume variations of the vortex cavity.
The in-phase, high-amplitude oscillations observed on vortex machines
at critical conditions were due to the frequency coincidence of the preces
sion and the free oscillations. Similar tests on turbine models with actual
runners (by changing Thoma number a but remaining constant <f> and tp)
showed similar evolutions of amplitudes, but systematic risings of precession
frequency at resonance and at the lower a values.
Figure 7.27 (right) shows the evolution of part load oscillations on a
v — 0.515 Francis turbine model versus a. The pressure fluctuation ampli
tude associated with the precession drops sharply as it goes off resonance
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

390 Cavitation
i—n i -i— r-T-n-

St~K
• -
• -
9 J_ j i.l,. ,.i , . _ . ! I.LJ—
a
* * «*io * * 'W
— •r-i i i 1 ?—TTI
12 -
•{ O t m m t n f t r«W *Met =0.730

f/n " ^
11 A IMal tortiM
« i V
• 1/n V AH/H
( •-
<I
f « a tHJH [V.ftMS]
0.5+ • •
4 .
*"4k** L ^
./
• 4^
V
. 1 1 • 1 1 ... 1 • 1 1 at.
1 « «l l f J 4 « « M . 020 0.2S 0.30
O A pressure fluctuations associated with precession

• pressure fluctuations associated with the 2nd harminic of precession
xy ji pressure fluctuations associated with free oscillations
O relative frequencies
Figure 7.27: Resonance at part load in a vortex machine [7.40] and in a

turbine model.
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,

(resonance is at the extreme left of the diagram). The free oscillations am
plitude is also plotted (A'), which clearly indicates the amplification when
the free oscillations frequency coincides with the precession second harmonic
frequency (black circles) atCT= 0.273, f/n — 0.46. The rising of precession
frequency at resonance is more evident for its second harmonic.
Since the tracking of the free oscillation frequency outside the resonance
conditions has not been performed before, the aim of the research of tur
bine manufacturers has previously been only towards the reduction of the
synchronous excitation related to the precession rather than the understand
ing of free oscillation frequencies. That is why nearly all published works
mention only the precession frequency.
There is no sudden modification of draft tube dynamics at the critical a.
The free oscillations frequency exists in the whole operating range [7.22]. It
dominates the machine dynamic behaviour at the part load resonance, but
it is also excited in other operating conditions by the wide band hydraulic
noise generated by the turbine flow.
With a free trigger on the data acquisition and a Hanning weighting
function for instance, the frequency resolution of signals is enhanced without
giving excessive focus on the dominant component. Averaging on a set of
time-sequenced FFTs and then highlighting all relevant frequencies in the
signal, free oscillations stand out with a phase shift close to zero, uniform
amplitudes and a high coherence among signals from the periphery of a
section of the draft tube cone. They can be amplified by various excitations:
the precession and its harmonics, the rotational frequency etc. The free
oscillation results from the joint actions of the water plug inertia and the
vortex cavity compliance. Observation of cavity volumes helps to track the
free oscillation frequency [7.21].
The part load operating range is studied with constant ip and a. For
each test point, the precession frequency and its second harmonic (likely to
excite a resonance) are collected along with the free oscillation frequency.
The amplitudes and phase shifts for theses frequencies are also recorded.
These data represent the typical information that should be retained from
the model test. The highest amplitudes associated with the precession are
generally found in the draft tube cone, on the inside of the bend (downstream
side). The lowest amplitudes are found on the outside of the bend (upstream
side). These two kinds of signals provide the best sensitivity for tracking
resonance through the phase shift. That is why the data related to the
downstream cone and upstream cone sensors are plotted in the synoptic
diagram (Figure 7.28). Pressure oscillations at the spiral case inlet result
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

392 Cavitation
tt <f>/<P,
O A □ pressure fluctuations associated with precession

• ▲ pressure fluctuations associated with the 2nd harminic of precession
XT /£ pressure fluctuations associated with free oscillations
0 relative frequencies
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
□ phase shift (referring to cone downstream)

1 cone downstream
3 cone upsream
Figure 7.28: Synoptic diagrams of distinctive frequencies and amplitudes.

from the pulsating component of the precession pressure field. They only
have a qualitative value, as their amplitude is highly dependent of the feed
pipe dynamic behaviour.
The synoptic diagram reads in the same way as the previous figures. In
the upper frame (1), we mainly see the frequency (o) and amplitude (A),
sensed by the downstream cone sensor, of the pressure oscillation associated
with the precession, as well as with its second harmonic (same symbols but
in black). The frequency of free oscillations of the water column in the
draft tube is plotted using the symbols (o) and dotted lines. Corresponding
ampUtudes are also plotted here, although this is not always the case. In
the lower frame (3), the amplitudes associated with the precession and its
second harmonic and the free oscillations sensed by the upstream cone sensor
are presented. The phase shift ( □ ) of the precession between the upstream
and down-strewn cones is also shown.
Free Oscillations at Part Load and Vortex Cavitation Compliance

Let's pay some special attention to the dynamics of a Francis turbine draft
tube. It actually can be viewed as such a tube that it is closed at the tapered
end (inlet) by the runner and opened with a zero impedance limit at the out
let, forming a water plug (referring to Figure 7.29). A gas volume occupies
*-'••'/
, / \ /
j \ /
Vj *-»-*- /
«.'» 1.1 ••« ~>/_ u»
Vft
Figure 7.29: Francis turbine draft tube (schematic) and cavitation compli
ance.
part of the inlet region. The following simplifications are also assumed:
• Due to the fact that the runner admittance is very small compared to
the gas volume, the tube can be approximated as closed at its tapered
end;
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

394 Cavitation
• The vortex cavity is lumped at the runner outlet;
• For the lumped parameter analysis (it is acceptable if the draft tube
is relatively short), the water plug compliance is lumped at the draft
tube inlet and incorporated in the vortex cavity's;
• There are no mass transfer between the vortex cavity and the adjacent
water. The total compliance of both the cavitation and the water plug
remains constant for a given operating condition.
The frequency of free oscillations of the water plug in the draft tube is then
1
/o =
with
SVT
-L? * - & ♦ * - / , £ dNPSE
If the net positive suction energy, NPSE, represents the energy around the
vortex cavity, the compliance associated with the volume VT of the cavity is
—dVr/dNPSE. If 7 is the polytropic coefficient for small variations of the
cavity volume, then
dVT _ VT
~ dNPSE ~ jNPSE
If all the vapour evaporated from the water is assumed to be within the
vortex cavity, the wave propagation speed, S, used for the evaluation of Cc
is then rather high: around 800 m/s for a real turbine, 200 m/s for a model
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
turbine with elastic walls. The draft tube of a typical laboratory model with
a <f> 400 mm runner would then have a water plug compliance equivalent to
0.19 x 10 _ 3 m 3 at NPSE = A.hJ/kg. This additional compliance is quite
insignificant if compared with a fully developed vortex cavity. That's why
the following is often used:
VT V1/2
J0
2n \JL A VPSE)
JNPSEJ
Now, let us see how the similitude of free oscillations comes out.
A/o — yXR AflA7 \a \E j = —— if ACT = A7 — 1

If the similitude of flow is achieved, the rotational frequency is linked to the

specific hydraulic energy and turbine dimensions by the same law,
\ _ i _. \ _ v ^ g
A^ — 1 ->• A n — —
Aft
That's why the frequency of free oscillations of a non-compliant draft tube
is in similitude with the runner rotational frequency, i.e.
Ay = An.
A more accurate approximation may be achieved by transposing not the

frequency of free oscillations but the cavitation compliance itself [7.21, 7.22,
7.24]. This is evaluated from /o after the estimation of the water plug com
pliance CL- The non-dimensional cavitation compliance is
C* = C-E/R£; A6. = 1.
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
If the draft tube extends with a tail-race tunnel, the lengthwise distribu
tion of its compliance may not be overlooked. C* is still transposable, but
impedance calculations must be performed in order to evaluate the frequency
of free oscillations.
Full Load Pulsations

The vortex machine theory predicts an instability of the rotating flow what
ever the direction of the runner-exit-swirl is. If, however, we simplify the
moment of momentum number as the ratio, Cu/Ca, of the peripheral to
axial velocities, we can easily understand that the precession limit will be
much farther from zero swirl flow for full load than for part load. Meanwhile,
the operational limits imposed by the runner cavitation and the generating
efficiency breakdown prevent a Francis turbine from running beyond 130%
of the best efficiency flow. That's why the rotating pressure field feature of
the precession is never significant at full load.
Beyond the best efficiency flow, the meridional speed is greater in the
central region of the draft tube flow, with a sharp drop close to the axis, un
der the runner cone [7.27], referring to Figure 7.30. The rotation is organised
in an axially centred free vortex. The static pressure close to the machine
axis is low and an axially centred cavity forms. In low (and medium, to
a lesser extent) specific speed turbines, the axially symmetric cavity is fol
lowed by a helical tail, indicating a rotation instability, although no rotating

396 Cavitation
. .- CJ^-r t.OSS
VÔJO f-t.9
—.- 0s-r 0.0it
Figure 7.30: Velocity distribution at the runner outlet at full load, with and
without vortex cavity (Kazan, 1962 [7.27]).
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

pressure field is ever observed on the draft tube cone wall. Unlike the part
load cavity, the full-load one is often larger when the energy coefficient xp is
high with constant gate opening and a.
s. .
PSl'PStr.f. 1.8093
H*»«, «CE « 9.9902
Slgaa « ■.2782
2. 3. fM
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
Figure 7.31: Waterfall diagram: free oscillations at the draft tube wall at
full load.
The presence (at full load) of a vortex cavity with its volume comparable
with the one at part load means that free oscillations must exist for these
high flows; and, have the same frequency band as for those fractional flows.
The oscillation is excited by the wide band hydraulic noise, which may be
come quite intensified beyond the best efficiency flow. It is also excited by the
instabilities developed near the cavity such as in the region of strong merid
ional velocity gradients. Note that such instabilities would follow Strouhal's
similitude, so that their frequency would fall in similitude with the runner
rotational frequency and the draft-tube free-oscillation frequency.
As a matter of fact, we can see one or more ridges of full load pulsations
on the waterfall assembly of amplitude spectra of pressure oscillations from
the draft tube cone wall (Figure 7.31). These oscillations, read from different
sensors on a given cone section, feature a high coherence, equal amplitudes
and no phase shift. With an external exciter, we may show that the am-

398 Cavitation
plification is strong for all these oscillation frequencies [7.21]; they are free
oscillations. These external excitations may only have very low amplitudes,
interacting merely through the draft-tube response, if in the resonance mode.
7.3.5 Auto-oscillations
Preliminary Discussion, Non-linear Representation of t h e Vortex

Cavity
Some operating conditions of Francis turbines feature free oscillations with

an apparently spontaneous amplitude growth [7.8, 7.14, 7.50]. Violent bursts
and shocks may also occur without apparent excitations. These particular
instabilities are called auto-oscillations. The explanation about the inability
by the impedance approach has been critisized because it does not include
the non-linear mechanism of this phenomenon. A non-linear representation
of the vortex cavity compliance indeed modifies the system damping, and
some parameter combinations make the draft tube free oscillations unstable.
Pressure fluctuations at the part load resonance, particularly for the low
amplitudes, and at the full load pulsation are visualised by a periodic varia
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
tion of the vortex cavity volume. We may reasonably wonder whether or not
it would be appropriate to introduce a dynamic variation of the compliance
into the draft-tube-behaviour model. The global cavitation compliance Ca
would then be the result of a first-order development around the steady-state
compliance C3,
with
dC. 1; 8CS
= 1/p
~di ~ -dE
Here, the time variables are the pressure and the flow at the runner outlet
(the steady-state pressure being zero), pa and qa; and, the flow-rate in the
incompressible water plug in the draft tube. The introduction of these partial
derivatives seems reasonable, as p„ and qa determine the flow around the
vortex cavity for a given global moment of momentum. Although this is a
very simplified representation of the actual phenomenon, the resultant state
equation does reveal some interesting points.

Non-linear State Equation
The system consists of a Francis turbine fed by a single penstock, referring

to Figure 7.32. The analysis is performed with lumped parameters. The
Figure 7.32: System dynamic parameters.
dynamic constants in the system are the variable cavitation compliance Ca

as discussed before, the penstock inertance Lc and the non-linear damping
in the turbine runner treated as a simple head loss. The state variables are:
the flow-rate qc (significant for the inertia) and the pressure pc (significant
for the compliance) in the penstock; and, the flow-rate qa (significant for
the inertia) and the pressure pa (significant for the compliance) in the draft
tube.
To start with, the runner is assumed to have a zero admittance: only
draft tube dynamics are considered and the flow-rate through the runner is
constant,
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
U—^
with
- / I dCs 1 dCs, ~\
Then, the inertance and compliance in the penstock and the damping due
to the runner are introduced,
Qa — j
Pa
I ~ Ca

400 Cavitation
with
I Pa > -pNPSE
This state equation can be solved numerically by Runge-Kutta approach:
X = [q;p] and X{t) = f(X(t) + U(t));
X(t + At) = X(t) + l/6(o + 26 + 2c + d)

with
' a= At-f(X(t),U(t))
b = At- f(X(t) + a/2, U{t + At/2))
' c = At- f(X(t) + 6/2, U(t + At/2))
> d = At ■ f(X{t) + c, U(t + At))
System Stability
We can show that the system with a zero admittance at the runner is stable
if dC/dQ and dC/dE have opposite signs and unstable if they have the
same sign. After applying an unsustained disturbance, the free oscillation
occurs with a positive damping for a stable situation, and with a negative
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
damping for an unstable parameter combination.

With a dynamic involvement of the penstock, free oscillations are rapidly
attenuated. If, however, the penstock dynamic parameters are such chosen
that the free oscillation frequencies of penstock and feed pipe are close, free
oscillations with negative damping may still occur.
Generally speaking, at Â, dC/dQ is positive from very small flows up
to 60% or 75% of the best efficiency flow. Then, dC/dQ is negative until
the vortex-free region. At full load, dC/dQ is positive again after passing
through a partial load point where dC/dQ = 0. This tendency remains
for neighbouring ip values except the shifts of the part-load flow at which
dC/dQ = 0. However, for further lower rp values (around 60% of ip\), the
evolution of dC/dQ becomes quite different.
The variation of dC/dE is rather subject to the machine design. At part
load, dC/dE is close to zero for xp\. It is always negative for lower ip values.
At full load, dC/dE is nearly always positive for i])\ and above; and, may
be positive or negative for the values below Â- The above discussion is
schematically shown in Figure 7.33.

Figure 7.33: Schematic representation of the cavitation compliance in the

Energy-Flow diagram: the compliance is greater if the area is more shaded.
(A) stands for best efficiency. Arrows point to possible auto-oscillations.
Bursts of auto-oscillations are sometimes reported for fractional flows,

below 50 % of QA, or for full load [7.8, 7.12, 7.14]. In these two operation
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
ranges, dC/dQ and dC/dE may have opposite signs. These reported cases
of auto-oscillations could be the type we analysed here. Although the auto-
oscillations at partial flows are annoying, they are not too bad since the
turbine normally will not operate in such condition over a long period. The
full-load auto-oscillations, however, are a major nuisance and may cause
catastrophic consequences. The unavailability of the turbine in its maximum
output region is completely unacceptable to the user.
7.3.6 Influence of the Test Head

The model-turbine operating conditions in similitude with those of the pro
totype machine may be achieved within the whole range of test head by
adjusting the rotational speed accordingly. Limits will be imposed by the
feed pump and the brake power towards the high test heads, or by the sta
bility of operation and the sensitivity of instruments towards the low test
heads.
The test head is chosen favourably for the setting of operating conditions
and the measurement of fluctuating quantities. The test head for Reynolds
similitude is beyond reach, and the Froude similitude is difficult to achieve in
most times. This means that the similitude of cavitation conditions will be

402 Cavitation
v = 0.515
Cdneaval
Figure 7.34: Influence of test head on part load fluctuations (cone down
stream).
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
i4. E,J/kg
* v = 0.516 a = 0.280
Cdneaval
Figure 7.35: Influence of test head on full-load pulsations (cone downstream).

truly met only at the reference elevation for which a is calculated. Therefore,
it is essential that the selection of this elevation must minimise the marginal
influence of the test head on the observation of dynamic behaviour.
Many authors proposed a reference elevation that is one runner diameter
below the runner, or at the mid-depth of the draft tube, without giving a
full experimental evidence to back their choice [7.39]. Actually, the vapour
formations in the draft tube are visually independent of the test head if the
reference elevation is set at the runner outlet [7.21, 7.24]. This observation is
confirmed by the fact that the relative frequencies of free oscillations don't
really vary with the test head if such reference is used (Figures 7.34 and
7.35).
Relative frequency and amplitude of oscillations associated with the part
load precession are basically independent of the test head. Actually, subject
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
to the dynamic characteristics of the test circuit, the relative amplitude

of pressure fluctuations may vary with the test head. These distortions,
however, may be eliminated by performing an acoustic power analysis on
the pressure signals in the feed pipe.
The amplitude of full load pulsations depends precisely and strongly on
the feed pipe dynamics. Considering the circuit configuration, a variation
of the test head may strongly alter the pressure fluctuations level, and even
the evolution of amplitude versus test parameters (flow coefficient, a). In
this case, the acoustic power analysis is not very helpful because the power
emissions associated with the disturbances at full load are very feeble or even
not existent.
7.3.7 Thoma Number

Thoma number, a, mainly influences the vortex cavity volume in the draft
tube for a given operating condition: the lower the a , the larger the vol
ume. As the frequency of free oscillations is determined by the draft tube
compliance, it will be thus affected by a.
Experience shows that, at part load as well as at full load, the frequency
of free oscillations varies linearly with a (Figures 7.27 and 7.36).
At part load, it is shown that the free oscillations occur at the precession
frequencies for a critical a value. This leads to the resonance of the water
column in the draft tube. The main feature of this resonance is the zero phase
shift of pressure oscillations in a draft tube section. This does not mean that
there will be higher amplitudes at a given location of the draft tube wall: the
pulsating field of the response has to be opposed to the precession rotating

404 Cavitation
Cone aval f/Yr^-UO

2j.H=15m CE -1
Y/H»ref=1.300 AH/H
o f/n [•/.RMS]
A AH/H
f/n
1--
• — **— r
<TTe
O20 0.25 0.30
d K oscillations libres dc la colonnc d'eau dans l'aspirateur
free oscillations at the cone downstream
Figure 7.36: Influence of a at full load (free oscillations at the cone down
stream).
field in a portion of the circumference (Figure 7.24), where the resonance

may appear as a drop in the observable amplitudes. Most of the time, the
part load resonance will bring a strong acoustic power emission to the feed
pipe.
The vortex cavity volume marginally modifies the velocity fields in the
draft tube cone and the precession frequency. This effect generates a jump
between the two constant values of the rotating field frequency at part load.
For very low a values, at full load, the cavitation on the runner blades
develops into such a severity that there is almost a water column separation.
For such a full load pulsation, a global low frequency surge with high ampli
tudes will dominate the system dynamics [7.19]. However, these operating
conditions rarely exist for the prototype turbines.
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,

7.4 Cavitation Resonance in Hydraulic Machinery

Installations
P. Henry
7.4.1 Introduction
The resonance induced by cavitation in a flow system is not a well understood
subject. Cavitation resonance was initially named from the investigation into
the low-frequency fluctuations associated with cavitation in cavitating ven-
turi flows [7.29, 7.30 and 7.31]. It suggests that the coincidence of the cavity
natural-frequency (which varies with cavitation development) with the flow-
circuit natural-frequencies leads to the cavitation resonance in the system 1 .
In a model Francis turbine study [7.32], a similar fact that the cavity in the
draft-tube, in some operating conditions, can behave as an oscillator itself
rather than a lumped capacitor which possesses only a compliance to be
combined with the water portion (of the cone) then forming an oscillator
(ie. the water plug as described before). This indicates that the mode in
which cavitation resonance can occur varies subject to various factors; and,
the lumped compliance model is only a first order approximation.
Due to such complexity and uncertainty of this phenomenon, its predic
tion is not a easy task and not always successful. Here, such a practical
case of cavitation resonance, based on the water plug model, is presented to
demonstrate the phenomenon and some combating measures.
This case study deals with the turbines (u = 0.516) of the Saucelle-
Huebra plant (Figure 7.37) in the Salamanca province of Spain [7.24, 7.25].
The control of stability of operation was performed during the model accep
tance tests in 1986, according to the concepts then. The interpretation of the
model test results failed to predict the instability occurred when the full-size
turbines operated at full load. This problem has motivated an additional
series of laboratory tests.
7.4.2 Prototype Installation

Configuration
The Saucelle-Huebra power plant is set on the rio Duero, on the border
between Spain and Portugal. It reinforces the 240 MW Saucelle power-
'Also referring to footnotes 12, 13 and 14 in 'Bubble-Bubble (Large-Scale) Interaction'
of §3.7.3.
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

406 Cavitation
Figure 7.37: Saucelle-Huebra installation and its operating range and full-
load instability.
house and will also, in a second phase, handle the lower portion of the rio
Huebra. Starting from the existing dam on the Duero, a 1330 m long tunnel
takes the water to the surge tank. A branching penstock with gate shafts
then feeds the two Francis turbines in the underground power-house. The
5.23 m diameter runners have a 137 MW rated power under the rated head,
62 m (608 J/kg). The Thoma number under 62 m is a = 0.28, the reference
elevation is set at the runner outlet. The draft tubes are connected to the
tailwater through two tailrace tunnels of 109 m and 118 m respectively.
The operating range of the turbines goes from 427 to 612 J/kg, with flows
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
of 147 to 246 m3/s. Operation is guaranteed beyond rated gate opening in

the high head region, up to 140 MW.
Field Observations
Upon the commissioning of the machines, pulsations were encountered above

130 MW with the minimum tailwater level (Figure 7.38). The venting de
vice (through the shaft seal and runner cone) did not allow a natural air
admission. Although the pulsation was acceptable for higher tailwater lev
els, operation was temporarily restricted to 128 MW.

15
t/n f\ H
• f/n _ — \ - RMS -T.0
aAH/H
7 \
■L -V 0.5
f/f ref
12 US U
Figure 7.38: Full load pressure fluctuations at the prototype draft tube wall.
0 l I I * ,
Figure 7.39: Pressure fluctuations with the cut-off cone and the Snorkel tube
extension.
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

408 Cavitation
5-
AP 1 , I39MW
MWRMS i
% ft
\.
0 05
ir,
i
W
fto
15
t. I38MVV
A
I.
0- A f.'Hz
Figure 7.40: Power output f at unit 2. — unit 2 alone, unit 2 + unit 1.
Modifications were made to unit 2 to ease the natural air admission
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
through the runner cone [7.16, 7.46]: a funnel-shaped Snorkel tube (Figure
7.39) was such fitted to the runner cone that it can reach the level where the
top of the vortex cavity had been observed during the model tests. Some
instruments were installed to record the pressure and power output fluctu
ations. Compressors were also provided for a broader investigation of air
admission possibilities. With the air admission shut off, the overload range
up to 140 MW was tested under a constant test head. The full load pul
sation reduced remarkably with a much lower amplitude. The unmodified
unit 1 still generated excessive oscillations in that range of operation. Figure
7.40 shows power output fluctuations of unit 2 at 138 and 139 MW. If unit
2 runs alone, the oscillations are acceptable. If unit 1 also runs at full load,
the dominant frequency is shifted and the annoying fluctuations originating
from unit 1 are transmitted all the way to unit 2.
Additional tests with small forced air admission flows showed a further
reduction of both the frequencies and the amplitudes of full load pulsations.
Since the use of the Snorkel tube itself was enough to calm down the full
load pulsation, the air admission was finally abandoned. The runner cone
of unit 1 was also modified in the same way as the unit 2. Since then, the
operation of the Saucelle-Huebra plant machines is unrestricted all the way
up to 140 MW, without any losses of generating efficiency.

7.4.3 Model Tests

Preliminary Discussion
Although the user was fully satisfied with such a simple solution for this
stability problem, several questions were still needed to be answered:
• Why was not the full-load instability detected during the model tests
conducted in the designer's laboratory?
• How could the runner cone extension suppress the pulsation without
the air admission?
• How are the oscillations influenced by the long tailrace tunnels?

To answer these questions, the Hydro Vevey model was erected on IMHEF's
test installation. Its dynamic behaviour was investigated in detail by new
evaluation methods. The diameter of the model runner was 0.28 m. The
main test head was 15 m.
For the first test series (Figure 7.39), the original and the modified (with
the Snorkel tube extension) runners were tested at the plant Thoma number
<7P with the operating range covering the best efficiency energy coefficient;
and, the flow-rate was adjusted from part load to full load. The amplitude of
oscillations at the draft tube cone wall was plotted versus non-dimensional
frequency f/fn- The part load pressure fluctuations were not much modified,
but the amplitude difference of full load pulsations was remarkable, which
indicated a moderate improvement owing to the extension tube.
Proceeding from this encouraging result, an investigation into the influ
ence of the test head on the full load pulsation was performed. Figure 7.41
shows the dominant frequencies and the corresponding amplitudes under 10
m and 20 m heads. Note: Under these test heads, the Snorkel tube exten
sion made the pulsation stronger! Relative frequencies, however, were not
influenced by the test head.
Influence of the Test Circuit

Tests were also performed with a different feed pipe configuration (Figure
7.42). The tank inserted into the high pressure pipe modified its dynamic
response. The modification became strongest at twice the runner rotational
frequency. Amplitudes were globally lower, but the relative frequencies re
mained unchanged. This exploration result gives an important information
that the relative frequencies of this significant phenomena are influenced
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

410 Cavitation
cutoff cone . f. with Snorkel tube extension

1.0 0^20™
•0.5 0.5 ""*-
1.25 1.3 1.35 V*nt 1.25 1.3 1.35 ^ref

AH/H AH/H .
.%RMS 0m
20 m,\ ,n
%RMS yOv
1.0
0.5
^
J> V* 'Wjpm
5
°- c^S^
// \20m
^
Figure 7.41: Full load pulsation at the draft tube wall, cut-off cone and
Snorkel tube extension, under 10 m and 20 m heads.
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
Figure 7.42: Pressure oscillations at the draft tube wall, with different feed
pipe configurations.

t/»„.
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
Figure 7.43: Tailrace tunnel of unit 2 and full load pulsation at the draft
tube cone wall for three different tailwater configurations.

412 Cavitation
neither by the test head nor by the feed pipe dynamic response. While
its pulsation amplitude varies according to the test conditions, its signal is
well-organised in all cases.
Precautions should be taken if the turbine outlet is connected to tailrace
channels, instead of a free surface basin, such as the the Saucelle-Huebra
machines. Of course, we can build a model geometrically similar to the pro
totype tailwater works, but dynamic similitude will not be satisfied. Figure
7.43 shows pressure oscillations at the draft tube cone wall with three dif
ferent tailwater configurations: the short draft tube, an additional tunnel
length and the full tailwater tunnel (geometrically similar to the prototype
layout). The full load pulsation is clearly revealed for the short draft tube
configuration. It remains at a less severe level for the half tunnel configura
tion. It is hardly noticeable for the full tunnel configuration.
Prediction Elements
The variable quantity is the cavitation compliance. If the draft tube free
H=15m
a = 0.280
Figure 7.44: Non-dimensional cavitation compliance evaluated from model

tests, with two different tailwater configurations.
oscillations frequency is close to a resonance frequency of the full circuit,

cavitation resonance (instability) may occur. The estimation of cavitation
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`-
compliances from the observation of free oscillation frequencies on the model

is quite easy (Figure 7.44).

7.4.4 Stability of the Prototype Installation

Free Oscillations
The knowledge about cavitation compliance makes it possible to consider
the dynamic behaviour of a model turbine in the stability analysis of the
whole hydraulic system, as shown in Figure 7.45. The detailed data of the
Figure 7.45: Schematic representation of the waterworks for impedance cal

culations.
system are as follows.

Pipe Length Diameter Wave Speed Head Loss Coeff.
771 m m/s
1 1330.0 12.0 1300.0 0.02
2 38.0 8.0 800.0 0.02
3 30.0 12.0 1300.0 0.02
4 30.0 8.0 1300.0 0.02
5 38.0 6.5 1300.0 0.02
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
6 17.0 6.1 800.0 0.02

7 20.0 6.25 1300.0 0.02
8 40.0 10.7 1300.0 0.02
9 110.0 12.9 800.0 0.02
(10 30.0 8.0 1300.0 0.02)
(11 48.0 6.5 1300.0 0.02)
(12 17.0 6.1 1300.0 0.02)
(13 20.0 6.25 1300.0 0.02)
(14 40.0 10.7 1300.0 0.02)
(15 100.0 12.9 1300.0 0.02)

414 Cavitation
Surge tank: Ajieg = 1963.5 m2 kreq = 0.025 s/m2

Gate shafts: Achva = 50.0 m 2 fcrtJa = 1.5 s/m2
Operating conditions: Q = 258 m3/s H = 61 m l f C
The boundary conditions of the system are imposed:
• the free surface at the headwater,
• the free surfaces in the surge tank and gate shafts,
• unit 1 being shut,
• the free surface at the tailwater of unit 2.

--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
Figure 7.46: Critical cavitation compliances of unit 2.
Having imposed a flow-rate fluctuation on the system, the associated

flow-rate and energy fluctuations at the runner outlet of unit 2 are com
puted. The local impedance is computed from the other boundary condi
tions. The cavitation compliance that tunes the system at the considered
frequency is then evaluated. The critical cavitation compliance is also cal
culated for the zero admittance boundary at the runner outlet. Acoustic
transmissions are computed using distributed parameters. The tailrace of
unit 2 is represented by 10 equal-length sections (6 for the draft tube and
4 for the tunnel). The calculation (Figure 7.46) can single out both the

oscillation frequency obtained previously from the field tests; and, the criti
cal cavitation compliances predicted previously by the model tests for these
operating conditions. The results show the likely tunings of the prototype
installation for wave speeds of 1300 m/s or 320 m/s (harmonic mode). This
seems to be a good calculation if we consider that:
• the field recordings show no oscillations in the 0.4—0.8 Hz band, which
has been revealed by the calculation;
• the non-dimensional compliance, C* = 0.26, obtained by the model

tests, which is a lumped parameter, includes all additional compliances
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
owing to machine operation; the wave propagation speed of 1300m/s
then stands for the water under a mean pressure of 1.5 bar in fully
rigid ducts.
During the field tests, the operation of unit 2 (modified runner cone)
alone at 4>/cf>ref = 1.255 generated oscillations which were acceptable: AH/H
= 1.5% RMS and AP/P = 1.0% RMS at 0.75 n. On the other hand, the
parallel full load operation of unit 1 (cut-off cone) with unit 2 induced exces
sive fluctuations on unit 2: AH/H = 3.7% RMS and AP/P = 3.1% RMS
at 0.81 n.
AH, m f=1.68Hz,a=lO- s AQ.mVs

10.0 5.Q
Figure 7.47: Oscillation mode of the prototype waterworks at 1.68 Hz.
The critical cavitation compliance for unit 1 is also computed for the
frequency of 1.68 Hz which was obtained from the field tests. The boundary

416 Cavitation
conditions are:
• the free surface at the headwaters;
• the free surfaces in the surge tank and gate shafts;
• the free surface at the tailwater of unit 1;
• the free surface at the tailwater of unit 2, with the cavitation compli
ance as previously calculated.
The resultant compliance of unit 1 is a complex value, and only its real
part is retained (the imaginary part is very small). The frequency and
damping of free oscillations are then identified; and, the pressure and flow-
rate amplitudes are computed throughout the system. The amplitude is
adjusted to agree with the field recordings. The oscillation mode for the
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
complete waterworks can then be obtained as shown by Figure 7.47.
The cavitation compliance computed for unit 1 agrees well with the one
obtained from the model tests around <f>/4>ref = 0.25. The flow-rate ampli
tude jumps at the surge tank insertion, manifold and gate shafts are evident
in the computed mode shape. Although these elements may seem insignifi
cant at first glance, they must be included event in a preliminary calculation
as they modify the vibration mode.
Comments
The stability analysis of the Saucelle-Huebra installation clearly shows the
risk of cavitation resonance associated with the full-load operations. Labo
ratory tests can indicate neither its amplitude nor the exact operating con
ditions which will cause the cavitation resonance. Nevertheless, the model
tests for a given runner design can indicate, to some extent, the likelihood
of this full-load pulsations caused by cavitation resonance. They can also
provide the dynamic characteristics of the turbine, which are necessary for
the numerical analysis of this system-instability.

References
7.1 Acosta, A. J. (1958), ' An Experimental Study of Cavitating Induc-
ers', Second Symposium on Naval Hydrodynamics, Office of Naval
Research, Dept. of Navy, ACR-38, August 1958.
7.2 Angelico, G., Muciaccia, F. Rossi, G. (1986), 'Part Load Behaviour of

a Turbine: a Study on a Complete Model of Hydraulic Power Plant',
IAHR Symposium, Montreal.
7.3 Angelico, G., Fanelli, M., Siccardi, F. (1987), 'Hydraulic Machinery

as a Transfer Element for Steady Pulsating Quantities', IAHR W.G.,
Line.
7.4 Badie-cassagnet, A. et al. (1981), 'Application de I'intensimetrie

Acoustique a Pidentification des Sources de Pulsation de Pression
dans les Circuits', CETIM, Senlis.
7.5 Brennen, C , Acosta, A.J. (1976), 'The Dynamic Transfer Function

for a Cavitating Inducer', ASME J. Fluids Eng., Vol.98, No.2, 1976-6,
pp.182-191.
7.6 Brennen, C. (1978), 'Bubbly Flow Model for the Dynamic Character
istics of Cavitating Pumps', J. Fluid Mech., Vol.8-9, part 2, pp.223-
240.
7.7 Brennen, C. E., Meissner, C , Lo, E. Y., Hoffman, G. S. (1982), 'Scale

Effects in the Dynamic Transfer Functions for Cavitating Inducers',
ASME J. Fluids Eng., 104-4, pp.428-233.
7.8 Campmas, P. (1960), 'Stabilite du Regime de Fonctionnement des

Turbines Francis,' AIRH Symposium, Nice.
7.9 Campmas, P., Giraud, H. (1960), 'Analyse du Fonctionnement d'une

Turbine Francis: Comparaison des Investigations sur Modele Reduit
et sur Turbine Industrielle', AIRH Symposium, Nice.
7.10 Dorfler, P. (1983), 'Francis Turbine Impedance-Experimental Results

vs. Theoretical Model,' IAHR W.G., Milan.
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
7.11 Dorfler, P. (1983), 'Evaluation Concepts for Low Frequency Oscilla

tions in Hydraulic Reaction Turbines', IAHR W.G., Milan.

418 Cavitation
7.12 Dorfler, P. (1985), 'Francis Turbine Surge Prediction and Prevention',

Waterpower' 85, Las Vegas.
7.13 Fanelli, M. (1991), 'Research on Off-design Behaviour of Francis Tur
bines: an Overview of Present State, Difficulties, Open Problems,
Needs and Strategies', IAHR W.G., Milan.
7.14 Fisher, R. K., Ulith, P. (1982), 'Comparison of Draft Tube Surging
of Homologous Scale Models and Prototype Francis Turbines', Voith
Research and Construction, Vol.28e.
7.15 Fritsch, A., Maria, D. (1987), 'Dynamic Behaviour of a Francis Tur
bine at Partial Load: Model-prototype Comparison', IAHR W.G.,
Lille.
7.16 Grein, H. (1980), 'Vibration Phenomena in Francis Turbines: Their
Causes and Prevention', IAHR Symposium, Tokyo.
7.17 Greitzer, E. M. (1981), 'The Stability in Pumping Systems - The
1980 Freeman Scholar Lecture', ASME J. Fluids Eng., Vol.303, No.2,
pp.193-242, 1981-6.
7.18 Henry, P. (1984), 'La Cavitation dans les Machines Hydrauliques',
Publication EPFL, Lausanne.
7.19 Henry, P., Graeser, J. E., Prenat J. E., Wegner M. (1984), 'Scale
Effects Concerning Hydraulic Quasi-stationary Oscillations on a Tur
bine Model and Its Test Circuit', IAHR Symposium, Stirling.
7.20 Henry, P. (1984), 'La Cavitation dans les Machines Hydrauliques',
Publication EPFL, Lausanne.
7.21 Jacob, T., Maria, D., Prenat, J.-E. (1987), 'Comportement Dynani-
ique D'une Turbine Francis 1 Forte Charge. Comparaisons modele -
prototype', SHF Comite technique 134, Paris.
7.22 Jacob, T., Prenat, J. E., Grenier, R. (1988), 'A Characterization
Procedure for the Dynamic Behaviour of Francis Turbines: Practical
Comparison of Elbow and Moody Type Draft Tubes', IAHR Sympo
sium, Trondheim.
7.23 Jacob, T., Prenat, J. E. (1990), 'Generation of Hydro-acoustic Dis
turbances by a Francis Turbine Model and Dynamic Behaviour Anal
ysis', IAHR Symposium, Belgrade.
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

7.24 Jacob, T., Prenat, J. E., Vullioud, G., Lopez Araguas, B. (1992),
'Surging of a 140 MW Francis Turbine at High Load, Analysis and
Solution', IAHR Symposium, Sao Paulo.
7.25 Jacob, T. (1993), 'Evaluation sur modele Reduit et Pre'ction de la

Stabilite de Fonctionnement des Turbines Francis', These EPFL,
Lausanne.
7.26 Kamijo, K., Shimura, T., Watanabe, M. (1977), 'An Experimental
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
Investigation of Cavitating Inducer Instability', ASME Paper 77-
WA/FE-14, 1977-11, and also 'A Visual Observation of Cavitating
Inducer Instability', NAL TR-598T, 1980-5.
7.27 Kazan, M. N. (1962), 'Turbine Francis: Etude de L'ecoulement a la

Sortie D'une Roue Francis', These EPFL, Lausanne.
7.28 Kercan, V., Bajd, M. (1988), 'Computation and Experimental Inves

tigation of Draft Tube Vortex in Francis Turbines of Various Specific
Speed', IAHR Symposium, Trondheim.
7.29 Li, S. C , Zhang, Y. J., Hammitt, F. G. (1983), 'Low-frequency Fluc

tuations Associated with Cavitating Venturi Flows', DRDA Rept.
No.UMICH 014571-64-1, University of Michigan, Ann Arbor, USA.
7.30 Li, S. C , Zhang, Y. J., Hammitt, F. G. (1986), 'Statistical Consid

erations of Pressure Pulses from a Cavitating Venturi', Proc. Instn.
Mech. Engrs. 200, N0.C6, pp381-387.
7.31 Li, S. C. et al (1989), 'Cavitation Resonance in a Venturi Loop',

ASME FED, Vol.79, pp71-74.
7.32 Li, S. C. (1992), 'Pressure Fluctuations in Cavitating Draft-tube

Flows', ASME FED, Vol.136, ppl-6.
7.33 Massey, I. C. (1976), 'The Suction Instability Problem in Rotody-

namic Pumps', International Conference on Design and Operation of
Pumps and Turbines. NEL, East Kilbride, Scotland, Paper 4.1.
7.34 Muciaccia, F., Rossi, G. (1983), 'Experimental Approach to the Study

of Stationary Pulsations at the Rope Frequency on Francis Turbine
Models: Measurement Techniques and Some Results', IAHR W.G.,
Milan.

420 Cavitation
7.35 Muciaccia, F., Rossi, G., Zanetti V. (1984), 'Study on Model of the
Dynamic Behaviour of a Hydraulic Machine', IAHR Symposium, Stir
ling.
7.36 Muciaccia, F., Rossi, G. (1985), 'Experimental Approach to the Study

of the Dynamic Behaviour of a Francis Turbine Model at High Specific
Speed when Operating at Reduced Load', IAHR W.G., Mexico.
7.37 Muciaccia, F. (1991), 'Influence of the Extension of the Pier Nose,

Inside the Draft Tube Elbow of a Francis Turbine, on the Pressure
Pulsations and Rotating Radial Thrust at Suction Vortex Frequency',
IAHR W.G., Milan.
7.38 Murai, H., Takeuchi, H. (1969), 'Observation of Cavitation and Flow

Pattern in an Axial Flow Pump at Small Flow Rate', Mem. Inst. high
Sp. Mech., Japan, Vol.254, (1968/1969), No.246, pp.315-333.
7.39 Nishi, M., Kubota, T., Matsunaga, S., Senoo, Y. (1980), 'Study of
Swirl Flow and Surge in an Elbow Draft Tube', IAHR Symposium,
Tokyo.
7.40 Nishi, M., Matsunaga, S., Kubota, T., Senoo, Y. (1984), 'Surging
Characteristics of Conical and Elbow-type Draft Tubes', IAHR Sym
posium, Stirling.
7.41 Nishi, M. (1992), 'Pressure Data Analysis by Using Nishi's Method',

Kyushu Institute of Technology.
7.42 Numachi, F. (1954), 'Summary Report on the Research of Cavita
tion Phenomena Obtained Hitherto by our Institute', Rep. Inst. High
Speed Mech., Tohoku Univ., Vol.14, No.40, ppl59.
7.43 Palde, U. (1974), 'Model and Prototype Turbine Draft Tube Surge
Analysis by the Swirl Momentum Method', IAHR Symposium, Vi
enna.
7.44 Rosenman, W. (1965), 'Experimental Investigations of Hydrodynam-

--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
ically Induced Shaft Forces with a Three Bladed Inducer', Proc. Sym
posium on Cavitation in Fluid Machinery, ASME WAM, pp.172-195.
7.45 Sack, L. E., Nottage, H. B. (1965), 'System Oscillations Associated

with Cavitating Inducers', ASME J. Basic Eng., Vol.87, Ser.D, No.4,
pp.917-925.

7.46 Thicke, R. H. (1980), Methods of Controlling Turbine Draft Tube

Vibrations and Stability, Canadian Electrical Association, Hydraulic
Power Section.
7.47 Tsujimoto, Y., Kamijo, K., Yoshida, Y. (1993), 'A Theoretical Anal
ysis of Rotating Cavitation in Inducers', ASME J. Fluids Eng.,
Vol.115, No.l, pp.135-141, 1993-3.
7.48 Ulith, P., Jaeger, E. U., Strcheletzky, M. (1974), 'Contribution to

Clarifying the Inception of Nonstationary Flow Phenomena in the
Draft Tube of High Specific Speed Francis Turbines Operating at
Part Load', IAHR Symposium, Vienna.
7.49 Vatcher, T. R., Hunter, M., Perreault, G., Coulson, D. M. (1984),

'Hydraulic Resonance in a 78 MW Francis Turbine - Field Test Re
sults', IAHR Symposium, Strirling.
7.50 Viegner, M. (1983), Cavitation as Erosion and Instability Cause-

experiments and Remedies, Lahore.
7.51 Wade, R.B. and Acosta, A.J. (1967), 'Investigation of Cavitating

Cascades', Trans. ASME, Ser. D., Vol.89, No.4, pp693.
7.52 Wylie, E. B., Streeter, V. L. (1985), Fluid Transients, FEB Press,

Ann Arbor.
7.53 Yamaguchi, Y. and Miura, S. (1981), 'Characteristics of Radial Force

Fluctuation on Runners of Francis-type Pump-turbines', Bulletin of
the JSME, vol. 24, No. 195, PP 1586-1593.
7.54 Yamamoto, K. (1980), 'An Experimental Study on Instability in a

Cavitating Centrifugal Pump with a Volute Suction Nozzle', Proc.
IAHR Symposium, Tokyo, pp.303-312.
7.55 Yamamoto, K. (1992), 'Instability in a Cavitating Centrifugal Pump

(3rd Report, Mechanisms of Low-Cycle System Oscillation)', Trans.
JSME, 58-545 (1992-1), pp.180-186.
7.56 Yedidiah, S. (1986), 'Certain Effects of Recirculation on Cavitation

in Centdfugal Pumps', Proc. IMechE, 200, pp283-292.
7.57 Yokoyama, T., Hishida, Y. and Niikura, K. (1984), 'Some Operating

Experiences at Off-Design Points for Pump-Turbines and Hydraulic
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

422 Cavitation
Turbines', Proceedings of 13th IAHR Symposium on Fluid Machinery,

Stirling.
7.58 Young, W. E. et al (1972), 'Study of Cavitating Inducer Instabilities,
Final Report', NASA-CR-123939, 1972-8.
7.59 Zielke, W. (1980), Unsteady One-dimensional Flows in Complex Net
works and Pressurized Vessels: Vibrations and Resonance in Hy
draulic Systems, Von Karman Institute for Fluid Dynamics.
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

Chapter 8
Unsteady Cavitation Flows

Caused by Machine
Transients
Part 1: Turbine Transients
C. S. Martin
8.1 Introduction
The transient modes of operation that pumps and turbines must be designed
to accommodate can result in enhanced cavitation on the low-pressure side of
the machine and its associated water passage. This expanded cavitation can
potentially lead to more severe transient phenomena, and indeed, even wa
terhammer under more severe circumstances. There have been incidents of
waterhammer associated with load rejection of hydraulic turbines for which
a vapour cavity formed underneath the runner and subsequently collapsed
upon flow reversal. This phenomenon has often been labeled reverse water-
hammer because of the reversal of flow back into the void below the turbine
runner. Actually, a more representative expression draft-tube column separa
tion will be used in this volume. Draft-tube column separation has occurred
in installations of Francis and Kaplan turbines during load rejection.
The transient regimes for hydraulic turbines range from normal startup
(load acceptance) and shutdown to the more severe case of full load rejection,
the latter usually being the most severe case regarding potential draft-tube
column separation. For pumps and pump-turbines in the pumping mode
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
only pump startup with an abnormally long suction pipe would appear to
423
424 Cavitation
be a condition that would lead to pump suction column separation.

The occurrence and severity of turbine draft-tube column separation
clearly depends upon such factors as initial (steady-state) cavitation regime
and degree, amount of turbine load and associated vortex core and atten
dant cavitation, amount of free air present from both release of dissolved gas
and injection, draft-tube length, and the rate of closure of the turbine guide
vanes.
The current practice of longer tailraces coupled with more frequent cy
cling of units such as pump-turbines has led to the impetus to conduct more
studies into the nature of part-load swirling, draft-tube column separation
and any severe waterhammer.
8.2 Types of Turbine Cavitation

Cavitation in hydraulic machinery has been investigated extensively, as doc
umented in Chapters 5,6 and 7. Of interest in draft-tube column separation
is the initial cavitation condition and any impact it may have on the devel
opment of a large vapour-air filled void that may cause waterhammer upon
collapse. The effect of various hydraulic turbine designs as represented by
specific speed will be discussed with regard to cavitation number, percentage
of load, and draft-tube pulsations and surging.
The recognition of cavitation and its effect on the performance of hy
draulic machinery has been known for more than a century, resulting in
numerous laboratory and field investigations regarding the nature of the
phenomenon. Detailed accounts are provided in Raabe, 1985 [8.48] and in
Chapters 5 and 6. This section will only be concerned with brief descriptions
of types of cavitation on turbine runners, both Francis and Kaplan, and on
pump-turbine impellers.
8.2.1 Francis T u r b i n e s
One of the earlier studies with careful observations of the cavitation patterns
in draft-tubes of Francis runners was provided by Deriaz in 1960 [8.10],
who enhanced the investigation of power swings due to vortex shedding by
inducing cavitation to produce the visible rope vortex. The paper illustrates
the rope-like vortex made visible by cavitation and its precession in the
direction of runner rotation at part-load operation, Figure 8.1. Dziallas
in 1964 [8.13] added to this knowledge for a ns = 174 Francis turbine by
measuring pressure pulsations for part-, optimum-, and over-load conditions.
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

Unsteady Cavitation Flows: Turbine T\nnsients 425
Figure 8.1: Sketches from Deriaz, 1960 [8.10] defining vortex centre Q, pre
cession of vortex core P
Relative information regarding pressure excursions are provided for the range
of loads, as sketched in Figure 8.2. The effect of dissolved air content and
head on the cavitation pattern exhibited by a Francis runner was thoroughly
observed by Lecher in 1968 [8.29].
Wegner and Roman in 1984 [8.54] define four types of cavitation on a
Francis runner
• Outlet edge cavitation - the formation of small individual bubbles
attached to the blade. This type of cavitation has a strong dependence on
submergence and has a low noise level.
• Inlet edge cavitation between blades - the formation of large bubbles
between runner vanes. Inlet edge cavitation between blades yields a high
noise level with relative broad-band noise and sensitivity to unit speed.
• Inlet edge cavitation on suction side of blades - the formation of a
vortex or void on the lowest part of inlet edge, leading to a severe pressure
fluctuations.
• Part load vortex cavitation - the formation of a core vortex on turbine
crown, resulting in low-frequency noise and sensitivity to guide vane opening.
Borciani and Rossi in 1983 [8.3] evaluated the effect of the cavitation
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

426 Cavitation
Figure 8.2: Sketches from Dziallas, 1964 [8.13] showing zone of dead water
vortex for various loads
index on the operation of a Francis turbine and a pump-turbine over a range

of operating conditions, including such abnormal operation regions as partial
pumping, turbine part load, and runaway. The problem of more frequently
occurring over-load conditions and the attendant difficulty in modelling cav
itation bubble instability are addressed by Anderson in 1989 [8.1].
Draft-Tube Surging and Oscillations
Power swings due to draft-tube surging at part-load operation has been

recognised since 1912, according to Rheingans in 1940 [8.49], who developed
the initial correlations between the frequency of the draft-tube surging and
the runner speed. A carefully conducted model study of a hydraulic turbine
by Ulith et al. in 1974 [8.52] illustrated the pattern of low-frequency regular
oscillations associated with the so-called cavitating dead-water vortex core.
Portions of the unstable behaviour of the dead-water zone are illustrated
in Figure 8.3. Of importance to hydraulic transients is the interaction be
tween the draft-tube and the turbine and the overall pipe system and its
inherent natural frequencies. Clearly these acoustic characteristics can be
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

Unsteady Cavitation Flows: Turbine Transients 427
Figure 8.3: Unstable behaviour of dead zone during rotation (Ulith et al.,
1974 [8.52])
significantly modified by the presence of vapour and free air, as outlined

later. The beneficial effect of the injection of air into the draft-tube region
through various openings on reduction of draft-tube surging has been docu
mented by Grein in 1974 [8.21], 1980 [8.22], and Pejovic et al. in 1984 [8.45],
1986 [8.47], among others. For utilisation of hydraulic model results as well
as knowledge regarding analytical modelling Fisher et al. in 1980 [8.17] have
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
shown that laboratory measurements of pressure pulsations can be extrapo
lated to full scale provided that there is minimum dynamic action between
the draft-tube excitation and the hydraulic system itself. The mathemati
cal modelling of the pulsations and the associated hydraulic system under
nominal steady-state operation has been treated by Doerfler in 1980 [8.11],
1985 [8.12] and Fanelli in 1988 [8.16].
A series of papers by Nishii et al. in 1980 [8.34], 1982 [8.35], 1984 [8.36]
and 1986 [8.37] significantly enhanced the understanding of the effect of swirl
flow in conical and elbow-type draft-tubes, with and without the presence of
a runner. For an elbow-type draft-tube with induced swirl, Figure 8.4 shows
the effect of vortex core cavitation on the fluctuating pressure (Nishi et al. in
1980 [8.34]). As expected, a correlation existed between a Strouhal number,
the swirl rate (angular momentum function), and the cavitation index, as
shown in Figure 8.5. Questions regarding the use of a test rig without a
runner were answered by comparing pressure fluctuations with a draft-tube
equipped with a real model runner, Figure 8.6. Figure 8.7 depicts the types of
vortex cores identified with the draft-tube surging, Nishi et al. in 1980 [8.34].

428 Cavitation
—* 0 . 1 s e c
— l o . i sec
'21 VAKAAAAA-
L2 I— --I0.1 s e c
* T
\-c - i ^ O . l sec
f »
L3
•HO.l s e c
Figure 8.4: Pressure traces for non-cavitating vortex core (Nishi et ai, 1980
[8.34])
St
1 TTT
0.8
0.4
ra = 1 . 1
I 1 ,) 1 I L_L
4 6 8 10 4 6 8 102
Figure 8.5: Effect of cavitation on Strouhal number for vortex core, St vs.
K (Nishi et al. in 1980 [8.34])
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,

A* i—r "i ■ r-
12 -
i—rr
m - 1.1
4
'! O P r e s e n t Draft Tube
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
_L_L XX
4 6 8 10 4 6 8 10
Figure 8.6: Surging in simulated draft-tube compared with model draft-tube,

A * vs. K (Nishi et al. in 1980 [8.34])
(a) (b)
Figure 8.7: Types of vortex cores associated with draft-tube surge (Nishi et
al. in 1980 [8.34])
(a) At the crest of pressure; (b) At the trough of pressure wave.

430 Cavitation
Figure 8.8: Flow regimes and associated pressure fluctuations for various
swirl and cavitation values (Nishi et al. in 1986 [8.37])
These fundamental studies with induced swirl generated a large synchronous

pressure fluctuation at all measuring stations. A very fundamental study by
Nishii et al. in 1984 [8.36] allowed for the identification of four flow regimes
that depended on the swirl parameter, as observed through a transparent
elbow-type draft-tube. With reference to Figure 8.8 the four regimes are
• Flow regime I - at a small swirl flow the vortex core is a long straight
rope with minimum stall and small pressure fluctuations
• Flow regime II - at a higher swirl rate irregular vortexing occurs in a
transitory fashion
• Flow regime III - a stable spiral vortex forms that almost extends to
the draft-tube. There are quite regular pressure fluctuations
• Flow regime IV - the existence of a very strong swirl rate can lead to
the formation of two vortex cores with quite irregular pressure variation.
Muciaccia et al. in 1984 [8.33] provide data on the dynamic behaviour
of Francis turbines in terms of pulsation amplitude and frequency of the
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

±£g-%
discharge
n n h rfHIRHi«
—L=>=Msbjtl«^ shaft torque
AH
H
head
%
- H
pressure
/ta.
m pressure
MP5 ■Si-., I
±j$«
l_ ^ fisfis ■^+0*9****. S--....-
suction pressure
far from the runner
±KR
X
JffiF
.. J/fw.'
rotating radial
thrust
_(X
1.0 a
relative gate opening opt
Figure 8.9: Rope frequency variations with relative gate opening for Francis
turbine (Muciaccia et ai, 1984 [8.33])
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

432 Cavitation
rope vortex over a wide range of operation, concluding that only moderate
pressure pulsations can be scaled up using modelling laws. Figure 8.9 shows
the effect of relative gate opening on the rope frequency. They emphasise
that physical hydraulic modelling is not completely valid if the hydraulic
circuits are not also satisfying scaling laws. Recent detailed studies of cavi-
tating draft-tube flows by Li in 1992 [8.30] show that the natural frequency
of the cavitation cloud is an important parameter of the system characteris
tics and that there is an additional frequency component beyond that of the
Rheingans value.
Bhan et al. in 1988 [8.2] conducted a useful survey relative to practical
experience regarding the occurrence and means of reducing draft-tube surges
at a number of Francis turbine installations.
8.2.2 Kaplan Turbines

For draft-tube column separation occurring below Kaplan runners the run
away cavitation characteristics investigated by Ishii in 1962 [8.26] and Chen
and Whippen in 1978 [8.7] are of value. Ishii defined two types of area cav
itation under runaway conditions, the first being a normal occurrence with
similar cavitation on every blade, while the second is termed abnormal be
cause the cavitation is different on adjoining blades but similar on alternate
blades. Figure 8.10 shows the two different types of cavitation defined by
Ishii. Additional visualisation is provided by Eichler and Jaeger in 1979
Figure 8.10: Examples of first and second type of cavitation on Kaplan

turbine blades (Ishii, 1962 [8.26])
[8.14] with the purpose of evaluating scaling of erosion, Figure 8.11.

As documented by Fisher and Cybularz in 1985 [8.18] there are several
types of cavitation that can occur on Kaplan turbine blades and within the
draft-tube. Four principal types are defined that are not mutually exclusive
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

Figure 8.11: Effect of cavitation index on Kaplan blade cavitation (Eichler

and Jaeger, 1979 [8.14])
• Area cavitation - formation over an area of the suction side of a blade.

The bubbles form randomly and travel to higher pressures regions of flow
and collapse. Importantly, the effects on turbine performance are significant.
• Entrance edge cavitation - formation of bubbles in an off-design op
eration wherein the blade angle may have a negative or positive angle of
incidence.
• Cavitating vortices (vortex-type flow) - formation of bubbles in an
off-design operation that are relative insensitive to the cavitation index.
• Leakage cavitation - formation of bubbles in the gap region adjacent
to the low pressure side of the turbine blade.
8.3 Draft-Tube Column Separation (Reverse Wa-

terhammer) Incidents
There have been several reported incidents in Francis and Kaplan turbine
hydroelectric installations associated with load rejection transients. These
incidents, in some case accidents, albeit often referred to as reverse water-
hammer in the literature, for clarity the phenomenon is called draft-tube
column separation in this treatise.
Pejovic et al. in 1980 [8.42, 8.43, 8.44] reported on an accident experi
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
enced in a Kaplan installation due to governor failure at part load, followed

by a sudden power increase from 8 MW to 20-25 MW. The accident oc
curred subsequently due to emergency operation of a shutoff button, causing
a severe case of load rejection, most likely creating a large vapour cavity
within the draft-tube, that collapsed causing waterhammer once the sepa
rated water column rejoined. The shock broke a runner blade and lead to
a thorough investigation of axial thrust magnitudes. The solution to pre-

434 Cavitation
Figure 8.12: Site measurement of moderate effect of draft-tube column sep

aration reported by Chaudhry, 1982 [8.5], 1991 [8.6]
vent future failure was an extension of the closure time of the guide vanes.
Two-stage closure as well as air injection should be considered for such oc
currences.
In 1976 Unit 2 of a power station experienced a minor water-column sep
aration event as reported by Chaudhry in 1982 [8.5],1991 [8.6]. A prototype
test during load rejection of 257 MW yielded turbine inlet pressures almost
to design values for the penstock. The traces in Figure 8.12 illustrate a coin
cident pressure spike in the draft-tube, strongly suggesting the formation of
a cavitation bubble of moderate size. The partial-open position of the guide
vanes allowed transmission of the pressure pulse to the upstream side of the
unit.
Measured pressures in the draft-tube of a 160 MW Francis turbine by
Dahlbeck in 1986 [8.8] indicated the creation of large vapour cavity during
load rejection at a relatively fast rate of guide vane closure. By prolonging
the time of guide vane closure cavitation was suppressed.
At the Paulo Afonso IV Plant, 402 MW Francis turbines experienced at
least partial draft-tube column separation, as reported by D'Amorin in 1989
[8.9], and illustrated in Figure 8.13. The recommendation for suppression of
column separation was the extension of the time of guide vane closure from
8.5 to 10-11 seconds.
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,

Figure 8.13: Draft-tube column separation experienced by load rejection of

large Francis unit (D'Amorin, 1989 [8.9])
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

436 Cavitation
8.4 Physical Modelling of Cavitating

Transient Flows
One of the first physical model addressing the problem of draft-tube column
separation during load rejection was conducted by Watanabe and Mutaguchi
in 1979 [8.53]. A physical hydraulic model of a pump-turbine (N = 300 MW
Figure 8.14: Test rig used for model turbine by Matsumoto et al., 1986 [8.32]
(1) Scroll; (2) Diffuser; (3) Stay vane; (4) Guide vane; (5) Link; (6) Rotating
disk; (7) Screen.
and H = 300 m) was built to simulate various modes of transient operation,

principally load rejection. A physical model was chosen because of stated
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
difficulties in analytical modeling, realizing the problem of representing the

complex geometry and flow pattern. The 20:1 model generated water column
separation under the more severe condition it was subjected to. For some
lesser cavitation conditions there was no apparent bubble formation on the
measured pressure traces in the draft-tube.
Matsumoto et al. in 1986 [8.32] constructed a plexiglass model in or
der to gain more fundamental knowledge of the effect of swirl on hydraulic
transients, in particular draft-tube column separation due to load rejection.
The test rig for these studies is shown in Figure 8.14. Pressure traces sim
ilar to those later published by Nonoshita et al. in 1991 [8.40] in Figure
8.15 indicated that there was an effect of release of dissolved gas, effectively

Unsteady Cavitation Flows: Turbine TYansients 437
1602.2 1802.0 1951.8 2002.1 2051.9 2102.2 2151.9 2202.3
Figure 8.15: Measured pressure traces in a model draft-tube during load

rejection (Nonoshita et al. in 1991 [8.40])
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

438 Cavitation
(a)m-0
GVOR-100%
1
JJUg
(b)m-0.24
" GVOR-94%
0 1 2 » 4 5 «
Tlu •
Figure 8.16: Effect of swirl and effect of dissolved air content on draft-tube
column separation as measured by Nonoshita et al., 1991 [8.40]
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

aoo tUNH
fr> 4>0J>2
_ WOO
•oo
500
0 _ j L
JOO _
tllNM
.! CM
1 £
Mi MOO *
2ioo 2
500 2
200
M1N2.
( 0 C.-1JJ
100
MO
Figure 8.17: EflFect of dissolved gas on pipe column separation as measured

by Wiggert and Sundquist, 1979 [8.55]
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

440 Cavitation
attenuating the maximum surge. A bubbly region appeared due to the pen
etration of the core beneath the runner. Other work by Nonoshita et al.
in 1988 [8.38], 1990 [8.39], 1992 [8.41] expanded the knowledge to show the
effect of initial swirl at part load, and the increased effect of gas release
and attendant attenuation of the maximum pressure due to an increased
inter-facial area, as shown by the results in Figure 8.16. The fact that the
formation of a vapour-air mixture is similar to that in pipes is borne out
by the plots in Figure 8.17 taken by Wiggert and Sundquist in 1979 [8.55].
Nonoshita et al. in 1992 [8.41] point out that a two-step closure of the guide
vanes may create only local draft-tube column separation that could inter
act with the surge directly associated with the swirl. On the other hand,
single-stage guide vane closure can produce a large vapour cavity that is not
affected by the swirl.
8.5 Two-Phase Flow Modelling in Conduits

In any development of accurate and definitive comprehensive analytical mod
els, whether one-, two- or three-dimensional, one must recognise the fact that
two-phase flow in conduits, and to some extent in draft-tubes, can occur in
various regimes. Depending upon a number of factors, but in particular
upon the mass flow rates in the respective phases, the flow pattern may
range from completely dispersed gas (bubbly flow) to completely dispersed
liquid (mist or annular mist). Gravity can produce a significant effect on
the flow pattern development as the conduit inclination is changed. For
steady flow some of the flow regimes that have been identified for vertically
upward flow and horizontal flow are presented in Figure 8.18. The number
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
of possible regimes that can be identified in either case depends to some

extent on the observer. The vast difference between some of the flow patters
exhibited by Figure 8.18 would suggest that the possibility of the develop
ment of a universal two-phase analytical model is rather remote. In fact, the
effect of relative velocity between the phases, phase interaction, and heat,
mass and momentum transfer can have a much greater effect for one flow
regime than another. Although the most suitable model may vary depend
ing upon the flow regime analysed, for flow in long conduits the assumption
of one-dimensionality is usually not a severe limitation.
Unless there exists a significant difference in the velocities of the respec
tive phases the homogeneous model can be applied with moderate success
to dispersed flows, whether bubbly or droplet. For the so-called homoge-

Bubbly Slug Churn Annular ^ — — — ^ ^

Vertical Flow Horizontal Flow
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
Figure 8.18: Definition of two-phase flow regimes in pipes (Martin, 1993
[8.31])
neous model the components are treated together as a single pseudo-fluid

with average properties. Inasmuch as the slip between the gas and liquid
is not considered, the gas and liquid velocities are assumed equal at every
instant. In some transient and steady flows, inertial and gravitational effects
can play an important enough role that the relative velocity between the air
bubbles and the liquid should be considered. Hence, apart from the simplic
ity involved, a homogeneous model may not be appropriate in many cases.
In the separated-flow model the phases are considered to flow side-by-side,
interacting with each other. The fact that two-phase flow in conduits can
occur in many different regimes - bubbly, slug, churn, wispy-annular, and
annular, to name a few - complicates the development of comprehensive
one-dimensional models.
8.5.1 Acoustic Velocity

Equation (8.1) is the theoretical acoustic velocity or wave propagation speed
a of a pressure wave in a two-phase mixture in a conduit under the associated
assumptions of homogeneous flow and no slip,

442 Cavitation
The term D/Ee can be important in single-phase liquid flow or in two-

component flow with very low values of the void fraction a. For moderate to
low values of a the pipe wall elasticity effect is minimal, however. The effect
of the gaseous component is represented by two quantities - the void fraction
a and the pressure p, as represented by the bulk modulus of elasticity of the
gas Kg. Various investigators have measured the acoustic velocity in two-
phase and two-component flow situations. Many of the measurements have
been conducted in the bubbly-flow regime . Figure 8.19 shows the results
reported by Kobori et al. in 1955 [8.28], for an air-water mixture for low
void fractions.
°o 0002 oooS aooe 5355 aoio

Void Fraction a
Figure 8.19: Measured and theoretical wave speed in air-water mixture (Ko
bori et al, 1955 [8.28])
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
Researchers in acoustics have also investigated the effect of the frequency

of oscillation on the acoustic velocity. Figure 8.20 shows recent measure
ments by Ruggles, Scarton, and Leahy in 1989 [8.51] for different bubble
sizes in a bubbly batch-flow apparatus.

m
J5
e m
a
a
« m
K4
11*
VoU RKtiea OL'OMS

I
!* 1M
' k ' k ' \k ' ill "ft)
frequency Is Hs
Figure 8.20: Effect of frequency and bubble size on acoustic velocity (Ruggles
et al. , 1989 [8.51])
Slug-Flow Theory of
' Henry, Gxoliaes and ftuskc
• Experiment
*%%%£% C - M l » * « = ; . , - 6 1 6 m*c)
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
0.1 0L2
Void Fraction a
Figure 8.21: Wave propagation in slug-flow regime (Martin, 1993 [8.31])

444 Cavitation
8.5.2 Slug Flow

Slug flow may occur in conduits in all orientations - horizontal, sloping, or
vertical. As reported by Martin in 1993 [8.31], wave propagation of a two-
phase bubbly flow and slug flow of an air-water mixture was investigated in
a 26.0 mm clear lucite pipe 18.8 m in length. Figure 8.21 shows the results
for a range of void fractions in the slug-flow regime, compared to the theory
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
of Henry, Grolmes, and Fauske in 1969 [8.24]. Clearly, the homogeneous
model yields a much better representation than their theory (also see Hsu
and Graham in 1976 [8.25]), which is based upon time-of-travel between
idealised plugs of gas and liquid placed end-to-end.
8.6 Analytical Modelling of Cavitating Transient

Flows
Jordan in 1975 [8.27] refers to potential draft-tube column separation as dan
gerous to catastrophic. The paper considers both rigid column and elastic
waterhammer analytical approaches, as well as consideration given to the
introduction of air. Rongqiao in 1986 [8.50] developed a simple column sep
aration model that included the two-component equations given in Martin,
1993 [8.31], but no calculations.
Nonoshita et al. in 1991 [8.40] used the method of characteristics to
actually simulate one of the model test with quite reasonable results, as
shown in Figure 8.16. Their model included the effects of gas release by
assessing the gas transfer at the vortex core interface.

References
8.1 Anderson, A. (1989), 'Cavitation Driven Draft-Tube Oscillations of
Francis Turbines at Overload', International Congress on Causes and
Accidents in Fluid Systems, Sao Paulo, Brazil, Vol. 1, March 6-10,
pp. 33-39.
8.2 Bhan, S., Codrington, J. B., and Mielke, H. (1988), 'Reduction of

Francis Turbine Draft-Tube Surges', ASME Fifth International Sym
posium on Hydropower Fluid Machinery, Winter Annual Meeting,
Chicago, November 27 - December 2, pp. 95-102.
8.3 Borciani, G. A., and Rossi, G. (1983). 'Cavitation-Parameter Effects

on Francis Turbines and Francis Type Pump Turbines', ASME Sym
posium on Performance Characteristics of Hydraulic Turbines and
Pumps, Boston, November 13-18, pp. 75 - 82.
8.4 Cassidy, J. J. and Falvey, H. T. (1970), 'Observations of Unsteady

Flow Arising After Vortex Breakdown ', Journal of Fluid Mechanics,
Vol. 41, Part 4, pp. 727-736.
8.5 Chaudhry, M. H. (1982), 'Experience on Hydraulic Transients in Hy

dropower and Pumping Plants', 11th IAHR Symposium on Hydraulic
Machinery, Amsterdam, 1982, Vol. 2, Paper 29.
8.6 Chaudhry, M. H. (1991), 'Mathematical Model of Transient Flows

with Column Separation', 9th Round Table of IAHR Group on Hy
draulic Transients with Water Column Separation, Valencia, Septem
ber, pp. 211-220.
8.7 Chen, Y. W. and Whippen, W. G. (1978), "Turbine Runaway,

Hysteresis and Thoma's Sigma Characteristics of Francis Pump-
Turbines', 9th IAHR Symposium on Hydraulic Machinery, Fort
Collins, June 12-14, Vol. 2, pp 402.- 412.
8.8 Dahlbeck, N. (1986), 'Detailed Studies of Pressure Transients at Load

Rejection of a Water Power Plant', Fifth BHRA International Confer
ence on Pressure Surges, Hanover, West Germany, September 22-24,
Paper C4, pp. 73-79.
8.9 D'Amorin, P. M. (1989), 'Occurrence of Draft-Tube Waterhammer

in Francis Turbine of Paulo Afonso Power Station - Its Causes and
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

446 Cavitation
Adopted Solutions', International Congress on Causes and Accidents

in Fluid Systems, Sao Paulo, Brazil, Vol. 1, March 6-10, pp. 160-173.
8.10 Deriaz, P. (1960), 'A Contribution to the Understanding of Flow in

Draft-Tubes of Francis Turbines', 1st IAHR Symposium on Hydraulic
Machinery, Nice, Paper B-l.
8.11 Doerfler, P. (1980), 'Mathematical Model of the Pulsations in Francis

Turbines Caused by the Vortex Core at Part Load', Escher Wyss
News, Vol. -, No. 112, pp. 101 - 106.
8.12 Doerfler, P., (1985), 'Francis Turbine Surge Prediction and Preven
tion', Waterpower 85 -International Conference on Hydropower.
8.13 Dziallas, R. (1964), 'Francisturbinen bei Teil- und Uberlast', VDI

Bericht, No. 75, pp. 53-64.
8.14 Eichler, O. and Jaeger, E. U. (1979), 'The Assessment of the Cavi

tation Behaviour of Kaplan Turbines on the Basis of a Comparison
between Model Tests and Field Experience', Voith Research and Con
struction, Vol 25e, Paper 4.
8.15 Falvey, H. T. (1993), 'A Primer on Draft-Tube Surging', Hydro Re

view, Vol.12, No. 1, February, pp. 76-86.
8.16 Fanelli, M. (1988), 'Mathematical Models of the Vortex Rope Surge

Effects in a Draft-Tube of a Francis Turbine Working at Partial Load',
14th IAHR Symposium on Hydraulic Machinery, Trondheim, June
20-23, Vol. 1, Paper F l , pp. 341-354.
8.17 Fisher, R. K., Palde, U., and Ulith, P. (1980), 'Comparison of Draft-
Tube Surging of Homologous Scale Models and Prototype Francis
Turbines', 10th IAHR Symposium on Hydraulic Machinery, Tokyo,
pp 541-556.
8.18 Fisher, R. K., and Cybularz, J. M. (1985), 'Characteristics of Axial

Turbines in Cavitating Regimes', ASME Symposium on Cavitation in
Hydraulic Structures and Turbomachinery, Joint ASCE/ASME Me
chanics Conference, Albuquerque, June 24-26, pp. 119-127.
8.19 Gajic, A., Pejovic, S., Arautovic, D., and Ignjatovic, B. (1992), 'Re
verse Waterhammer Analysis in Kaplan Turbines', 16th IAHR Sym-
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

posium on Hydraulic Machinery, Sao Paulo, September, Vol. 1, pp.

161-171.
8.20 Graeser, J. E., Prenat, J. E., and Walther, W. (1982), 'Caracteris-

tiques de Machines Hydrauliques et de Vannes dans les Calculs de
Regimes Transitoires', 11th IAHR Symposium on Hydraulic Machin
ery, Amsterdam, 1982, Vol. 2, Paper 31.
8.21 Grein, H. (1974), 'Partial Pressure and Gas Content Inside a Cav-
itating Vortex Core and Influence on Pressure Fluctuations in The
Draft-Tube of a Francis Turbine', Proceedings, Institution Mechanical
Engineers, London, Vol. C182, pp. 313-319.
8.22 Grein, H. (1980), 'Vibration Phenomena in Francis Turbines: Their

Causes and Prevention', 10th IAHR Symposium on Hydraulic Ma
chinery, Tokyo, pp 527-539.
8.23 Grolmes, M. A., and Fauske, H. K. (1969), 'Propagation Charac

teristics of Compression and Rarefaction Pressure Pulses in One-
Component Vapour-Liquid Mixtures', Nuclear Engineering and De
sign, Vol. 11, pp. 137-142.
8.24 Henry, R.E., Grolmes, M.A., and Fauske, H.K. (1969), 'Propagation
Velocity of Pressure Waves in Gas-Liquid Mixtures', Cocurrent Gas-
Liquid Flow, Edited by E. Rhodes and D.S. Scott.
8.25 Hsu, Y. Y., and Graham, R. W. (1976), Transport Processes in Boil

ing and Two-Phase Systems, McGraw-Hill.
8.26 Ishii, Y. (1962), 'The Two Types of Runaway State of Propeller Tur
bine under Cavitation', 2nd IAHR Symposium on Hydraulic Machin
ery, Sendai, Paper B-5, pp. 247-263.
8.27 Jordan, V. (1975), 'Reverse Waterhammer in Turbine Draft-Tubes',

Water Power and Dam Construction, Vol. 55, Part One, February,
pp. 75-80, Part Two, March, pp. 116 - 118.
8.28 Kobori, T., Yokoyama, S., and Miyashiro, H., (1955), 'Propagation
Velocity of Pressure Wave in Pipe Line', Hitachi Hyoron, Vol. 37, No.
10, October , pp. 1407-1411 (in Japanese).
8.29 Lecher, W. (1960), 'Cavitation Research on Water Turbines', Escher

Wyss News, Vol. 23, No. 1, pp. 41-48.
--``,```,,``````,,,,``,`

448 Cavitation
8.30 Li, S. C. (1992), 'Pressure Fluctuations in Cavitating Draft-Tube

Flows', ASME Symposium on Hydropower Fluid Machinery, Winter
Annual Meeting, Anaheim, November 8-13, pp. 1-6.
8.31 Martin, C. S., (1993), 'Pressure Wave Propagation in Two-

Component Flows', NATO Advanced Study Institute on Computer
Modeling of Free Surface and Pressurized Flows, Washington State
University, Pullman, Washington, June 28 - July 9, 1993, Edited by
M. H. Chaudhry and L. W. Mays, NATO Series E: Applied Sciences,
Vol 274, Kluwer Academic Publishers, pp. 519-552.
8.32 Matsumoto, Y., Ohashi, H., Nonoshita, T., and Kubota, T. (1986),
"Transient Characteristics of Swirl Flow in Draft-Tube', 13th IAHR
Symposium on Hydraulic Machinery, Montreal, September, Vol. 1,
Paper 36.
8.33 Muciaccia, F. F., Rossi, G., and Zanetti, V. (1984), 'Study on Model
of the Dynamic Behavior of a Hydraulic Machine', 12th IAHR Sym
posium on Hydraulic Machinery, Stirling, August 2730, Vol. 1, Paper
3.10, pp. 341-355.
8.34 Nishi, N., Kubota, T., Matsunaga, S., and Senoo, Y. (1980), 'Study
on Swirl Flow and Surge in an Elbow Type Draft-Tube', 10th IAHR
Symposium on Hydraulic Machinery, Tokyo, pp 557-568.
8.35 Nishi, N., Matsunaga, S., Kubota, T., and Senoo, Y. (1982), 'Flow
Regimes in an Elbow-Type Draft-Tube', 11th IAHR Symposium on
Hydraulic Machinery, Amsterdam, Vol. 2, Paper 38.
8.36 Nishi, N., Matsunaga, S., Kubota, T., and Senoo, Y. (1984), 'Surging
Characteristics of Conical and Elbow-Type Draft-Tubes', 12th IAHR
Symposium on Hydraulic Machinery, Stirling, Paper 3.6, pp 272-283.
8.37 Nishi, M., Matsunaga, S., Kubota, T., and Senoo, Y. (1986), 'Effect of
Draft-Tube Shape on the Characteristics of Pressure Surge and Swirl
Flow', 13th IAHR Symposium on Hydraulic Machinery, Montreal,
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
September, Vol. 1, Paper 7.
8.38 Nonoshita, T., Matsumoto, Y., Ohashi, H., and Kubota, T. (1988),
'Transient Phenomena in a Draft-Tube during a Load Rejection', 14th
IAHR Symposium on Hydraulic Machinery, Trondheim, June, Vol. 1,
Paper F3, pp. 365-376.

8.39 Nonoshita, T., Matsumoto, Y., Ohashi, H., Kubota, T., and Fujita,
M. (1990), 'Effect of a Swirl Flow on Transient Phenomena in a Draft-
Tube during Load Rejection', 15th IAHR Symposium on Hydraulic
Machinery , Belgrade, September, Vol. 1, Paper J5.
8.40 Nonoshita, T., Matsumoto, Y., Ohashi, H., and Kubota, T. (1991),
'Model Analysis of Water Column Separation Accompanied with
Swirl Flow', 9th Round Table of IAHR Group on Hydraulic Transients
with Water Column Separation, Valencia, September, pp. 235-249.
8.41 Nonoshita, T., Matsumoto, Y., Ohashi, H., and Kubota, T., (1992),
'Effect of a Swirl Flow on Transient Phenomena in a Draft-Tube dur
ing Load Rejection', 16th IAHR Symposium on Hydraulic Machinery,
Sao Paulo, September, Vol. 1, pp. 141-150.
8.42 Pejovic, S., Krsmanovic, Lj., Gajic, A., and Obradovic, D. (1980a),
'Kaplan Turbine Accidents and Reverse Water Hammer', Third
BHRA International Conference on Pressure Surges, Canterbury,
March 12-15, Paper H2, pp. 391-400.
8.43 Pejovic, S., Krsmanovic, Lj., Gajic, A., and Obradovic, D. (1980b),
'Kaplan Turbine Incidents and Reverse Water Hammer', Water
Power and Dam Construction, Vol. 56, August, pp. 36-40.
8.44 Pejovic, S., Gajic, A., and Obradovic, D. (1980c), 'Reverse Water-
hammer in Kaplan Turbines', 10th IAHR Symposium on Hydraulic
Machinery, Tokyo, pp. 489-499.
8.45 Pejovic, S., Gajic, A., and Obradovic, D. (1984), 'The Effects of
Air Supply to the Draft-Tube upon Hydraulic Oscillations in a Hy
dropower Plant', 12th IAHR Symposium on Hydraulic Machinery,
Stirling, August 27-30, Vol. 1, Paper 3. 4, pp. 242-253.
8.46 Pejovic, S., Obradovic, D., and Gajic, A. (1986a), 'Field Tests and
Calculations of the Effects of Air Introduction into the Draft-Tube
upon Hydraulic Oscillations in a Hydropower Plant', 13th IAHR
Symposium on Hydraulic Machinery, Montreal, September 2-5, Vol.
1, Paper 9.
8.47 Pejovic, S. (1986b), 'Pressure Surges and Vibrations in Hydropower

Plants - Experiences in Yugoslavia', Current State of Technology in
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

450 Cavitation
Hydraulic Machinery, Edited by A. P. Boldy and D. C. Guo, Inaugu

ral Meeting of the International Editorial Committee for Book Series
on Hydraulic Machinery, Beijing, July, Paper 10, pp. 177-204.
8.48 Raabe, J. (1985), Hydropower - The Design, Use, and Function of

Hydromechanical, Hydraulic and Electrical Equipment, VDI Verlag.
8.49 Rheingans, W. K. (1940), 'Power Swings in Hydroelectric Power

Plants', TRANS ASME, Vol. 62, April, pp. 171-184.
8.50 Rongqiao, W. (1986), 'Analysis of Reverse Waterhammer in Draft-

Tubes of Turbines', ASME International Symposium on Fluid Ma
chinery Troubleshooting, Winter Annual Meeting, Anaheim, Decem
ber 7-12, FED-Vol. 46, PWR-Vol. 2, pp. 75 - 82.
8.51 Ruggles, A. E., Scarton, H. A., and Leahy, R. Y. (1989), 'Relationship

between Standing Waves, Pressure Pulse Propagation, and Critical
Flow Rate in Two-Phase Mixtures', TRANS ASME, Journal of Heat
Transfer, Vol. I l l , No. 2, May , pp. 467-473.
8.52 Ulith, P., Jaeger, E. U., and Strscheletzky, M. (1974), 'Contribution

to Clarifying the Inception of Nonstationary Flow Phenomena in the
Draft-Tube of High Specific Speed Francis Turbines Operating at
Part Load', 7th IAHR Symposium on Hydraulic Machinery, Vienna,
September, Paper III-4.
8.53 Watanabe, T., and Mutaguchi, K. (1979), 'Research of Hydraulic

Transient Phenomena in Pump Turbine', Pump Turbine Schemes -
Planning, Design, and Operation, Joint ASME-CSME Conference on
Applied Mechanics, Fluids Engineering and Bioengineering Confer
ence, Niagara Falls. New York, June 18-20, pp. 57-65.
8.54 Wegner, M. and Roman, J.M. (1984), ' Control of Erosion and Insta
bility Effects Induced by Cavitation in Francis Turbines', 12th IAHR
Symposium on Hydraulic Machinery, Stirling, August 27-30, Vol. 1,
Paper 15, pp. 77-89.
8.55 Wiggert, D. C , and Sundquist, M. J., (1979), 'The Effect of Gaseous

Cavitation on Fluid Transients', ASME Transactions, Journal of Flu
ids Engineering, March, Vol. 101, pp. 79-86.
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

P a r t 2 : P u m p Transients
H. Tsukamoto
8.7 Introduction
Transients are generated in pumping system by pump start, stop or speed
change, or by valve-opening change. During such a process, a transient
cavitation will occur in the system if the fluid pressure drops to or below
the vapour pressure in the system. The most serious type of these possible
cavitation transients is the water-column separation which occurs in pump
discharge lines owing to the negative pressure caused by machine transients,
for example a power failure to the pumps (Wylie and Streeter in 1978 [8.69];
Chaudhry in 1987 [8.56]). Many studies on the transient cavitation occurring
in liquid-filled pump discharge lines have been reported (Fan and Tijsseling
in 1992 [8.57]).
Another problem is the transient performance of the pump involved. The
pump may operate away from its normal head-capacity and speed range
during a transient period while its operating point changes from one steady-
state to another. In some particular circumstances, the pump may even
operate as a turbine or a pump rotating backwards. However, little is known
about such transient performance of cavitating pumps.
Usually attention has been confined to the transients in the pump dis
charge line, because the suction pipes are usually not long enough to induce
any dangerous pressure rise. However, in this section, the transient cavita
tion phenomena occurring in the suction lines will also be described.
8.8 Transient Cavitation in Discharge Lines

8.8.1 Transient Cavitation Type
Transient cavitation frequently occurs in liquid-filled pipelines during un
steady flow processes. Two types of transient cavitation can be distin
guished: cavitating flow and water column separation (Kranenburg in 1974
[8.60]).
Cavitating flow, some times called gas release, is a type of bubbly flow
which can extend over quite a considerable length in the pipeline. These
dispersed gaseous bubbles usually possess a small void fraction. Vapour
--``,```,,``````,,,,``,`,,`,`,`
451
452 Cavitation
_ \ ^.Wottr eotwm uporott
/Upptr nlmnn
Wottr MlaiMis rtjoim i
(a)
_l_l I I I I I I I I ■ ' J l_
iwc III MTCRwu or ^ » F I H rowed nn.uRC
I L_ i I I I I I I I
tare ia $ECO«M »"£« POW:» ruLun
(b)
" I l 3 4 J « 7
TIUC IN IHTERWLS OF ^ AFTER POWER FAILURE
I I I I J I I I I I L
o i t > 4 » « r * t i o
T l « III SECONDS AFTER POffER FAILURE
Figure 8.22: Water column separation caused by a power failure at pump

motor; L/a: wave travel time of discharge line, Vo: velocity of water in
discharge line for initial steady conditions (Parmakian in 1958 [8.65])
(a) Water column movements; (b) Time histories of water columns.
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

Unsteady Cavitation Flows: Pump Transients 453
bubbles are also dispersed within the liquid. Such a gaseous cavitation is a
slow process, in the order of one to several seconds.
Column separation is a vaporous cavitation occurring in a liquid pipeline,
where the bubble population is sufficiently large to coalesce, forming a vapour
pocket within the pipe. The process of vaporous cavitation is much quicker
taking only a few microseconds. Figure 8.22 illustrates the water column
separation at the pump discharge line due to a power failure to the pump.
Figure 8.22(a) shows the movement of the upper and lower columns; and,
Figure 8.22(b) shows the corresponding process of water column separation
(Parmakian in 1958 [8.65]). When these two separated water columns rejoin,
a high pressure-rising (shock wave nature) will be generated which may be
strong enough to destroy the pipeline.
8.8.2 Water Column Separation

The transient phenomena in the discharge line due to a power failure to the
pump is shown in Fig. 8.23 (Richards in 1956 [8.66]). Figure 8.23(b) presents
the time histories of pressure at the valve house No. 8, which falls rapidly
to vapour pressure when a void starts developing (as long as as the pressure
remains at vapour pressure), then rises sharply (due to the void collapsing)
followed by an oscillation (due to the repeated parting and joining of the
water columns). About 64 sec after the trip-out, the first vaporous volume
collapses, producing a maximum surge of 75 psig ( = 517 kPa). Normal
operating pressure is minus 9 in Hg ( = 30.5 kPa). In this case, water
column separations occurred at all of the nine measuring points after the
pump trip-out.
8.9 Transient Cavitating Flow in Turbopumps

8.9.1 True Total Pressure Rise
For transient calculation, the meaning of the true total pressure rise, P,
generated solely by pumping action must be clarified. In steady state oper
ation, it is equal to the total pressure increase, Pi, from the pump suction
to discharge. However, when the flow rate is unsteady, a part of the total
pressure difference measured across the pump is attributed to the inertia
effect of the water contained in the pump impeller and casing. This part
is called as the apparent total pressure rise, Pc, (Ohashi in 1968 [8.64]). In
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

454 Cavitation
V . -oasirwi
tar mniy j ^
(a)
aiiMJii • u m i 11
MAX 7t M K ( t n r»r)
f"T«lfBUT
y— aiHOMHcme r t m u t i
N_«iiiJIITtie LIMIT
i i ■
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
TIME I* H I I I T t l
(b)
Figure 8.23: Variation of line pressure caused by power failure to pump

(Richards in 1956 [8.66]).
(a) Profile of water pumping plant: (b) Surges at valve house No. 8 following
pump trip-out.

cavitating transients, it can be calculated as
Pc(t) = -{{pLegs/A0)dQs{t)/dt + {PLeqd/A0)dQd(t)/dt} (8.2)
where the pump is represented by a straight pipe with the reference cross
sectional area AQ, length (Leqs + Leqd). The equivalent pipe length Leqs and
Leqd can be obtained as follows,
Leqa= / {A0/A{s)}-ds (8.3)

Js=Q
Leqd = [ {A0/A(s)} ■ ds (8.4)

J s=L\
where s is the distance measured from the suction port, and L is the total
path length (Yamamoto in 1990 [8.70]). Thus, the true total pressure rise
(solely attributed to pumping action), P, is
P = Pi-Pc (8.5)
8.9.2 Transient B e h a v i o u r of a C a v i t a t i n g P u m p
As an example, the transient behaviour of a cavitating centrifugal pump after
the sudden opening and closure of a discharge valve is presented here. The
variations of instantaneous rotational speed, suction and discharge pressures,
and flow rate are shown (Fig.8.24) where a cavitating transient is generated
by such a transient process (Tsukamoto and Tanaka in 1994 [8.68]). It shows
the variations of the measured suction and discharge pressures, ps and pd;
the suction and discharge flow rates, Qs and Qd; the true total head rise Hp
(= P/(pg)); and the apparent total head rise Hc (= Pc/(pg)) at the highest
valve opening/closing rates with a low suction-pressure of a — 0.085 . Here,
a = 2gHsv/{u\) is the cavitation coefficient, and Hsv = Hs — pv/(pg) (ie
NPSH), Hs: the total head at pump suction port.
For a cavitating transient process, the flow-rate difference between the
inlet and the discharge is caused by the growth and shrinkage of the total
volume of cavitation within the pump, while for steady-state operation, the
flow-rate difference is zero. The mass conservation equation in terms of the
changing rate of cavitation volume, i), reduces to (Sack and Nottage in 1965
[8.67]; Greitzer in 1981 [8.58]):
v = Qd~Qs (8.6)
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

456 Cavitation
NorKllnraiwkMtalHm* t/T Nondlmenslonal d m * 1/T

--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
(a) (b)
Figure 8.24: Time histories of Qs, Qd,Ps,Pd,Hp,Hc, and v; N0 = mean rota

tional speed, T = valve opening/closure time, subscripts N = noncavitating
case, 0 = initial value, / = final value (Tsukamoto and Tanaka in 1994
[8.68]).
(a) Quick valve opening case at NT = 2.99, af = 0.085 and $Nf = 0.107;
(b) Quick valve closure case at NT = 2.99, a0 = 0.085 and $NO = 0.107.

The variation of v is also shown in Fig.8.24.

In the case of quick valve opening, the discharge pressure and flow rate at
the discharge show a slow fluctuation before reaching the final steady-state
(Fig.8.24(a)). The time interval required for one full fluctuation decreases
with time and as does the cavitation volume. These pressure and flow-
rate oscillations are mainly attributed to the growth and shrinkage of the
total volume of cavitation within the pump. Because the pressure-wave
propagation in a liquid-filled pipeline can be greatly reduced if gas bubbles
are present in the liquid (Wylie and Streeter in 1978 [8.69] and Chaudhry
in 1987 [8.56]), the small air/vapour bubbles produced by cavitation lead to
a reduction in wave speed and consequently an increase in the time interval
for one full pressure-fluctuation.
On the other hand, the valve closure case shows a different phenomenon.
That is, there are suction and discharge pressure fluctuations after reaching
zero flow rate as shown in Fig.8.24(b). The time interval required for the
discharge pressure fluctuation coincides with that for the suction pressure
fluctuation, which suggests that the water hammer is accompanied by water
column separation.
8.9.3 Transient Characteristics of Pump

In many studies of pump systems which involve water column separation
or cavitating flow, it is assumed that the transient process is slow enough
for the pump to respond quasi-statically, ie. moving from one steady-state
operating-point to another along its steady performance curve. The validity
of such a quasi-steady characteristics assumption is discussed here. A com
parison of the true total pressure rise, P, with its counterpart quasi-steady
value is made in order to see how the real dynamic performance deviates from
its quasi-steady assumption. Figures 8.25 (valve opening) and 8.26 (valve
closing) indicate the measured loci, along which the coordinates of the in
stantaneous pump discharge-flow-coefficient <f> and total-head-rise-coefficient
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
ip move during the transient process. These two coefficients are defined as
follows,
<t> = Qd/{*d2b2U2) (8.7)

V> = 2gHp(t)/(ul) (8.8)
where, 62 is the outer passage width of pump impeller; cfe is the outer di
ameter of impeller; U2 is the peripheral speed of impeller (= nd^N). The

458 Cavitation
$ flow coefficient
(c)
Figure 8.25: Dynamic relationship between discharge flow coefficient <f> and
total head rise coefficient ip at rapid opening of discharge valve (Tsukamoto
and Tanaka in 1994 [8.68]).
(a) Cavitating case at NT = 2.99, of = 0.085 and $Nf = 0.107; (b) Non-
cavitating case at NT = 2.99, CT/ = 0.3 and $Nf = 0.093; (c) Cavitating
but slow valve opening case at NT — 17.3, aj — 0.085 and <!>#/ = 0.107.
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

Figure 8.26: Dynamic relationship between discharge flow coefficient <f> and
total head rise coefficient ip at rapid closure of discharge valve (Tsukamoto
and Tanaka in 1994 [8.68]).
(a) Cavitating case at NT = 2.99, a0 = 0.085 and $N0 = 0.107; (b) Non-
cavitating case at NT — 2.99, <7o = 0.3 and $NO = 0.093; (c) Cavitating
but slow valve closure case at NT = 17.3,CTQ= 0.085 and $m = 0.107.
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

460 Cavitation
measured steady-state characteristics curves for ajfi = 0.085 are indicated

by a dot-dash-line. If the transient process is sufficiently slow, the instanta
neous pressure rise of the pump moves closely along the steady-state 4> — ips
curve, which may be referred as a quasi-steady case. As can be seen clearly
from these figures, the actual deviations of transient characteristics from the
quasi-steady ones are significant as reported by Kaneko and Ohashi in 1982
[8.59]. In noncavitating cases (Figs.8.25(b) and 8.26(b)), the total head rise
(ipi) for quick valve opening is greater than the quasi-steady value (tps) and
is smaller for quick valve closing. In cavitating cases, the transient total
head rise coefficient (tpi) is greater than V>s during most time of the transient
process.
As can be seen from these figures, the deviation of transient character
istics from the quasi-steady ones is significant if a sudden change of pump
operating conditions is involved. However, the application limit of the quasi-
steady assumption is still not very clear. Nevertheless, it is practically ac
ceptable for most calculations provided the rate of change between the oper
ating points is reasonably slow (Nguyen Due et al in 1993 [8.63] and Lefebvre
and Barker in 1994 [8.61]).
8.10 Computer Simulation of Cavitating Tran

sient Flows in P u m p System
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
In order to simulate the transient flows in both the pump and the discharge
line, three effects must be considered; namely the pump and motor inertia,
the pump characteristics, and the pressure wave behaviour in the discharge
line (Miyashiro in 1967 [8.62]). The pump transient characteristics are usu
ally assumed to be quasi-steady as described above. As to the transient
cavitation in pipeline, there are many unknowns such as the variation of
the void fraction along the pipeline and consequently the variation of the
wave speed, the energy dissipation, gas release, and so on. Many approaches
for such transient cavitating flow have been reported (Wylie and Streeter in
1978 [8.69]; Chaudhry, 1987 [8.56]). The partial differential equations gov
erning the unsteady fluid flow in a pipeline may be transformed into a set
of finite difference equations which can be solved numerically. For details,
readers are referred to Streeter and Wylie in 1978 [8.69]. As an example, Fig
ure 8.27 shows the calculated pressure and velocity variations for the cases
of pump trip-out and restart which causes cavitating flows in the horizon
tal pipeline (Kranenburg in 1974 [8.60]). In this example, a large reservoir

provides a constant downstream pressure. After pump failure, a negative

pressure wave is seen to propagate downstream; and, in certain part of the
pipeline the pressure reduces down to the vapour pressure as shown by the
low constant pressures which indicate the existence of cavitating flows there.
After a certain time, this cavitating region is nullified by the positive pres
sure waves travelling from the ends of the pipeline. The calculated values
(dash lines) agree well with the measured ones (solid lines).
hlo> prttsvrt upttrum : MS ItN/^t

low prtfiur* uptb-tim : 101 t«N/_i
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
Figure 8.27: Simulation of pump failure causing cavitating flow in horizontal

pipeline ; length — 1,450m, diameter = 0.10m, u\ = upstream velocity, U2=
downstream velocity, £1,2,3,4,5 = pressures at 3m, 299m, 597m, 887m, and
1,168m from the upstream end ( Kranenburg in 1974 [8.60]).
8.11 Concluding Remarks

Although further work is needed to establish a more complete theoreti
cal model of the cavitating transient in pipe flow is needed, most prac
tical cases of cavitating transients in pump discharge lines can be pre
dicted, to a reasonable degree of accuracy, by the existing approach. As
to the transient characteristics of a cavitating pump used for some extreme
conditions such as liquid propulsion, further investigation into the pump-

462 Cavitation
cavitating-dynamic-performance is required so that a better prediction for

these transient-cavitating-pumping-systems can be achieved.
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

References
8.56 Chaudhry, M. H., (1987), Applied Hydraulic Transients, Second Edi
tion, Van Nostrand Reinhold, New York.
8.57 Fan, D., and Tijsseling, A., (1992), 'Fluid-Structure Interaction With
Cavitation in Transient Pipe Flows', ASME Journal of Fluids Engi
neering, Vol. 114, pp.268-274.
8.58 Greitzer, E.M., (1981), 'The Stability of Pumping Systems,' ASME

Journal of Fluids Engineering, Vol. 103, pp. 193-242.
8.59 Kaneko, M., and Ohashi, H., (1982), "Transient Characteristics of a
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
Centrifugal Pump During Quick Change of Flow Rate,' Trans. JSME,
Ser.B, Vol.48, No.426, pp.229-237 (in Japanese).
8.60 Kranenburg, C , (1974), 'Gas Release During Transient Cavitation in

Pipes,' Journal of the Hydraulic Division, ASCE, Vol.100, No. HY10,
pp.1383-1398.
8.61 Lefebvre, P. J., and Barker, W. P., (1994), 'Cavitation in Transiently

Operating Centrifugal Pump,' ASME FED Vol.190, pp.11-17.
8.62 Miyashiro, H., (1967), 'Waterhammer Analysis of Pump Discharge

Line With Several One-Way Surge Tanks,' ASME, Journal of Engi
neering for Power, Vol. 89, pp.621-627.
8.63 Nguyen Due, J.M., Von Kaenel, A., and Danguy, F., (1993), "Tran
sient Behavior of Liquid Hydrogen Pumps During Start-Up and Shut
down of Rocket Engines', ASME FED, Vol.154, pp.159-170.
8.64 Ohashi, H., (1968), 'Analytical and Experimental Study of Dynamic

Characteristics of Turbopumps,' NASA TN D-4298.
8.65 Parmakian, J., (1958), 'One-Way Surge Tanks for Pumping Plants',
Trans. ASME, Vol. 80, pp.1563-1573.
8.66 Richards, R.T., (1956), 'Water-Column Separation in Pump Dis

charge Lines', Trans. ASME, pp. 1297-1304.
8.67 Sack, L.E., and Nottage, H.B., (1965), 'System Oscillations Associ
ated With Cavitating Inducers,' ASME Journal of Basic Engineering,
Vol. 87, pp.917-924.

464 Cavitation
8.68 Tsukamoto, H., and Tanaka, T., (1994), "Transient Characteristics

of a Cavitating Centrifugal Pump System at Rapid Change in Flow
Rate,' ASME FED, Vol.190, pp.175-183.
8.69 Wylie, E.B., and Streeter, V.L., (1978), Fluid Transients, McGraw-
Hill, New York.
8.70 Yamamoto, K., (1990), 'Instability in a Cavitating Centrifugal Pump
(2nd Report, Delivery of Mechanical Energy during Oscillation)',
Trans. JSME, Ser. B, Vol.56, No.523, pp.644-650 (in Japanese).
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---

Cavitation f Hydraulic M<achinery
iinor\/
inery Rnnlf
This volume ini tthe Hydraulic Machiner Qorioc
Book Series deal with cavitation
ies deals cavitati and
jrbines and pumps. After
its effects in tur Afte introducing caviti ion and its rrelation
ig cavitation
machines, the invited
with hydraulic n through ut the world review
contributors throughout
>dc<
ts from
mt cavitation subjects
in detail relevan fn pf nomena to \various
ental phenomena
fundamental
nachines.;. The authors
problems andI isolution measures: in hydraulic machine authc are
internationally recognized
r experts in their
tr fields.
--``,```,,``````,,,,``,`,,`,`,``-`-`,,`,,`,`,,`---
MBlifl
mi I I I I I I II minium mi ■
ico.uk
www.i
depress. |9''781860''94257d'|

Cavitation of Hydraulic Machinery

Încărcat de

Informații document

Drepturi de autor

Formate disponibile

Partajați acest document

Partajați sau inserați document

Opțiuni de partajare

Vi se pare util acest document?

Este necorespunzător acest conținut?

Drepturi de autor:

Formate disponibile

Cavitation of Hydraulic Machinery

Încărcat de

Drepturi de autor:

Formate disponibile

SERIES ON HYDRAULIC MACHIWEW - VOL.

Copyright World Scientific Publishing Co.

- ImperiaI College Press

Copyright World Scientific Publishing Co.

Prof H Netsch (Canada) Prof H Ohashi (Japan)

British Library Cataloguing-ln-Publlcation Data